Updated
Updated What changed on
- Added the honest range on AI Overview prevalence, because the single figures quoted in most guides disagree with each other by a factor of nearly three.
- Added Pew Research's click-through measurement.
Can I do GEO myself with ChatGPT? About 60% of it, yes.
The 30-second answer
Roughly 60% of GEO work is genuinely DIY-able with free tools and 2 to 4 hours a week: baselining, schema basics, llms.txt, bot access audit, monthly query sampling, restructuring your top pages. The 40% you cannot easily DIY is the strategic work, the at-scale rewrites, and the diagnostic work when citations stop landing. For solo founders and small UK businesses, DIY makes sense up to the point AI search becomes a real lead channel (around 5% of revenue-influenced traffic). For mid-market and up, the consultant hour pays back faster than the founder hour.
Numbers worth knowing before you start
of Google searches trigger an AI Overview, depending entirely on whose panel you read.
How it was measured: Pew measured 18% on real-user browsing data. Ahrefs measured 48% of its keyword set in March 2026. Similarweb reported 43%, up from 15% a year earlier, via TechCrunch on 27 July 2026. The panels sample different query mixes, so the spread is a methodology artefact. Anyone quoting a single figure as the number is overstating what is known.
Sources disagree on this one. The range is the honest answer.
is how often people clicked a search result when an AI summary was present, against when it was not. Roughly half the click rate.
How it was measured: Browsing data from 900 US adults who agreed to share their activity, covering March 2025. US only, one month, and now over a year old. It remains the largest published real-user measurement rather than a keyword-panel estimate.
of visits to a page carrying an AI summary resulted in a click on a source cited inside that summary.
How it was measured: Same 900-person US panel, March 2025. This is the figure that matters most for anyone selling AI citation as a traffic channel.
The four engines worth testing yourself
Manual citation sampling is the single highest-leverage DIY task. Open one browser tab per engine, run the same five queries on the first of every month, log who got cited. This is what every DIY guide skips and every consultant charges for.
ChatGPT
Perplexity
Claude
Gemini
AI Overviews
BraveThe DIY toolkit (all free, all browser-based)
| Job | Tool | Time per use |
|---|---|---|
| Baseline visibility | AI Search Visibility tool | 5 min |
| Crawler access audit | AI Bot Robots.txt Tester | 2 min |
| Schema validation | AI Schema Validator + Google Rich Results Test | 10 min/page |
| llms.txt starter | Firecrawl generator + hand-edit | 20 min once |
| WordPress schema | Yoast or Rank Math free plugin | 15 min setup |
| SEO baseline | Google Search Console | weekly review |
| Citation sampling | ChatGPT, Perplexity, Claude, Gemini + a Google Sheet | 30 min/month |
Total cash cost: zero. Total time: 2 to 4 hours a week for an organised owner, less once the setup work is done.
The 90-day DIY playbook
Run it in this order. Out of order, you waste time fixing pages that bots cannot even reach. If you want the wider context for each step before you start, the complete GEO field guide on this site walks through the same stages without the time-budget constraint.
Week 1: baseline (90 minutes total)
- Run the AI Search Visibility tool on your domain.
- Run the AI Bot Robots.txt Tester. Capture which bots are blocked.
- List your 5 highest-intent pages. Validate their schema.
Weeks 2 to 3: unblock and structure
- Unblock GPTBot, ClaudeBot, PerplexityBot, Google-Extended in robots.txt unless you have a legal reason not to.
- Add missing Article, FAQPage, Person and Organization schemas. ChatGPT writes the JSON-LD if you paste the page content and ask, then validate with the Rich Results Test before pushing.
Week 4: llms.txt
- Generate the starter with Firecrawl, then hand-edit to your 30 highest-value URLs only. Auto-generated full-site llms.txt files add no signal because they list every URL including thin tag pages.
Month 2: the 30-second answer pattern
- Add a 30-second answer block to the top 5 buyer-question pages. Direct answer in 2 to 3 sentences, then supporting detail. AI engines extract these blocks far more reliably than narrative paragraphs.
Month 3: monthly sampling, every month thereafter
- Pick 10 queries that real prospects ask. Run each on all four engines. Log: did your brand appear, what was the citation, who else got cited. One reading proves nothing - AI outputs vary 30 to 50% session to session. The trend over 3 months is what tells you whether your work is moving the needle.
What monthly citation sampling actually looks like
Worked example. Your domain is acme-roofing.co.uk. The query is "best commercial roofing contractors UK". Here is the spreadsheet row format that works:
One month of this across 10 queries shows you which engines are easy wins, which need work, and what each engine values (Perplexity loves source diversity, ChatGPT often defers to industry bodies, Gemini leans on Google's own SERP). No tool sold to consultancies does anything fundamentally different - they just automate the same loop.
Where ChatGPT itself genuinely helps
Five things ChatGPT does well as your DIY GEO assistant. Five things it cannot.
| Yes, use ChatGPT for | No, do it yourself |
|---|---|
| Basic schema JSON-LD generation | Strategic engine prioritisation |
| Rewriting paragraphs into 30-second answer blocks | Choosing which 50 pages to rewrite (depth audit) |
| Brainstorming FAQ questions from a topic | Reading the citation trend over 3 months |
| Drafting llms.txt entries from page summaries | Diagnosing why a specific competitor outranks you |
| Generating buyer-intent query lists for sampling | Anything requiring access to your GSC data |
The four failure modes I see in DIY GEO
- Skipping the bot access audit. Months pass, then someone notices GPTBot has been blocked since 2024 because a security plugin added it. OpenAI's bot docs and Anthropic's crawler list tell you exactly which user agents to permit.
- Auto-generated llms.txt left untouched. The file lists 400 URLs including tag pages. No engine treats that as signal. Trim to 30 high-value URLs, write the summary line by hand.
- Single-sample citation reads. One query, one answer, one conclusion is data theatre. AI outputs vary so much session to session that Search Engine Journal's 2026 strategy guide and most working GEO practitioners advise 3-month trend lines over snapshots.
- Optimising the homepage. AI engines almost never cite homepages for specific buyer questions. They cite the page that answers the specific question. Spend the work where the questions land.
When DIY stops paying back: three triggers
- AI referrals cross 5% of revenue-influenced traffic. Check the AI source segment in GA4 - chatgpt.com, perplexity.ai, gemini.google.com, claude.ai. At 5% it stops being a curiosity and starts being a channel, and channels deserve specialist attention.
- Your priority page count crosses 50. Rewriting 5 pages is a Saturday. Auditing depth, restructuring and re-validating 50 is a project, and the project takes weeks of founder time that earns less than the consultant fee.
- Citations still are not landing at month 4. If you have run the 90-day playbook, sampled, fixed, and citations remain flat, the bottleneck is usually entity strength or third-party signal, not on-page. Diagnostic takes experience.
Honest framing: the GEO industry has a strong financial incentive to tell you that this work is too complex to do yourself. The market is full of dashboards that mostly automate manual sampling. For solo founders and small UK businesses with one or two priority pages, DIY is realistic and the ceiling is much higher than the agency pitch suggests. The thing to hire out is the strategic call about where to focus - everything downstream is execution you can do.
Who should DIY, who should hire
| Profile | DIY or hire | Why |
|---|---|---|
| Solo consultant / sole trader | DIY | One or two priority pages, hours-a-week budget fits easily |
| Small UK service business (under 20 staff) | DIY + one scoping call | Playbook works, but a single hour on engine priorities pays back fast |
| SaaS at PMF, under 1m ARR | Hybrid | DIY the on-page; hire the strategic prioritisation and quarterly review |
| Established mid-market (over 50 staff) | Hire | Page count and time-cost flip the maths against DIY |
| Enterprise / regulated industry | Hire + internal owner | Risk, scale and compliance all require senior input |
Sources I trust on DIY GEO
- Heinz Marketing - 5 steps to make your website GEO-ready (B2B-focused, structure-first)
- Search Engine Journal - 5 GEO strategies to make AI search recommend you (March 2026)
- Reply - GEO practical guide
- HubSpot - GEO tools marketing teams use
- Profound - 2026 GEO tools review (useful to see what dashboards do, before you decide to pay for one)
More on DIY GEO.
Can I genuinely do GEO myself with free tools?
About 60% of it yes. Baselining, schema basics, llms.txt, bot access audit, monthly query sampling and basic page restructuring are all doable with free tools and 2 to 4 hours a week. The 40% you cannot easily DIY is the strategic work: which engines to prioritise for your specific buyer mix, content depth audits across 50+ pages, and the diagnostic work when citations stop landing.
What free tools do I actually need?
Google Search Console for SEO baseline. The free AI Search Visibility tool, AI Bot Robots.txt Tester and AI Schema Validator on this site for AI-specific checks. Firecrawl\'s free llms.txt generator. Yoast or Rank Math free plugin on WordPress for schema. ChatGPT, Perplexity, Claude and Gemini browser windows plus a Google Sheet for monthly citation sampling. Hour budget: 2 to 4 a week. Cash budget: zero.
How long until DIY GEO shows results?
Schema and bot access fixes can show up in AI answers inside 2 to 4 weeks because crawlers re-index quickly. Content restructuring takes 6 to 12 weeks because engines re-evaluate context slowly. Citation lift on a competitive query can take 3 to 6 months for a small site, longer if you have weak third-party mentions. If nothing has moved at month 4 across all four engines, the problem is usually entity gaps or weak third-party signal, not on-page work.
Can ChatGPT write my schema markup for me?
Yes for basic Article, FAQPage, Person and Organization schemas. Paste a page URL or content, ask for valid JSON-LD with the relevant @type, then run the output through Google\'s Rich Results Test before publishing. Where ChatGPT fails: nested entity relationships (HowTo with embedded ImageObject and tool lists), Product schema with offers, and anything requiring real persistent IDs. For those, hand-write or use the schema-aware plugins.
Where do most DIY GEO attempts go wrong?
Four failure modes. First, skipping the bot access audit and finding out months later that GPTBot or ClaudeBot is blocked. Second, auto-generating llms.txt then never hand-editing it, so the file lists every URL and adds zero signal. Third, sampling citations once and treating the result as data, when AI outputs vary 30 to 50% session to session. Fourth, optimising the homepage instead of the specific pages that answer buyer questions.
When should I stop DIY and hire someone?
Three triggers. First, AI search referrals cross 5% of revenue-influenced traffic in GA4 (real lead channel territory). Second, your content base passes 50 priority pages (the time to audit and restructure that many breaks the DIY hour budget). Third, you have run the 90-day playbook, citations are still not landing, and you cannot work out why. At that point a 90-minute scoping call usually identifies what is missing faster than another month of guessing.
Jason Burns
Independent UK SEO, GEO and AI consultant. 17 years in search. Portfolio includes 3M, BlackRock, Unilever and E.ON. Owner of SEO Moves Ltd since 2014.
More about Jason →DIY hit a ceiling and you want help?
Tell me what you have already done and where citations are not landing. I cover the 40% you cannot easily DIY - strategic prioritisation, scale audit, diagnostic - on a single call rather than a retainer if that is what you need.
Book a strategic call