How to start an SEO career in 2026: a 90 day plan

The short version

6 things to remember

01SEO work in 2026 is still mostly about helping search engines find, understand and trust a site, then proving what that work did for the business. Google's AI Overviews and AI Mode run on its core Search systems, and Bing applies the same foundation to Copilot, so the fundamentals did not move.
02AI tools have made drafting, formulas, scripts and summaries cheap. Judgement, diagnosis, prioritisation, stakeholder work and measurement are still scarce, and they are what employers pay for.
03Learn from primary material first: How Search Works, the SEO Starter Guide, Search Essentials and the spam policies, then the Search Console and GA4 help centres. Blogs and courses are easier to judge once you know what the providers actually say.
04A small site you built, grew and measured yourself, with Search Console screenshots and honest notes, proves more than a certificate. Add a respectful written audit, a technical write-up and a content brief.
05In interviews and pitches, report results as absolute numbers with dates and context, and say what you cannot attribute. Inflated claims are easy to spot and hard to recover from.
06Much of the value in a first SEO job comes from saying what is not worth doing. Search engines document plenty of things they ignore, and pointing to those pages saves teams weeks.

Before you start

6 things to have ready

01Five to ten hours a week for 12 weeks, blocked out in your calendar.
02A Google account for Search Console and Google Analytics, and a Microsoft account for Bing Webmaster Tools.
03A subject you know well enough to write about from experience, for your lab site.
04A spreadsheet tool you are comfortable in, such as Google Sheets or Excel.
05A small budget for a domain name, and for hosting if you build with WordPress. A GitHub account if you plan to build a static site with an AI coding tool.
06A plain text notes file or document for a learning log, kept from day one.

SEO is the work of making a website easy for search engines to find, understand and show, and then making sure that visibility turns into something the business wants: sales, enquiries, sign-ups or readers. Day to day that means reading data in Search Console and analytics, working out why pages do or do not appear, writing and editing, briefing developers, and explaining to people who are not SEOs what is worth doing and what is not. What an SEO can help with includes reviewing content and structure, technical advice on development, content development, training, and optimising for generative AI.¹

The job suits people who like finding out why something happened, who can write clearly, and who are comfortable with slow feedback. Some changes take effect in hours and others take several months, and it is worth waiting a few weeks before judging whether work helped.² If you need quick, certain results to stay motivated, or you dislike explaining your reasoning to sceptical colleagues, you will find it frustrating. If you want to be told exactly what to do by a checklist, the job has fewer of those than it used to.

This tutorial runs over roughly 90 days of part-time study, then the job hunt and your first year. Stage one explains the job and the routes into it. Stages two and three cover the fundamentals, in order, from Google's and Bing's own documentation. Stage four has you build and run a small site as a lab. Stage five turns that work into a portfolio. Stage six is getting hired or finding a first client, and stage seven is how to do the first year well and keep current. A hypothetical career changer called Dan, who spent four years as a marketing coordinator and restores secondhand bikes at the weekend, is the worked example throughout.

Diagram

The 90 day plan, then the job hunt

Week

135791113151719

Baseline skills scoreLab site liveSecond skills score

Week 1

Understand the job, choose a route

Days 1 to 3: the job, the specialisms, what AI tools changed, and the decision tool.

A part-time pace of five to ten hours a week. If the skills scorecard in stage two puts you in the lowest band, stretch the fundamentals and everything moves right by two weeks.

Stage 01 / 07Days 1 to 3

Understand the job and choose a route in

Know what SEO work involves in 2026, which parts AI tools now handle, and which route into the work fits your background and circumstances.

What changed with AI Overviews and assistants, and what did not

Search results now include AI-written answers. Google shows AI Overviews and has AI Mode; ChatGPT, Perplexity and Microsoft Copilot answer questions with cited links. It is easy to read this as the end of SEO. Google's own guide to its generative AI features says the opposite: SEO best practices remain relevant because those features are rooted in its core Search ranking and quality systems, and they retrieve pages from the Search index to ground their answers.³ Bing's guidelines say the same of Copilot, which relies on the same crawling, indexing and ranking foundation as traditional Bing search.⁴ There are no additional requirements to appear in AI Overviews or AI Mode beyond being indexed and eligible to show with a snippet, and clicks and impressions from those features are counted in Search Console's Performance report under the Web search type.⁵

From Google's perspective, AEO (answer engine optimisation) and GEO (generative engine optimisation) are still SEO.³ Bing uses GEO for the work of being eligible to be cited in AI answers, and is clear that neither SEO nor GEO guarantees rankings or citations.⁴ Other providers run their own crawlers and indexes, so there is real extra work in understanding how OpenAI, Perplexity and Microsoft find and cite pages, and that work has become a specialism. The core skills underneath it are the same ones this tutorial teaches. For a fuller picture, see the AI search guide.

What has changed is how the work is measured. Fewer searches may end in a click. Pew Research Center tracked the Google searches of 900 US adults in March 2025 and found that people clicked a traditional result in 8% of visits when an AI summary appeared, against 15% when none appeared, and clicked a link inside the summary in 1% of visits.⁶ The study measured browsing behaviour in the US, not rankings. More visibility now happens inside answers, and some assistants publish no data at all. Being able to say honestly what you can and cannot measure is now part of the job, and stage four covers the reports that exist.

Where SEOs work

Setting	What the work is like	Good for a beginner because	Harder because
In-house	One business, one site or a few. You own the roadmap, compete for developer time and report to people outside marketing.	You see the effect of work over months and learn how decisions get made in a business.	You may be the only SEO, with nobody to check your work or teach you.
Agency	Several clients at once, often in different sectors. Audits, monthly reports, recommendations that someone else implements.	Variety and volume. You see many sites and many problems in a year, usually with seniors to learn from.	Less control over whether recommendations get implemented. Time is billed, so work is rationed.
Freelance	You find clients, scope work, deliver it and chase invoices. Often small businesses.	Freedom to choose clients and sectors.	No salary, no colleagues, and you have to sell before you can do any SEO. Hard without experience to point to.

The three main settings for SEO work, and what each is like to start in

Diagram

In-house, agency and freelance compared for a first SEO job

	Steady income	Training and review	Variety of sites	Seeing work through	Control over choices	Easy to start with no track record
In-houseTraining depends on whether there is another SEO on the team. Many in-house SEOs are on their own.
AgencyThe usual entry point for people with no SEO history, because agencies hire juniors and review their work.
FreelanceFull control and no salary. Clients expect you to know more than they do from day one.

Weak Strong Best in the row

My judgement of a typical first role in each setting. Individual employers vary a lot, so ask about these points in interviews.

The specialisms

Most people start as generalists and specialise after a year or two, once they know which problems they enjoy. It still helps to know what the specialisms are, because job adverts use the names and each has its own primary reading.

Specialism	What the work involves	Read further
Technical SEO	Crawling, indexing, rendering, site architecture, redirects, canonicals, migrations, log files and working closely with developers.	Website migrations, JavaScript sites
Content SEO	Research into what people search for and why, briefs, editing, content plans, updating and pruning existing pages.	When more content is not the answer
Local SEO	Google Business Profiles, reviews, location pages and service areas for businesses that serve a place.	Local SEO
Ecommerce SEO	Category and product pages, faceted navigation, product data and feeds, variants and stock.	Ecommerce SEO, Faceted navigation
International SEO	Several countries or languages: URL structure, hreflang, localisation.	International SEO, Auditing hreflang
Digital PR	Earning coverage and links from journalists and publications with stories, data and expert comment.	Google's link spam policy and Bing's rules on link schemes are required reading before any outreach.⁷⁴
Analytics and measurement	Tracking set-up, GA4, Search Console data, dashboards, attribution and forecasting.	Measuring SaaS SEO in pipeline
AI search visibility	How assistants find and cite sites, crawler access, measuring citations, and content that answers well.	AI search, Measuring AI visibility

SEO specialisms, what each involves, and where to read further on this site

What AI tools have made cheap, and what they have not

A beginner in 2026 can do in an afternoon what took a week a few years ago: draft twenty title tag options, write a regular expression to filter Search Console queries, classify two thousand keywords by intent, write a Python script to check status codes, or summarise a crawl export. That is good news for you, because it removes the grind that used to fill a junior's first year. It also means those tasks are no longer what you are paid for.

The scarce skills are the ones that decide whether the cheap output is right and whether it matters. An AI tool can list fifty issues from a crawl; it cannot tell you which three will move revenue for this business, or notice that the traffic drop started the day the tracking code was changed. It can draft an article, but content that recycles what is already available, or that a generative AI model could easily produce, is the kind least likely to stand out; first-hand experience and a unique point of view are what Google's guidance rewards.³ Using generative AI to produce many pages without adding value for users falls under Google's scaled content abuse policy,⁷ and Bing's guidelines say large-scale content generated without oversight or editorial review may be excluded from indexing.⁴ Knowing where that line sits is judgement, and it is part of the job.

Diagram

How much AI tools have reduced the cost of common SEO tasks

Now cheap with AI toolsStill needs a person

Spreadsheet formulas, regex and small scriptsSpreadsheet formulas, regex and small scripts: 8 percent of the way from Now cheap with AI tools to Still needs a person.
First drafts of titles, descriptions and briefsFirst drafts of titles, descriptions and briefs: 18 percent of the way from Now cheap with AI tools to Still needs a person.
Grouping keywords by topic and intentGrouping keywords by topic and intent: 30 percent of the way from Now cheap with AI tools to Still needs a person.
Listing technical issues from a crawlListing technical issues from a crawl: 40 percent of the way from Now cheap with AI tools to Still needs a person.
Diagnosing why traffic changedDiagnosing why traffic changed: 72 percent of the way from Now cheap with AI tools to Still needs a person.
Prioritising what to do firstPrioritising what to do first: 82 percent of the way from Now cheap with AI tools to Still needs a person.
Getting changes approved and builtGetting changes approved and built: 92 percent of the way from Now cheap with AI tools to Still needs a person.
Designing measurement and reporting honestlyDesigning measurement and reporting honestly: 90 percent of the way from Now cheap with AI tools to Still needs a person.

Hover a row for the main thing that goes wrong.

My judgement, not a measurement. Tasks at the left are now quick with an AI assistant; tasks at the right still depend mostly on a person who understands the business and the data.

Choose a route in

Use the tool below to pick a starting route. It is a starting point, and plenty of people move between in-house, agency and freelance work. Dan answers that he comes from marketing, needs a salary within six months, and likes variety, which points him to agency work. That shapes his portfolio: agencies want to see audits and clear written recommendations, so he will put more care into the audit piece in stage five.

Tool / DecisionYour answers stay in this browser

Which route into SEO suits you?

Question 01

What are you bringing with you?

Pick the background closest to the work you have done most recently, paid or unpaid.

See all 6 possible results

Aim for a technical SEO role
Your background is the hardest part of technical SEO to teach. Developers and data people who learn how search engines crawl and index are useful from early on, in-house or at an agency.
Aim for a content SEO role
Writers and editors who learn search research and measurement are well suited to content SEO. The risk is staying a writer who knows some SEO; the work you need to show is research, briefs and measured results.
Start at an agency
An agency gives you many sites and problems in a short time, and usually more experienced people to learn from. It suits people who like variety and can work to deadlines.
Start in-house
In-house work gives you depth: you see work through from idea to result and learn how a business decides. It suits people who prefer ownership and long projects.
Freelancing can work for you, with care
You have runway and a network, which are the two things most new freelancers lack. You still have less experience to fall back on than your clients will assume, so keep the work small and specific at first.
Get employed first, freelance later
Freelancing without a network or experience means learning SEO and learning to sell at the same time, with no salary. Most people find it easier to spend a year or two employed and build both.

Answer honestly for where you are now. The answer can change after your first year.

Tasks0 of 4 done

01Read Google's page "Do you need an SEO?" from start to finish. It is written for people hiring SEOs, which makes it a useful description of what employers and clients will expect of you and what they are told to be wary of.
02Use the decision tool above and write your result, and the reasons you chose each answer, at the top of your learning log.
03Find five current SEO job adverts in your area or for remote roles. Copy the responsibilities and requirements into your log and highlight each skill that appears in three or more of them.
04Write two sentences in your log on what you think AI tools will do for you in this job and what they will not. You will check this against your experience at day 90.

Stage 02 / 07Weeks 1 and 2

Score your starting point and learn how search works

Know where your gaps are, and understand from the search engines' own documentation how pages are found, indexed and ranked, and what the rules are.

Score yourself before you start

Rate yourself on each skill below. Be strict: a 5 means you could teach it to someone else today. The advice for any skill you rate 1 or 2 is the first thing to study for it, and the band tells you where to start the 90 days. Score again at day 90 and keep both results in your log; the difference is useful to show in interviews.

Tool / ScorecardYour answers stay in this browser

SEO skills self-assessment

01How search works×3 weight
1: you could not explain crawling, indexing and serving. 5: you can explain all three, why a page might not be indexed, and how AI Overviews draw on the index.
New to meCould teach it
02Technical foundations×3 weight
1: HTML source and status codes are unfamiliar. 5: you can read page source, explain 301, 302, 404, 410 and 503, and write a correct robots.txt and canonical tag.
New to meCould teach it
03Search Console and analytics×3 weight
1: you have not used Search Console or GA4. 5: you can explain the Performance report metrics, find why a page is not indexed, and set up key events in GA4.
New to meCould teach it
04Keyword and intent research×2 weight
1: you have not done keyword research. 5: you can group queries by what the searcher wants, judge intent from the results page, and pick targets a site can realistically win.
New to meCould teach it
05Writing and editing for search×2 weight
1: you rarely write for work. 5: you can write and edit clear, useful pages and brief other writers well.
New to meCould teach it
06On-page and information architecture×2 weight
1: titles, headings and internal links are new to you. 5: you can plan a site structure, write title tags, and fix internal linking.
New to meCould teach it
07Local or ecommerce specifics
1: you do not know how the map pack or product results work. 5: you understand Business Profiles or product data well enough to audit them.
New to meCould teach it
08AI search and assistants×2 weight
1: you have not looked at how AI Overviews or ChatGPT cite sites. 5: you can explain how they find pages, how crawler access is controlled, and what can be measured.
New to meCould teach it
09Communication and reporting×2 weight
1: you have not written recommendations or reports for others. 5: you can write a one-page recommendation that a developer and a director would both act on.
New to meCould teach it
10Spreadsheets and data handling×2 weight
1: you avoid spreadsheets. 5: you are comfortable with lookups, pivot tables, filters and cleaning exported data.
New to meCould teach it

Your result

Rate the 10 criteria above to see where you stand.

Rate each skill from 1 (new to me) to 5 (I could teach it).

Dan scores 4 on writing and communication from his marketing work, 3 on spreadsheets, 2 on Search Console and analytics, and 1 on technical foundations and how search works. His result lands in the second band from the bottom, so he takes the full four weeks and spends extra time on stage three.

Turn your score into a week-by-week plan

A score is only useful if it changes what you do on Monday. Paste your scores and the advice the scorecard gave you into an AI assistant with the prompt below, and ask it for a plan that fits your hours. Then check the plan against this tutorial's order: assistants like to front-load exciting topics such as AI search, and the foundations have to come first.

Prompt /ChatGPT, Claude or similar

Turn the skills scorecard into a study plan

You are a senior SEO who trains juniors. I am starting a 12 week, part-time plan to move into SEO from [your current or last job]. I can study [hours per week] hours a week, mostly on [days or times you have free]. Here are my self-assessment scores from 1 (new to me) to 5 (could teach it), with the advice I was given for my weakest skills: [paste each skill, your score, and the advice text] The plan must follow this order: weeks 1 and 2 how search works and Google's rules; weeks 3 and 4 technical foundations; weeks 5 to 8 building and running a small site of my own; weeks 9 to 12 portfolio pieces. Only use free primary sources from Google, Bing and the tool makers themselves, and name the document for each task. Write the plan as a table with columns: Week, Session (day and length), Task, Primary source, What I should be able to do afterwards. Put my two weakest skills into every week in some form. Keep each session within my hours. If you are unsure whether a document or tool exists or is still free, say so in a final list called 'Check these' instead of guessing.

Fill in the parts in brackets before you send it

CheckOpen every document the assistant names before you rely on it. Assistants invent plausible titles and links, and Google renames pages. Replace anything you cannot find with the documents listed in this stage and the next.

Diagram

The skills an SEO builds, from the foundations up

01
Judgement and communication
Prioritising, saying what is not worth doing, getting changes approved and built, reporting honestly.
- Monthly reports
- Recommendations
- Stakeholder meetings
02
Specialisms
Local, ecommerce, international, digital PR, AI search visibility. Added after a year or two of general work.
- Local SEO
- Ecommerce SEO
- International SEO
- AI search
03
Content and site structure
Keyword and intent research, writing and editing, briefs, titles, headings and internal links.
- Search Console queries
- Google Trends
- Keyword Planner
04
Measurement
Reading Search Console and GA4, knowing what each counts and where they disagree.
- Search Console
- GA4
- Bing Webmaster Tools
05
Technical foundations
HTML, HTTP status codes, robots.txt, sitemaps, canonicals and rendering.
- Chrome DevTools
- Screaming Frog
- Rich Results Test
06
How search works and the rules
Crawling, indexing and serving, Search Essentials and the spam policies.
- How Search Works
- SEO Starter Guide
- Spam policies

Each layer depends on the ones beneath it. Learn from the bottom of this stack (shown last) upwards; jobs advertise mostly the top layers, and interviews test the bottom ones.

Read the primary material in this order

Most SEO advice online is secondhand. Some of it is good, some is out of date, and some is wrong. Google's guidance on third-party advice draws a useful line: good advice either qualifies its claims as opinion based on data or experience, or cites official guidance, and that third-party tools do not have access to Google's internal ranking data.⁸ Reading the primary documents first gives you something to check everything else against. Read them in this order, because each builds on the one before.

In-depth guide to how Google Search works. There are three stages: crawling, where its crawlers download text, images and videos from pages they find; indexing, where it analyses the page, picks a canonical from similar pages and stores signals; and serving, where it returns results relevant to the query, using factors that can include the searcher's location, language and device.⁹ Google does not accept payment to crawl or rank a site, and does not guarantee to crawl, index or serve any page.⁹
SEO Starter Guide. The practical basics: how to help Google find and understand pages, titles and snippets, links, images and site structure. It also covers things not to spend time on, such as the keywords meta tag, which Google Search does not use.²
Google Search Essentials. Three parts: technical requirements, spam policies and key best practices.¹⁰ The technical requirements are short: Googlebot is not blocked, the page works and returns an HTTP 200 status, and the page has indexable content.¹¹
Spam policies for Google web search. The behaviours that can lower a site's rankings or remove it, including cloaking, doorway abuse, expired domain abuse, link spam, scaled content abuse, site reputation abuse and thin affiliation. Violations can lead to a manual action, reported in Search Console.⁷ Know these well enough that you can tell a client or manager plainly when a tactic breaks them.
Creating helpful, reliable, people-first content. Google's self-assessment questions for content, covering originality, expertise and whether the page is made for people or for rankings.¹²
Optimizing your website for generative AI features on Google Search. How AI Overviews and AI Mode use the index, and a list of things you can ignore for Google Search, including llms.txt files, splitting content into small chunks for AI, and seeking inauthentic mentions.³
Bing Webmaster Guidelines. Bing's equivalent of Search Essentials and the spam policies on one page, written for Bing search and Copilot. It covers discovery, sitemaps, IndexNow, redirects and robots.txt, then the abuse Bing acts on, including cloaking, link schemes, keyword stuffing, automatically generated content at scale and prompt injection aimed at its language models.⁴ Reading it next to Google's pages shows how much the two agree.

Use an assistant as a tutor, with the document in front of it

An AI assistant is a good quizmaster and a poor source. Asked from memory, it will mix current Google guidance with old forum lore. Paste the text of the Google page you have just read into the conversation, tell it to work only from that text, and have it test you. Being made to explain something in your own words, and corrected against the source, is how the reading sticks.

Prompt /ChatGPT, Claude or similar

Have an assistant quiz you on a Google document

You are a strict but friendly tutor helping me learn SEO from primary sources. Below is the full text of a Google Search Central document I have just read: [document title and URL]. [paste the full text of the page] Work only from this text. Do not add facts from your own knowledge, and if a question I ask is not answered by the text, say 'The document does not say'. Ask me 10 questions, one at a time, and wait for my answer before the next one. Mix three kinds: recall ('What does Google say about...'), application ('A site does X. What does this document say will happen?'), and myth-checking ('Someone tells you Y. Is that supported by the document?'). After each answer, mark it right, partly right or wrong, quote the sentence from the document that settles it, and explain my mistake in one or two sentences. At the end, list the three ideas I struggled with most.

Fill in the parts in brackets before you send it

CheckCheck any quotation the assistant gives against the page itself. If it marks you wrong and you think you were right, find the sentence and decide for yourself: the document is the authority, not the assistant.

Two free research tools, and what their numbers mean

Google Trends shows how interest in a search term changes over time and by place. It works from a sample of searches, divides each data point by total searches for that place and time, and scales the result from 0 to 100, so the figures show relative popularity, not volumes.¹⁴ It is good for spotting seasonality and comparing two phrasings of the same idea. Keyword Planner, inside Google Ads, suggests related keywords and estimates monthly searches and advertising costs. To reach its basic features you must complete your Google Ads account setup, including billing information, and the estimates are meant for planning ad campaigns.¹⁵ Read each setup screen carefully, because the flow is built for advertisers and you only want the planner, not a live campaign. Treat its volumes as rough guides for comparing keywords, and once you have a site, trust your own Search Console impressions over any estimate.

Exercises for weeks 1 and 2

Reading is not enough on its own. Each exercise below forces you to apply a document to real search results. Dan does them using searches about bike repair, because he can judge the results from experience.

Tasks0 of 7 done

01After reading How Search Works, write a one-page explanation of crawling, indexing and serving for someone who has never heard of SEO, without looking at the source. Then check it against the page and correct anything you got wrong.
02Search for 20 queries in a subject you know. For each, record in a spreadsheet: the query, whether an AI Overview appears, the type of pages in the top five (guide, product page, forum, video, local listing), and in one sentence what you think the searcher wants.
03Read the spam policies and, for each of five policies, write one sentence describing a tactic you have seen offered online or in your own work that would break it. This is the list you will use when someone asks you to do something risky.
04Pick three pages that rank well for queries in your subject and score each against ten of Google's helpful content questions. Note which questions were hard to answer from the page alone.
05Read the mythbusting section of Google's generative AI guide and write down each thing it says you can ignore. You will be asked about at least one of them in your first month in a job.
06Open Google Trends, enter three of your 20 queries, set the region to your country and the range to the past five years, and note which are seasonal. Remember that Trends shows relative interest scaled from 0 to 100, not search counts.
07At the end of week 2, write a 300 word summary in your learning log of what surprised you, and any claim you had believed that the primary documents contradicted.

Stage 03 / 07Weeks 3 and 4

Learn the technical foundations

Be able to read a page's HTML, check its status code and response headers, and explain robots.txt, sitemaps and canonicals well enough to spot problems on a real site.

You do not need to be a developer to work in SEO, but you need to read what developers build. The tools are free: your browser's developer tools, the View Source option, and, if you are comfortable with a command line, curl. Work through the topics below in order, doing each exercise on two or three public sites.

Basic HTML

Learn to find these in a page's source: the <title> element, the meta description, the <h1> and other headings, links (<a href>), images and their alt text, the <link rel="canonical"> tag, any <meta name="robots"> tag, and structured data in a <script type="application/ld+json"> block. MDN Web Docs has a free reference page for each of these elements.¹⁶ Google builds the title shown in its results from several sources, including the <title> element and the main visible heading, so give every page a distinct, descriptive title without keyword stuffing or repeated boilerplate.¹⁷

There is a difference between the raw HTML a server sends and the rendered page after JavaScript runs. Google processes JavaScript as long as it is not blocked, though SEO for JavaScript-heavy sites is generally more complex.³ Other crawlers may not run it at all: Vercel's analysis of crawler traffic on its own network, published in December 2024, found that none of the major AI crawlers it tracked, including OpenAI's, Anthropic's and Perplexity's, executed JavaScript.¹⁸ Compare View Source with the Elements panel in developer tools on a modern site and note what appears only after rendering. The post on checking what crawlers see on a JavaScript site takes this further.

HTTP status codes

Every request returns a status code. What each code means is defined in the HTTP standard, RFC 9110;¹⁹ what a search engine does with it is documented by each engine. For Google, a 301 is a strong signal that the redirect target should be canonical and a 302 is a weak one; a 404 or 410 leads it to drop previously indexed URLs over time; a 429, 500 or 503 makes it slow its crawling, and URLs that keep returning server errors are eventually dropped.²⁰ Watch for soft 404s too: pages that return 200 but show an error or empty content, which Search Console flags.²⁰

Code	Meaning	How Google treats it
200	Success	Content goes on to indexing, which is not guaranteed.²⁰
301	Moved permanently	Followed; a strong signal the target should be canonical.²⁰
302	Found (temporary)	Followed; a weak signal the target should be canonical.²⁰
404 / 410	Not found / gone	Indexed URLs are removed over time and crawled less often.²⁰
429 / 500 / 503	Too many requests / server error / unavailable	Crawling slows; persistent errors lead to URLs being dropped.²⁰

The status codes you will meet most often, and what Google does with each

robots.txt

A robots.txt file tells crawlers which URLs they may access. The standard behind it, RFC 9309, says its rules are not a form of access authorisation,²¹ and it is not a way to keep a page out of search results either: a disallowed URL can still be indexed if other pages link to it, and appear without a description. To keep a page out of results, use noindex or password protection.²² This is one of the most common mistakes you will find, so learn the distinction early. AI crawlers are controlled through robots.txt too, covered in stage four and in the post on robots.txt for AI crawlers.

Sitemaps

An XML sitemap lists URLs you want search engines to know about. The format is an open protocol, and each file can hold up to 50,000 URLs and 50MB.²³ A site of about 500 pages or fewer that is well linked internally may not need one, but most sites benefit, and a sitemap helps discovery without guaranteeing that the URLs in it will be crawled or indexed.²⁴

Canonicals

When the same or very similar content is available at several URLs, Google picks one as canonical. Redirects and rel="canonical" are strong signals of your preference, and sitemap inclusion is a weak one.²⁵ They are signals, so Google may choose a different canonical from the one you set. Canonical tags are common: HTTP Archive's 2024 Web Almanac, which looked at the home page and one inner page of each site in its crawl, found them on 65% of mobile pages and 69% of desktop pages.²⁶ Duplicate content across several URLs is inefficient but will not cause a manual action.² You will hear people treat duplicate content as a penalty; Google's documentation says otherwise.

The free toolkit, and how to start with each tool

You can learn almost all of technical SEO with free tools. The four below are examples that most SEOs use at some point; others do similar jobs. Product details here are as published by each maker at the time of writing, and they change, so check the current terms before you rely on a limit.

Tool	What it does	Free use	Use it for
Chrome DevTools	A set of developer tools built into Chrome, including the Elements, Console and Network panels.²⁷	Built into the browser	Reading the rendered DOM, checking status codes and headers in the Network panel, and comparing raw and rendered HTML.
Screaming Frog SEO Spider	A desktop crawler for Windows, macOS and Linux that finds broken links, redirects, duplicate titles, robots directives and canonicals.²⁸	Crawls up to 500 URLs free; a paid licence removes the limit.²⁸	Crawling your lab site and small public sites, and seeing a site the way a crawler does.
Rich Results Test	Google's test for structured data on a URL or a pasted code snippet, using a smartphone or desktop user agent.²⁹	Free, no account needed	Checking structured data on any public page, including sites you do not own.
PageSpeed Insights	Google's report on a page's speed and experience, with lab data from Lighthouse and real-user data from the Chrome UX Report where there is enough of it.³⁰	Free	Understanding Core Web Vitals and the difference between lab and field data.

Free tools for learning technical SEO, as described by their makers at the time of writing

Tasks0 of 5 done

01In Chrome, open a page, right-click and choose Inspect (or press Control+Shift+C on Windows, Command+Option+C on a Mac) to open DevTools on the Elements panel. Then open the Network panel, reload the page, click the first request (the page itself) and read its status code and response headers.
02Download Screaming Frog SEO Spider from the maker's site and install it. By default it obeys robots.txt. The maker lists crawl configuration among the paid features, so in the free version you cannot slow the crawl yourself: keep free crawls to small sites, run them at a quiet time of day, and stop the crawl if pages start responding slowly.
03Crawl a small public site (or your own, once it exists). When it finishes, look at the Response Codes tab for anything that is not 200, the Page Titles tab for missing or duplicate titles, and the Directives tab for noindex and canonicals.
04Paste the URL of a recipe, product or event page into the Rich Results Test. Note which structured data types it finds and any errors, then use View Source to find the JSON-LD block the test read.
05Run the same page through PageSpeed Insights. Write down whether it shows real-user data or only lab data, and which Core Web Vitals pass or fail.

Exercises for weeks 3 and 4

Tasks0 of 6 done

01Open View Source on three pages from different public sites. For each, record the title, meta description, h1, canonical URL and any robots meta tag in your spreadsheet. Note any page where the canonical points somewhere other than the page itself, and work out why.
02Use the Network panel in your browser's developer tools (or curl -I followed by a URL) to check the status code of a working page, a URL you know does not exist on the same site, and an old URL you suspect redirects. Record whether the missing page returns a real 404 or a 200, which would be a soft 404.
03Read the robots.txt file of five sites by adding /robots.txt to the domain. For each, write in plain English what it blocks, which crawlers it names, and whether any rule looks like a mistake.
04Find the XML sitemap for two sites, usually listed in robots.txt or at /sitemap.xml. Open five URLs from each and check that they return 200 and are canonical to themselves.
05Pick one modern site built with a JavaScript framework. Compare View Source with the rendered Elements panel and list what content or links appear only after rendering.
06Write a one-page note in your log explaining, without notes, the difference between a 301 and a 302, between noindex and a robots.txt disallow, and between a canonical tag and a redirect.

Dan finds, on a bike shop's site, that a discontinued product URL returns a 200 with a message saying the product is no longer available. That is the soft 404 pattern from earlier in this stage. He notes it as a finding to practise writing up in stage five.

Stage 04 / 07Weeks 5 to 8

Build and run your own site as a lab

Have a small live site, set up with Search Console, GA4 and Bing Webmaster Tools, that you are growing and measuring yourself.

Reading teaches you what should happen. A site of your own shows you what actually happens: how long Google takes to find a page, which queries arrive that you did not expect, what Search Console reports and what it leaves out. It is also the only place you can make mistakes without costing anyone money, and the basis of the strongest portfolio piece you will have.

Choose the subject

Pick a subject you know from experience, because first-hand expertise is what makes content worth reading,¹² and because you will need to write 10 to 20 good pages. Avoid subjects where bad advice could hurt someone, such as health or money, unless you are qualified in them. A hobby, a trade you have worked in, or a place you know well all work. The test is whether you could write ten pages that contain something a reader could not get from the first page of results.

Dan chooses secondhand bike restoration, written from his own repairs: how to check a used frame for damage, how to replace brake cables, what a fair price is for a 1990s steel road bike. He photographs his own work. Nobody else has his photos or his specific mistakes, which is exactly the kind of non-commodity content that stands out.³

Choose and register the domain

Choose a name you can say out loud and spell over the phone. Keywords in a domain name have hardly any effect on ranking in Google beyond appearing in breadcrumbs, and the ending (the TLD, such as .com or .co.uk) only matters if you target one country's users, and even then is usually a low impact signal.² For a UK audience, a .co.uk or .uk name is a sensible choice; for a wider audience, .com. Avoid buying an old domain for its existing links: Google's spam policies name expired domain abuse.⁷

Any accredited registrar will do. Two examples with straightforward pricing: Cloudflare Registrar says it sells registrations, transfers and renewals at cost with no markup, and includes free WHOIS redaction and one-click DNSSEC.³³ Porkbun lists free WHOIS privacy and free SSL certificates among its domain features.³⁴ Some hosts include a domain with a plan: at the time of writing, Krystal lists a free domain on its WordPress hosting plans from the Ruby tier upwards.³⁵ Whichever you use, check the renewal price as well as the first-year price, and turn on the registrar's two-factor sign-in.

Tasks0 of 3 done

01Shortlist three names for your subject, each under about 15 characters, without hyphens, and easy to say aloud. Search each name in Google to check it is not already a business or brand you would be confused with.
02Check availability and the first-year and renewal prices at two registrars. Buy the name for one year in your own name, with WHOIS privacy or redaction on, and turn on two-factor authentication for the registrar account.
03Note where the DNS for the domain is managed (usually the registrar until you change it). You will add a TXT record there to verify Search Console.

Pick a build route

There are three realistic routes for a lab site. Pick the one closest to the jobs you want: content and agency roles mostly meet WordPress, and technical roles increasingly meet sites built with modern frameworks. Avoid platforms that do not let you edit robots.txt, set redirects, or add a verification record, because those are the lessons.

Route	Examples	Good for learning	Watch out for
WordPress on managed hosting	WordPress software on a host such as Krystal's managed WordPress plans.³⁵ WordPress itself is free, open-source software licensed under the GPL.³⁶	The CMS you will meet most in agency and small business work. Since version 5.5 WordPress core has produced basic XML sitemaps without a plugin.³⁷	Plugins and themes can add bloat and duplicate URLs (tag and author archives). Keep plugins few and updated.
Static site built with an AI coding tool	Astro, a framework designed for content-driven sites that ships no client-side JavaScript by default,³⁸ built with Claude Code³⁹ or Cursor,⁴⁰ hosted on Cloudflare Pages,⁴¹ Netlify⁴² or GitHub Pages.⁴³	You see and own every file: robots.txt, sitemap, canonicals, redirects. Good preparation for technical roles and for working with developers.	The AI tool will make SEO mistakes you must catch. Needs a GitHub account and some comfort with a terminal.
Hosted website builder	Builders offered by many hosts and registrars.	Quick to launch if you mainly want to practise writing and Search Console.	Check before you start that you can edit robots.txt, set redirects and add a DNS or meta tag verification.

Three ways to build a lab site, with named examples. Product details are from each provider's own pages at the time of writing.

Route A: WordPress

Tasks0 of 5 done

01Choose a host with a managed WordPress plan, buy the smallest plan that covers one site, and install WordPress using the host's installer. Point the domain at the host as the host's instructions describe, and make sure HTTPS works before you go further.
02In the WordPress admin, open Settings, then Reading, and confirm that the option asking search engines not to index the site is not ticked once you are ready to launch. A forgotten tick here is one of the most common launch mistakes.
03Open Settings, then Permalinks, and choose the Post name structure so URLs read like /check-a-used-frame/.
04Install a lightweight theme and as few plugins as possible. Visit /wp-sitemap.xml on your domain to confirm the core sitemap exists (an SEO plugin may replace it with its own; either is fine, but there should be one).
05Create the pages from your site sketch, then check the source of one page for a single h1, a sensible title, and a canonical tag pointing to itself.

Route B: a static site built with an AI coding tool

Claude Code is an agentic coding tool that reads a codebase, edits files and runs commands, and works in the terminal, IDEs, a desktop app and the browser; most of its surfaces need a Claude subscription or an Anthropic Console account.³⁹ Cursor describes itself as a coding agent for building software.⁴⁰ Either can build a small Astro site for you from a clear brief. Your job is to brief it well and check the output, which is exactly the skill an employer wants from someone who works with developers.

Tasks0 of 5 done

01Install Git and Node.js, create a free GitHub account, and install your coding tool by following its own documentation. Create an empty folder for the project and open the tool in it.
02Give the tool the build brief from the prompt card below, with your subject and page list filled in. Let it scaffold the Astro project, then run it locally and click through every page.
03Read the generated files yourself: robots.txt, the sitemap, the head of one page (title, meta description, canonical, robots meta) and the 404 page. Ask the tool to fix anything that is wrong, one issue at a time.
04Push the project to a GitHub repository. In Cloudflare Pages, Netlify or GitHub Pages, connect the repository, accept the build settings for Astro, and deploy.
05Add your domain in the host's custom domain settings and create the DNS records it asks for at your registrar. Wait for HTTPS to be issued, then confirm that the http and www versions redirect to one canonical address with a 301.

Prompt /ChatGPT, Claude or similar

Brief an AI coding tool to build the lab site

You are a senior web developer who also knows technical SEO. Build a small static website with Astro in this folder. It is a practice site about [your subject], written for [who the readers are]. The domain will be https://[your domain]. Pages: a home page, these section pages: [list 3 to 5 sections], and these articles under them: [list article titles and which section each belongs to]. Use placeholder text I will replace; do not write the articles. Requirements: - Every page has a unique title element, a meta description, one h1, and a self-referencing canonical using the full https URL. - Generate an XML sitemap of every indexable page and a robots.txt that allows all crawlers and links to the sitemap. - A real 404 page that returns a 404 status. - Clean URLs in lowercase with hyphens, and breadcrumbs on article pages. - Article pages link to their section and to two related articles. Section pages link to all their articles. - No client-side JavaScript unless a feature needs it. Images have width, height and alt attributes. - Do not add analytics yet. Before writing code, show me the file and URL plan and wait for my approval. After building, list every SEO decision you made and anything you were unsure about.

Fill in the parts in brackets before you send it

CheckCheck the built site yourself in a browser and with View Source. Coding assistants often get canonicals, trailing slashes and 404 status codes wrong, and confidently report that they did not.

Audit the build before launch

If you built with an AI coding tool, the free SEO for AI Agents skill pack on this site gives the agent a method for checking its own work. It is an open-source, MIT-licensed set of skills for Claude Code, Cursor, Codex, Gemini CLI and Zed, marked experimental. Every skill audits by default and changes nothing until you approve a fix. The Launch QA skill runs the checks that most often stop a new site being indexed, against the live URL: staging noindex and robots.txt blocks, preview hosts in canonicals and sitemaps, redirects, soft 404s and crawlers blocked at a CDN, and splits the results into launch blockers and fix-this-week. The Measurement setup skill checks analytics and Search Console verification are installed correctly. In Claude Code you install the pack by copying its skills folder into your skills directory; in any other chat you can paste a skill's copy-paste version instead. Treat its findings the way you would treat a colleague's: check each one yourself before you accept it, because explaining why a finding is right is what you will be asked to do in an interview.

Set up Search Console

Search Console supports two kinds of website property. A Domain property covers every subdomain and protocol and can be verified only through a DNS record; a URL-prefix property covers only URLs that start with the exact prefix, including http or https.⁴⁴ Use a Domain property for your own site so that nothing is missed, and learn the DNS step, because you will do it for employers and clients. DNS verification through the domain name provider is more complex than the other methods but the only way to verify a Domain property, and simple if your provider appears in the verification wizard.⁴⁵ Data starts collecting as soon as the property is added, and takes a few days to build up.⁴⁵

Tasks0 of 5 done

01Go to Search Console, sign in, and add a property. Choose Domain, enter your domain without https or www, and continue.
02Copy the TXT record Search Console shows you. In your registrar's or DNS host's settings, add a TXT record at the root of the domain with that value, save it, then return to Search Console and click Verify. If it fails, wait an hour for DNS to update and try again.
03Open the Sitemaps report, enter the address of your sitemap (wp-sitemap.xml for WordPress core, or whatever address your static build produces) and submit it.
04Open the Search generative AI setting under Settings and confirm the site is set to be included, which is the default, so the Generative AI report can collect data for you.
05Use the URL Inspection bar at the top of Search Console on your home page and run the live test. Read the result so you know what a healthy page looks like before anything goes wrong.

Set up GA4

Setup has three parts: create an account, create a property with your reporting time zone and currency, then add a Web data stream with your site's URL and install the Google tag. Turn on enhanced measurement, which collects page views and some other events automatically. Many CMSs have a field for the Google tag ID (it starts with G-); otherwise you paste the full tag into every page.⁴⁶

Tasks0 of 6 done

01Go to analytics.google.com and click Start measuring. Name the account, set the data-sharing options, and click Next.
02Name the property, set the reporting time zone and currency, choose your industry and business size, choose your objectives, then click Create and accept the terms.
03Choose Web, enter your site's URL and a stream name, leave enhanced measurement on, and click Create stream. Copy the Google tag ID or, under View tag instructions, the full tag.
04Add the tag: in WordPress through your theme or a plugin's Google tag field; on a static site, ask your coding tool to add the tag to the shared page layout, then check View Source on two pages to confirm it is there once.
05Decide what counts as success on your site (Dan chooses newsletter sign-ups and downloads of his parts checklist). In Admin, under Data display, open Events and mark those events as key events with the star.
06Link Search Console: in Admin, under Product links, open Search Console links, click Link, choose your Search Console property, then select your web data stream and submit.

You can mark up to 30 events as key events on a standard property, and marking an event applies from that point forward and does not change historic data.⁴⁷ The Search Console link adds two reports to GA4, for organic search queries and organic landing pages, with Search Console data available about 48 hours after collection and a maximum of 16 months of history.⁴⁸

Set up Bing Webmaster Tools

Tasks0 of 3 done

01Sign in to Bing Webmaster Tools, or create an account.
02On the My Sites page, click Import, sign in with the Google account that owns your Search Console property, and click Allow.
03Select your site from the list Bing shows, with its sitemaps, and click Import. The site is verified automatically.

Imported sites are verified automatically, sitemaps found in Search Console are imported too, and traffic data can take up to 48 hours to appear.⁵⁰ Bing's tools include a Search Performance report that draws on web search and Copilot responses among other sources, URL inspection, backlink data, sitemap reporting, and support for IndexNow, an open protocol for notifying search engines of added, updated or deleted URLs.⁵¹ Bing also has an AI Performance report, introduced in public preview in February 2026, that shows how often your pages are cited as sources in Copilot and AI summaries in Bing.⁵²

Decide what AI crawlers may do

Write your robots.txt deliberately. OpenAI separates OAI-SearchBot, which surfaces sites in ChatGPT search, from GPTBot, which crawls content that may be used to train its models, and the two can be allowed or blocked independently.⁵³ Perplexity says PerplexityBot surfaces and links sites in its results and is not used to crawl content for foundation models.⁵⁴ Anthropic runs three: ClaudeBot collects content that may be used for training, Claude-SearchBot improves Claude's search results, and Claude-User fetches pages when a person asks Claude a question, so blocking the last two can reduce your visibility in Claude.⁵⁵ Google-Extended is a control token for whether Google may use your content to train Gemini models and to ground answers in Gemini Apps, and it does not affect inclusion or ranking in Google Search.⁵⁶ Separately, Search Console now has a Search generative AI control that lets a site include or exclude itself from AI Overviews, AI Mode and generative AI features in Discover; including is the default.⁵⁷ For a lab site, allowing search crawlers is sensible because you want to see where you are cited.

Build, publish and watch

Tasks0 of 8 done

01Before writing anything, sketch the site structure on paper: the home page, three to five section pages, and the articles under each, with how they link to each other. Build from that sketch.
02Test each GA4 key event yourself and check it appears in the Realtime report within a few minutes.
03Check your robots.txt allows the crawlers you want and blocks nothing important, and confirm the sitemap shows as successfully read in both Search Console and Bing Webmaster Tools.
04Crawl the whole site with Screaming Frog and fix every non-200 internal link, missing title and wrong canonical before you publish more pages.
05Publish your first five pages, each written from your own experience, with a distinct title, a clear h1, and internal links to and from at least two other pages. Record the publish date of each in a spreadsheet.
06Use URL Inspection in Search Console on each new page. Record when Google first reports each one as indexed, and if it is not indexed, the reason it gives.
07Publish five to fifteen more pages over weeks 6 to 8, choosing topics from the queries that start to appear in Search Console's Performance report as well as from your own plan.
08Every Friday, spend 30 minutes in Search Console and GA4 and write a short entry in your log: what changed, what you think caused it, and what you will try next.

Learn what the data does and does not tell you

Search Console's Performance report shows clicks, impressions, click-through rate and average position. Average position is the average of your site's topmost result, not every result.⁵⁸ Data usually appears two to three days after collection, and Google may not report queries made a very small number of times or that contain personal or sensitive information.⁵⁹ That is why the rows in the table rarely add up to the chart totals. Search Console and GA4 also count differently: GA4 records only visitors whose browsers run its JavaScript, Search Console reports in Pacific time, and the two can cover different sets of URLs.⁵⁹ Expect the numbers to disagree, and learn to explain why.

Search Console also has a Generative AI performance report that shows impressions for your pages in AI Overviews and AI Mode, by page, country, device and date. Google rolled it out to all sites on 31 August 2026, and it may not appear until a site has had enough impressions in those features.⁶⁰ On a new lab site it may stay empty for weeks, which is itself a useful lesson in what small sites can measure. For assistants that publish no data, the post on measuring visibility in AI search explains how to sample prompts and record results.

Say that by the end of week 8 Dan's site has 14 pages, 11 of them indexed, and Search Console shows a few hundred impressions and a handful of clicks across the period. Those are illustrative numbers and his could be higher or lower. They will not impress anyone as traffic. What matters is that he can show the indexing timeline for each page, explain the three pages that were not indexed and what he did about them, and point to queries he did not plan for that told him what readers actually wanted.

Stage 05 / 07Weeks 9 to 12

Build a portfolio that proves skill

Turn your work into four pieces an employer or client can read in ten minutes each: a lab site case study, a public site audit, a technical write-up and a content brief.

Hiring managers want evidence that you can do the work, and most SEO hiring weighs demonstrated work more than qualifications. Free certifications are worth having for what they teach: Google's Analytics Academy on Skillshop offers free GA4 courses and a certification.⁶¹ Treat a certificate as a supporting line on your CV. The four pieces below are what you will talk about in interviews.

Publish them on your own site or in a simple shared folder, each as a page or PDF with a short summary at the top. Keep each one short enough to read in ten minutes. Long documents suggest you cannot prioritise.

1. The lab site case study

Write up the lab site as a case study: what you built and why, how you set up measurement, the indexing timeline, what the Search Console data showed, what you changed in response, and what happened next. Include Search Console screenshots with the date range visible. Show absolute numbers, not only percentage changes, because growth from 10 clicks to 40 is a 300% rise and still 40 clicks. Say what you could not attribute.

A structure that works: a two-sentence summary; the site and why you built it; set-up (property type, key events, Bing, robots.txt decisions); a table of pages with publish and index dates; the Search Console chart for the whole period with the date range visible; three things you changed because of the data, each with what happened next; what you could not explain; and what you would do with another three months.

Prompt /ChatGPT, Claude or similar

Draft a portfolio case study that reports results honestly

You are an editor who has hired SEOs and dislikes inflated claims. Help me write a one-page case study of a practice website I built. Facts (use only these; do not add numbers): - Site and subject: [domain and subject] - Dates: built [start date], measured to [end date] - Set-up: [Search Console property type, GA4 key events, Bing Webmaster Tools, robots.txt choices] - Pages: [number published], [number indexed], with these indexing problems and fixes: [list] - Search Console totals for the period: [clicks], [impressions], top pages: [list] - Changes I made because of the data, and what happened after each: [list] - Things I cannot explain or attribute: [list] Write it with these sections: Summary (two sentences), What I built, How I measured it, What happened, What I changed and why, What I cannot attribute, What I would do next. Report every result as an absolute number with its date range. If a percentage appears, put the absolute numbers next to it. Flag any sentence in my facts that sounds like a claim I cannot support from the data, and list them at the end under 'Check before publishing'.

Fill in the parts in brackets before you send it

CheckRead the draft against your Search Console screenshots line by line. If a number is not in the screenshot or your spreadsheet, delete it, however good it sounds.

2. A written audit of a public site, done respectfully

Pick a real public site in a sector you want to work in, ideally a small business or charity, and audit it from the outside. You do not have its Search Console or analytics, so your findings are observations and hypotheses, and the audit should say so. Keep it to the ten findings that matter most, ordered by likely impact, each with the evidence, the fix, who would do it, and how you would check it worked. The first five checks for a local SEO audit is a useful model for ordering.

Crawl politely: keep crawls small, respect robots.txt, run them outside the business's busy hours, and stop if the site slows. Never run security scanning or load testing tools against a site you do not own. In the UK, doing an unauthorised act that you intend, or are reckless about, impairing the operation of a computer is an offence under section 3 of the Computer Misuse Act 1990.⁶²
Use only public information, and do not try to access anything behind a login.
Write findings in a neutral tone. You are showing how you think, not mocking someone's website.
Consider anonymising the business in the published version, or asking the owner's permission before you publish it.
Do not cold-email the audit to the business as a sales pitch. Unsolicited email offering SEO services is one of the warning signs Google gives site owners who are hiring.¹

Dan audits a hypothetical independent cycle shop in Norwich with an online store. His top finding is that discontinued product pages return 200 with an out of stock message, the soft 404 pattern from stage three, which he presents as a hypothesis with three example URLs and a proposed rule for when to redirect, when to return 404 or 410, and when to keep the page.

Prompt /ChatGPT, Claude or similar

Have your audit draft critiqued as a senior SEO would

You are a head of SEO at an agency reviewing a junior's first audit before it goes to a client. Be direct and specific, as you would be in a real review. Context: this is an outside-in audit of [type of business and sector], written without access to the site's Search Console or analytics. The business goal I assumed is [goal]. [paste your audit] Review it in this order: 1. Prioritisation: are the top three findings the ones most likely to affect the business goal? Which finding would you cut, and which is missing? 2. Evidence: for each finding, is the evidence enough to support it? Mark each Supported, Weak, or Unsupported, with one sentence on why. 3. Accuracy: list any statement about how Google or another search engine behaves, and say whether it needs a citation to the provider's documentation. Do not tell me a rule is correct unless you are confident; if unsure, say 'check this against Google's documentation'. 4. Tone: quote any sentence that reads as mocking or as a claim about the business I cannot know from outside. 5. Clarity: could a developer act on each fix without asking me a question? Finish with the three changes that would most improve the audit.

Fill in the parts in brackets before you send it

CheckAn assistant can sound like a senior reviewer without being one. Check every rule it raises in Google's documentation, and still ask a real SEO to read your final version.

What not to put in a portfolio or CV

Some portfolio habits only cost you credibility; others cross a line. Claiming a qualification or experience you do not have in order to get a job is a false representation, and in England, Wales and Northern Ireland dishonestly making a false representation to make a gain is fraud by false representation under section 2 of the Fraud Act 2006.⁶³ Scotland has its own law. Publishing an employer's or client's data without permission usually breaches the confidentiality terms you worked under, so get written permission or anonymise it. Running intrusive tools against other people's sites can be an offence, as above.⁶² A lab site built with bought links or mass-produced pages breaks Google's spam policies,⁷ and an audit that mocks the business tells an employer how you would talk about their clients.

Diagram

Portfolio and CV practices, sorted

Reporting results as absolute numbers with dates and caveats
Lets the reader judge the result for themselves, which is what interviewers want to see.
Fine to doFine to do
Auditing a public site from outside, crawled gently, in a neutral tone
Uses only public information and shows how you think without harming the business.
Fine to doFine to do
Turning small numbers into big percentages with no base
Not unlawful, but experienced interviewers spot it and stop trusting the rest of the portfolio.
A grey area: handle with careA grey area: handle with care
Publishing an audit that names and insults the business
It tells an employer how you would talk about their clients. Anonymise the business or write neutrally.
A grey area: handle with careA grey area: handle with care
Using a client's or employer's data without written permission
Usually breaches confidentiality terms; anonymise it or get permission in writing.
A grey area: handle with careA grey area: handle with care
Cold-emailing audits to businesses as a sales pitch
Site owners are told to be wary of unsolicited SEO emails, so it starts the relationship as a red flag.
A grey area: handle with careA grey area: handle with care
A lab site grown with bought links or mass AI pages
Link spam and scaled content abuse are named in Google's spam policies.
Against Google's policiesAgainst Google's policies
Claiming certifications or experience you do not hold
Dishonestly making a false representation to gain a job can be fraud under section 2 of the Fraud Act 2006.
Against the law (UK)Against the law (UK)
Running security scans or load tests on a site you do not own
An unauthorised act that impairs, or risks impairing, a computer's operation is an offence under the Computer Misuse Act 1990.
Against the law (UK)Against the law (UK)

UK law where law applies; other countries differ. Each verdict restates the text above.

3. A technical write-up

Pick one technical problem and explain it fully: what it is, how you found it, how you tested it, and what fixed it or would fix it. The best source is a real problem on your lab site, such as a page that would not index, a redirect chain after you changed a URL, or a canonical Google ignored. A clear 800 word explanation of one problem shows more skill than a 40-point checklist.

4. A content brief

Write a brief for one page that a freelance writer could follow without asking you questions. It should say who the page is for and what they are trying to do, the queries it should answer and the intent behind them (with what Google currently shows for those queries), the outline, what first-hand material the writer needs (photos, data, experience), internal links in and out, the title and h1, and how you will judge whether it worked after 8 to 12 weeks.

Tasks0 of 6 done

01Write the lab site case study, with Search Console screenshots that show the date range, absolute numbers next to every percentage, and a paragraph on what you could not attribute.
02Choose a public site, crawl it slowly while respecting robots.txt, and write a ten-finding audit ordered by likely impact, each with evidence, fix, owner and how to check it. State clearly that you had no access to their data.
03Write an 800 to 1,200 word technical write-up of one problem you found and fixed or tested on your lab site.
04Write a content brief for one page, either for your lab site or for the business you audited, and test it by asking a friend to read it and list anything they would need to ask you.
05Ask an experienced SEO, found through a community or event, to read one piece and tell you the weakest part. Fix it before applying anywhere.
06Score yourself again on the skills scorecard in stage two and record the change from your baseline.

Stage 06 / 07Weeks 13 onwards

Get the first role or the first client

Find openings, prepare for what interviewers and clients test, and talk about your results honestly.

Where to look

Junior SEO roles are advertised under several titles: SEO executive, SEO assistant, junior SEO, SEO analyst, content SEO, and sometimes digital marketing executive with SEO in the description. Look on general job boards and LinkedIn, but also on agency websites directly, since smaller agencies often hire from their own careers page. Local digital marketing meetups and SEO conferences are where many people hear about roles first. Google runs Search Central Live events in several countries, announced on its blog.⁶⁴ Answering questions well in the Google Search Central Help Community is also a public record of what you know.³

What interviewers test

Expect three kinds of test. The first is fundamentals: explain how Google crawls and indexes, the difference between noindex and a robots.txt disallow, what a canonical does. The second is diagnosis: a scenario where something went wrong. The third is communication: a take-home task, or explaining a recommendation as you would to a manager. The diagnosis question is where most candidates struggle, because it tests your method for finding causes.

Diagram

A method for the question "organic traffic dropped last week: what do you check?"

Confirm the drop is real

Compare the same period in Search Console and GA4. If GA4 fell and Search Console did not, suspect tracking, such as a changed tag or consent banner.

Step 1 of 7

01 Confirm the drop is real: Compare the same period in Search Console and GA4. If GA4 fell and Search Console did not, suspect tracking, such as a changed tag or consent banner.

02 Check the calendar: Seasonality, bank holidays, last year's same week, and any release or migration on the site that week.

03 Check Google's side: The Search Status Dashboard lists incidents and ranking updates, with dates.

04 Segment the loss: Which pages, queries, devices and countries lost clicks. A drop in one template or section points somewhere very different from a site-wide one.

05 Check indexing and access: Page indexing report, robots.txt changes, noindex tags, status codes, and the Manual actions and Security issues reports.

06 Separate clicks from impressions: Impressions steady with fewer clicks suggests a change in how results look; both falling suggests lost rankings or indexing.

07 Write it up: What happened, the evidence, what you are confident of and what you are not, and the next step.

One sensible order of checks. Interviewers care more that you have a method and say why each check comes where it does than about the exact order.

The Search Status Dashboard shows the status of Google's crawling, indexing, ranking and serving, with incidents and ranking updates listed by date.⁶⁵ Checking it before assuming a site-side cause is a habit interviewers notice. So is checking the Manual actions and Security issues reports in Search Console early, because a manual action for a spam policy violation is reported there.⁷

A sample audit task, and how to answer it

A common take-home task looks like this: here is a website; in two hours, tell us the five things you would do first and why. Practise with a timer on a site you have not seen. A good answer has a short summary at the top listing the five recommendations, then one section per recommendation with the evidence, the expected effect, the effort, and the order you would do them in. It states your assumptions about the business and what you would want to see in Search Console before committing. It leaves out the forty minor issues a crawler found. Choosing what to leave out is the skill being tested.

Dan practises on a hypothetical garden centre site. His five, in order: product pages for seasonal plants return 200 with a sold out message all year (soft 404s), the category pages have identical title tags, the opening hours page is reachable only through a JavaScript menu, the blog and shop each have their own sitemap but neither is listed in robots.txt, and the local pages for two branches share most of their text. He notes for each what he would check in Search Console first, and leaves out the 300 images without alt text the crawler flagged, with one line saying why they come later.

Prompt /ChatGPT, Claude or similar

Rehearse an interview audit task

You are the head of SEO at [an agency or in-house team, and its sector] interviewing me for a junior SEO role. Set me a realistic take-home task: describe a fictional website in that sector in enough detail to work from (its pages, templates, what the crawl and Search Console might show, and the business goal), and ask for my top five recommendations in priority order. When I send my answer, assess it as you would in a real hiring decision. Score each of these from 1 to 5 with one sentence of evidence: prioritisation (did I pick what matters for this business), diagnosis (is each finding supported by evidence), clarity (could a manager and a developer both act on it), honesty (did I state assumptions and limits), and effort and order (is the sequence realistic). Then ask me three follow-up questions an interviewer would ask, one at a time, and tell me what a strong answer to each would include. Do not invent Google rules; if my answer cites one, ask me where Google says it.

Fill in the parts in brackets before you send it

CheckThe fictional site is invented, so the assistant can make it as tidy or as messy as it likes. Practise on real public sites too, and check any SEO rule it states against Google's own documentation before you repeat it in an interview.

Talking about results honestly

You will be asked what results you have achieved. Answer with absolute numbers, dates and context: the lab site went from no impressions to a few hundred a week over eight weeks, here is the chart, and most of it came from three pages about brake repairs. Say what else could explain it. Do not convert small numbers into large percentages, do not count traffic you cannot attribute to your work, and do not imply experience you do not have. Google's own advice to people hiring SEOs is to ask about previous work and to be wary of anyone who cannot explain what they did or who guarantees rankings, since nobody can guarantee a number one ranking.¹ Experienced interviewers have heard every inflated claim. Honesty about modest results reads as competence.

If you are going freelance

In the UK, if you sell services regularly for profit you are likely to be trading and may need to register a business and report income to HMRC; GOV.UK's guidance on working for yourself explains how to check.⁶⁶ If your gross trading income is £1,000 a year or less, the trading allowance usually means you do not need to tell HMRC about it, unless you cannot use the allowance or must register for Self Assessment for another reason.⁶⁷ Rules differ in other countries, so check your own tax authority. Start with fixed-scope work, such as a Search Console set-up and review, so both sides know what is being delivered. Ask clients to add you to their Search Console property as a user; owners can grant full or restricted access without sharing a password, and you should never ask for a client's Google login.⁶⁸ Write down before you start what the client should expect and when, using realistic timescales, which range from hours to several months.²

Search Console's permission levels are worth knowing exactly. An owner has full control and can add users; a full user can view all data and take some actions; a restricted user can view most data. Owners manage the list under Settings, then Users and permissions.⁶⁸ For an audit, restricted access is usually enough; for ongoing work where you submit sitemaps or request indexing, ask for full access. The instructions to send a client are short:

Tasks0 of 3 done

01Send the client these instructions: in Search Console, choose the property, open Settings, then Users and permissions, click Add user, enter my email address, choose Full or Restricted, and click Add.
02For GA4, ask the client to add you as a user on the property with a Viewer role for an audit, or a higher role only if you will change the set-up.
03Record in your engagement letter which accounts you were given access to and at what level, and agree that the client removes your access when the work ends.

Tasks0 of 6 done

01Shortlist 15 to 20 openings or target agencies and companies, and record in a spreadsheet the title, the skills they ask for, and which of your portfolio pieces best matches each.
02Rewrite your CV so the top third lists the four portfolio pieces with links and a one-line result for each, stated with absolute numbers.
03Practise the traffic drop question aloud, timing yourself to three minutes, using the method above.
04Do one timed two hour audit task on a site you have not seen, then compare your five recommendations against the summary rules above.
05Prepare two questions for each interviewer: how juniors are trained and reviewed, and how SEO recommendations get prioritised and built.
06If freelancing, write a one-page fixed-scope offer, check the GOV.UK guidance (or your own country's rules) on registering, and agree in writing how you will get Search Console access.

Stage 07 / 07Months 4 to 15

Do the first year well and keep current

Report usefully, prioritise well, learn to say what is not worth doing, and keep up with changes from the providers themselves.

How to report

A report exists to help someone decide something. Lead with the outcome the business cares about, such as orders, enquiries or sign-ups from organic search, then the search data that explains it, then what you did, what you will do next and what you need from others. Keep the caveats short and consistent: Search Console data lags by two to three days and omits some rare queries;⁵⁹ average position is the average of your topmost result;⁵⁸ AI Overview and AI Mode visibility is included in the Web search type and has its own Generative AI report.⁵⁶⁰ If you report AI search visibility from assistants that publish no data, say how you sampled it.

The Search Console interface shows a limited number of rows in its tables, and the API returns more.⁶⁹ Once you are comfortable, learn to pull data through the API or bulk export so you are not reporting on a truncated table. This is a good use of an AI assistant: have it help write the script, then check its output against the interface for a few rows.

How to prioritise

You will always have more possible work than time. For each idea, write down the likely effect on the business outcome, how confident you are in that effect and why, and the effort including other people's time. Do the high-effect, high-confidence, low-effort work first, and say what you are not doing. The audits that crawlers produce rank issues by count, not by business effect; a missing alt attribute on 3,000 images may matter far less than one broken redirect from a page that used to take orders.

Learn to say what is not worth doing

One of the most useful things a junior can do is stop a team wasting effort, and Google's documentation makes this easier, because it lists what Google ignores. Google Search does not use the keywords meta tag.² E-E-A-T is not itself a ranking factor, and there is no minimum word count.² For Google's generative AI features, you do not need llms.txt files or special markup, do not need to break content into small chunks, do not need to rewrite content for AI, and structured data is not required to appear there.³ Creating separate pages for every variation of a query, or for fan-out queries, primarily to manipulate rankings or AI responses, violates Google's scaled content abuse policy.³ When someone proposes one of these, link to the relevant page and suggest what the time should go on instead.

Other assistants document their own behaviour separately, so a claim that is true for Google may not be true for ChatGPT or Perplexity. When the evidence is uncertain, say so, and propose a small test before a large rollout.

Keep current from the source

SEO news travels fast and is often wrong in the first day. Read the source first. The Google Search Central Blog announces documentation changes, new Search Console reports and policy updates.⁶⁴ The Search Status Dashboard confirms when ranking updates start and finish.⁶⁵ For AI crawlers, the providers' own pages are the reference: OpenAI's crawler documentation,⁵³ Anthropic's,⁵⁵ Perplexity's,⁵⁴ and Google's list of common crawlers, which includes Google-Extended.⁵⁶ The Bing Webmaster Blog covers Bing and Copilot reporting.⁵² Set aside 30 minutes a week to read these, and note anything that changes advice you have already given.

Tasks0 of 6 done

01Build a one-page monthly report template for your employer or client: business outcome first, then search data, then work done, next steps and what you need from others, with a fixed caveats box at the bottom.
02Keep a prioritisation spreadsheet with columns for effect, confidence, effort and owner, and review it with your manager or client every month.
03Keep a "not worth doing" note with links to the Google documentation for each item, and use it when those ideas come up.
04Subscribe to the Google Search Central Blog's feed and bookmark the Search Status Dashboard, OpenAI's, Anthropic's and Perplexity's crawler pages, Google's common crawlers page and the Bing Webmaster Blog. Block 30 minutes a week to read them.
05Keep your lab site running and use it to test anything new before you recommend it at work.
06At the end of month 12, write a one-page review of the year: what you got right, what you would do differently, which specialism you enjoyed most, and what you want to learn next.

Questions8 answered

Common questions

01Do I need a degree or a certification to get an SEO job?

Most SEO hiring weighs demonstrated work: a site you grew, an audit, a technical write-up. A degree can help with some graduate schemes, but it is rarely required. Free certifications such as Google's Analytics Academy courses on Skillshop are worth doing for what they teach, and work best as a supporting line on your CV.⁶¹

02Is SEO still a career now that AI answers so many searches?

Google's AI Overviews and AI Mode are rooted in its core Search ranking and quality systems, and Google's position is that optimising for them is still SEO.³ Bing says the same foundation underpins Copilot.⁴ Other assistants run their own crawlers and indexes, which adds work. What has changed is measurement, since fewer searches end in a click when an AI summary appears,⁶ and the routine tasks AI tools now do quickly; judgement, diagnosis and stakeholder work are still needed.

03Do I need to learn to code?

You need to read HTML, understand HTTP status codes, and know how robots.txt, sitemaps and canonicals work. Writing code is optional, though basic scripting in Python or spreadsheet formulas helps a lot with data work, and AI assistants make it easier to start. Technical SEO roles expect more.

04How long does it take to get a first SEO job?

It varies too much by person and location to give a figure. The 90 days in this plan get you to the point of having fundamentals and evidence to show. The job hunt after that depends on your local market, your background and how closely your portfolio matches the roles you apply for.

05Should I pay for an SEO course?

Start with the free primary material: Google's SEO Starter Guide, How Search Works, Search Essentials, the Search Console help centre and Google's curated YouTube learning paths.¹³ Once you know what the providers actually say, you will be better placed to judge whether a paid course adds anything, and to spot one that contradicts the documentation.

06Can I use AI to write the content on my lab site?

You can use it to help with outlines, editing and drafts, but the value of the lab site is in content from your own experience. Content that recycles what already exists, or that a generative AI model could easily produce, adds little, and Google's spam policies cover using AI to generate many pages without adding value.³⁷ An employer reading your site will also be judging your writing.

07Is it fine to publish an audit of someone else's website?

Auditing a public site from the outside is common practice for portfolios, as long as you crawl gently, respect robots.txt, use only public information and write in a neutral tone. Consider anonymising the business or asking permission. Do not send it to the business as an unsolicited sales pitch; unsolicited SEO emails are a recognised warning sign for site owners.¹

08Which specialism pays best or has the most jobs?

I would not choose on that basis, and I have not seen official figures good enough to quote. Choose the work you enjoy enough to get good at. Most people start as generalists and specialise after a year or two, once they know which problems they like solving.

References69 sources

Sources

Platform docs58
Regulator8
Industry study3

01Do you need an SEO?Google Search CentralPlatform docs
02Search Engine Optimization (SEO) Starter GuideGoogle Search CentralPlatform docs
03Optimizing your website for generative AI features on Google SearchGoogle Search CentralPlatform docs
04Bing Webmaster GuidelinesBing Webmaster Tools HelpPlatform docs
05AI features and your websiteGoogle Search CentralPlatform docs
06Google users are less likely to click on links when an AI summary appears in the resultsPew Research CenterIndustry study
07Spam policies for Google web searchGoogle Search CentralPlatform docs
08Google Search's guidance on using third-party SEO tools, services, and adviceGoogle Search CentralPlatform docs
09In-depth guide to how Google Search worksGoogle Search CentralPlatform docs
10Google Search EssentialsGoogle Search CentralPlatform docs
11Google Search technical requirementsGoogle Search CentralPlatform docs
12Creating helpful, reliable, people-first contentGoogle Search CentralPlatform docs
13New learning paths: curated YouTube playlists for Search fundamentals and Technical SEOGoogle Search Central BlogPlatform docs
14FAQ about Google Trends dataGoogle Trends HelpPlatform docs
15Use Keyword PlannerGoogle Ads HelpPlatform docs
16HTML elements referenceMDN Web DocsPlatform docs
17Influencing your title links in search resultsGoogle Search CentralPlatform docs
18The rise of the AI crawlerVercelIndustry study
19RFC 9110: HTTP SemanticsIETFRegulator
20How HTTP status codes affect Google's crawlersGoogle for DevelopersPlatform docs
21RFC 9309: Robots Exclusion ProtocolIETFRegulator
22Introduction to robots.txtGoogle Search CentralPlatform docs
23Sitemaps XML formatsitemaps.orgRegulator
24Learn about sitemapsGoogle Search CentralPlatform docs
25How to specify a canonical URL with rel="canonical" and other methodsGoogle Search CentralPlatform docs
26SEO: 2024 Web AlmanacHTTP ArchiveIndustry study
27OverviewChrome for DevelopersPlatform docs
28Screaming Frog SEO Spider Website CrawlerScreaming FrogPlatform docs
29Rich Results TestSearch Console HelpPlatform docs
30About PageSpeed InsightsGoogle for DevelopersPlatform docs
31SEO Spider ConfigurationScreaming FrogPlatform docs
32SEO Spider PricingScreaming FrogPlatform docs
33Cloudflare RegistrarCloudflarePlatform docs
34Free WHOIS PrivacyPorkbunPlatform docs
35WordPress HostingKrystalPlatform docs
36AboutWordPress.orgPlatform docs
37New XML Sitemaps Functionality in WordPress 5.5Make WordPress CorePlatform docs
38Why Astro?Astro DocsPlatform docs
39OverviewClaude Code DocsPlatform docs
40Cursor DocumentationCursorPlatform docs
41Cloudflare PagesCloudflarePlatform docs
42Pricing and PlansNetlifyPlatform docs
43What is GitHub Pages?GitHub DocsPlatform docs
44Add a website or platform property to Search ConsoleSearch Console HelpPlatform docs
45Verify your site ownershipSearch Console HelpPlatform docs
46Set up Analytics for a website and/or appGoogle Analytics HelpPlatform docs
47Mark events as key eventsGoogle Analytics HelpPlatform docs
48Connect Search Console to Google AnalyticsGoogle Analytics HelpPlatform docs
49What are the exceptions?Information Commissioner's OfficeRegulator
50Easily Verify Your Site in Bing Webmaster Tools by Importing from Google Search ConsoleBing Webmaster BlogPlatform docs
51Bing Webmaster Tools or Google Search Console: A ComparisonBing Webmaster BlogPlatform docs
52Introducing AI Performance in Bing Webmaster Tools Public PreviewBing Webmaster BlogPlatform docs
53Overview of OpenAI CrawlersOpenAIPlatform docs
54Perplexity CrawlersPerplexityPlatform docs
55Does Anthropic crawl data from the web, and how can site owners block the crawler?Claude Help CenterPlatform docs
56Google's common crawlersGoogle for DevelopersPlatform docs
57Search generative AI controlSearch Console HelpPlatform docs
58Performance report (Search results): Overview and basic setupSearch Console HelpPlatform docs
59About Search Console dataSearch Console HelpPlatform docs
60Generative AI performance report (Search)Search Console HelpPlatform docs
61[GA4] Analytics AcademyGoogle Analytics HelpPlatform docs
62Computer Misuse Act 1990, section 3legislation.gov.ukRegulator
63Fraud Act 2006, section 2legislation.gov.ukRegulator
64Google Search Central BlogGoogle Search CentralPlatform docs
65Google Search Status DashboardGooglePlatform docs
66Working for yourselfGOV.UKRegulator
67Tax-free allowances on property and trading incomeGOV.UKRegulator
68Managing owners, users, and permissionsSearch Console HelpPlatform docs
69A deep dive into Search Console performance data filtering and limitsGoogle Search Central BlogPlatform docs

How to start an SEO career in 2026

Before you start

Understand the job and choose a route in

What changed with AI Overviews and assistants, and what did not

Where SEOs work

The specialisms

What AI tools have made cheap, and what they have not

Choose a route in

Score your starting point and learn how search works

Score yourself before you start

Turn your score into a week-by-week plan

Read the primary material in this order

Use an assistant as a tutor, with the document in front of it

Two free research tools, and what their numbers mean

Exercises for weeks 1 and 2

Learn the technical foundations

Basic HTML

HTTP status codes

robots.txt

Sitemaps

Canonicals

The free toolkit, and how to start with each tool

Exercises for weeks 3 and 4

Build and run your own site as a lab

Choose the subject

Choose and register the domain

Pick a build route

Route A: WordPress

Route B: a static site built with an AI coding tool

Audit the build before launch

Set up Search Console

Set up GA4

Set up Bing Webmaster Tools

Decide what AI crawlers may do

Build, publish and watch

Learn what the data does and does not tell you

Build a portfolio that proves skill

1. The lab site case study

2. A written audit of a public site, done respectfully

What not to put in a portfolio or CV

3. A technical write-up

4. A content brief

Get the first role or the first client

Where to look

What interviewers test

A sample audit task, and how to answer it

Talking about results honestly

If you are going freelance

Do the first year well and keep current

How to report

How to prioritise

Learn to say what is not worth doing

Keep current from the source

Common questions

Sources

Reading that goes deeper on parts of this tutorial

Stuck on a stage?