How to Get Recommended by ChatGPT and Google AI (GEO)
AI answers are built from search results, so GEO is mostly SEO with a new scoreboard. Here is what Google, OpenAI and the research support, and what to skip.
Key takeaways
- ChatGPT search, Perplexity and Google's AI Overviews cite pages they find through search, so rankings and crawler access come before any AI-specific tactic.
- Allow OAI-SearchBot to appear in ChatGPT search; GPTBot only covers model training, so you can block it without leaving ChatGPT's answers.
- Only 38% of AI Overview citations rank in Google's top 10 for the original query (Ahrefs, March 2026), so cover the follow-up questions buyers ask.
- llms.txt and special schema don't earn citations: Google ignores llms.txt, and 97% of the files got zero requests in May 2026 (Ahrefs).
- Measure monthly: the same 20–30 buyer questions in each AI tool, plus Search Console's Generative AI report for AI Overviews and AI Mode.
Generative engine optimization (GEO) is the work of getting your business named and linked when ChatGPT, Perplexity or Google's AI Overviews answer a question, and most of it is SEO. Google's guide to generative AI search, updated July 2026, calls it "still SEO". These tools search the web before they answer, so the levers are crawler access, pages that answer the exact question and what other sites say about you.
This guide is for founders and marketing leads who ask ChatGPT for a provider in their niche and see competitors named instead. It covers how AI tools pick sources, which crawlers to allow, which GEO tactics hold up, how to measure AI visibility and what to fix first.
Sources are Google, OpenAI and Perplexity docs, two peer-reviewed studies and Ahrefs datasets, with correlations marked. Our examples are search projects; we have no measured AI-citation figures for them.
GEO tactics against the evidence: Google and OpenAI docs, two studies and Ahrefs data from 2024–2026, checked October 2026.
| Tactic | What the evidence says | Worth it? |
|---|---|---|
| Rank in regular search, including for follow-up questions | Google, ChatGPT and Perplexity answer from search results; 38% of AI Overview citations rank in Google's top 10 (Ahrefs, March 2026) | Yes: the base |
| Allow AI search crawlers | Sites opted out of OAI-SearchBot are not shown in ChatGPT search answers (OpenAI) | Yes: a quick fix |
| Earn mentions on other sites and YouTube | Branded web mentions correlate 0.66–0.71 with AI visibility; site page count about 0.19 (Ahrefs, December 2025) | Yes, but slow |
| Add statistics, quotes and sources | Up to 40% more visibility in lab tests on GPT-3.5 (GEO paper, 2024); mostly ineffective in a 2025 benchmark (C-SEO Bench) | For readers, not as a trick |
| Publish llms.txt | Google ignores it; 97% of the files got zero requests in May 2026 (Ahrefs) | Optional |
| Add special schema for AI | Not required; adding JSON-LD brought no major citation uplift (Ahrefs, May 2026) | For rich results only |
| Chunk pages, rewrite for AI, stuff keywords | Not needed, per Google; keyword stuffing did 10% worse than baseline on Perplexity in the GEO tests | No |
| A page for every phrasing of a question | Scaled content abuse when aimed at AI answers, per Google | No |
How do ChatGPT and Google AI choose which sources to cite?
They search first and write second. Google calls the method retrieval-augmented generation (RAG), or grounding: the model pulls pages from the search index, writes the answer from them and links the pages that support it.
To find those pages, the model runs query fan-out: related searches it generates from the user's question. In Google's example, "how to fix a lawn that's full of weeds" fans out into "best herbicides for lawns", "remove weeds without chemicals" and "how to prevent weeds in lawn". A page can be cited for ranking on one of those sub-questions, even if it doesn't rank for what the user typed.
Ahrefs data may show fan-out at work. In July 2025, about 76% of pages cited in AI Overviews ranked in Google's top 10 for the same query; by March 2026, across 863,000 searches and with improved parsing, it was 38%. Ahrefs suspects fan-out plays a bigger role since Gemini 3 began powering AI Overviews in January 2026.
ChatGPT works the same way. Ahrefs' study of ChatGPT prompts (April 2026) concluded that to be cited "you need to be in that search selection pool—which means your content needs to rank". Pages whose titles matched ChatGPT's internal sub-questions were cited more often, a correlation that points the same way as Google's fan-out.
Is GEO different from SEO?
Mostly no: the ranking work is the same, the scoreboard is new. Google's guide says it outright:
"From Google Search's perspective, optimizing for generative AI search is optimizing for the search experience, and thus still SEO." — Google Search Central, Optimizing your website for generative AI features on Google Search, July 2026
Answer engine optimization (AEO) is another label for the same work; our GEO and AEO glossary defines it and the other terms of AI search. Google's page on AI features and your website (December 2025) names one hard requirement: to be a supporting link in AI Overviews or AI Mode, a page must be indexed and eligible to show in Search with a snippet. The July 2026 guide adds a second: the site must be included in Search Console's generative AI features control, the default setting.
What changes is what winning looks like. In Pew Research Center's browsing data for 900 US adults (March 2025), people clicked a result on 8% of Google visits with an AI summary and on 15% without one, and a link inside the summary on 1%. Google says clicks from pages with AI Overviews are higher quality but publishes no figures for it.
So the goal moves from the click to the mention: your name and link in the answer, and a page that converts the visitor who does click.
Which AI crawlers should you allow in robots.txt?
Allow the search crawlers and decide on training crawlers separately: blocking a search crawler can take you out of that tool's answers, blocking a training crawler does not. OpenAI is explicit: "Sites that are opted out of OAI-SearchBot will not be shown in ChatGPT search answers, though can still appear as navigational links." Its settings are independent, so you can allow OAI-SearchBot and block GPTBot, per OpenAI's crawler documentation.
What blocking each crawler does. From OpenAI, Perplexity and Google documentation, checked October 2026.
| Crawler | What it does | If you block it |
|---|---|---|
| OAI-SearchBot (OpenAI) | Finds pages for ChatGPT search | Out of ChatGPT search answers; navigational links can remain |
| GPTBot (OpenAI) | Collects content that may train OpenAI models | Signals no training on your content; search unaffected |
| ChatGPT-User (OpenAI) | Opens a page for a user's request | robots.txt may not apply; not used for search inclusion |
| PerplexityBot | Finds pages for Perplexity search, not for model training | May not appear in Perplexity search results |
| Perplexity-User | Fetches a page for a user's question | Generally ignores robots.txt |
| Googlebot | Crawls for Google Search, AI Overviews and AI Mode included | Pages can't be crawled for Search or its AI features |
| Google-Extended (a robots.txt token, not a crawler) | Controls Gemini training and grounding in Gemini Apps and Vertex AI | Search and AI Overviews unaffected |
Blocking GPTBot or Google-Extended keeps your content out of model training without removing you from search answers, though Google-Extended also covers answers grounded in Gemini Apps.
How do you answer the exact question buyers ask AI?
Give each real buyer question one page that answers it in the first lines, with facts nobody else has. Fan-out finds the page for the sub-question; the facts give the model a reason to cite you.
We built this for Gastro.News, an Austrian food magazine. Diners don't search for a restaurant guide; they search for dinner in their district, such as top-rated restaurants in Vienna's 1070 district. So we built a page for every district: 32 ranked hubs for 9 states and 23 Vienna districts, plus 2,600+ restaurant pages from one template.
Each venue carries a Gastro.News Score that merges four guides (Michelin, Gault&Millau, Falstaff and A la Carte) into one number. Only Gastro.News publishes it, the opposite of what Google calls commodity content, "based on common knowledge, which could originate from anyone". Titles follow how people search, with menu, prices, photos and reviews.
Gastro.News draws 30K+ organic visitors a year. That is search traffic: we haven't measured AI citations for the site, so read it as proof that question-shaped pages rank, which a citation needs first. The full build is in our programmatic SEO case study.
Google draws a line here: separate pages for every variation of a query, fan-out queries included, made "primarily to manipulate rankings or generative AI responses" break its scaled content abuse spam policy. Each Gastro.News district page lists different restaurants; a page that only rewords the question adds nothing.
To find your questions, start with what buyers ask your sales team before they sign. Then put the main one to ChatGPT and Perplexity and note the sub-topics and sources the answers use.
Why do mentions on other sites matter more than page count?
AI answers repeat what the web says about a brand. In Ahrefs' study of 75,000 brands (December 2025), branded web mentions correlated 0.66–0.71 with how often ChatGPT, AI Mode and AI Overviews mentioned the brand. YouTube mentions correlated most, at about 0.74; the number of pages on the brand's own site, about 0.19.
Ahrefs notes that correlation isn't causation: famous brands get mentioned everywhere, AI answers included. Still, the work moves off your site: reviews where your buyers read them, industry directories, podcasts and videos, articles that compare options. Keep your name, services and cities consistent everywhere.
For local businesses and stores, Google's guide names two direct inputs: Google Business Profile and Merchant Center feeds help products and services show "in both AI responses and other Google Search results". It also warns that seeking inauthentic "mentions" "isn't as helpful as it might seem".
Do statistics and quotes make AI cite you more?
In lab tests they did; in a benchmark with competing pages, most such rewrites stopped working. Add numbers, quotes and sources because they make the page worth citing, not as a trick.
The paper that coined GEO (Aggarwal et al., KDD 2024) fed a GPT-3.5 Turbo answer engine the top five Google results per query and rewrote one page at a time. Citing sources, adding quotations and adding statistics raised that page's visibility by 30–40%, while keyword stuffing gave "little to no improvement". When all five pages were rewritten at once, pages ranked fifth gained most: citing sources lifted their visibility by 115.1%.
C-SEO Bench, a 2025 benchmark for conversational search engine optimization (C-SEO, its name for GEO), retested such methods across more tasks and domains, with several pages competing at once:
"Our experiments reveal that most current C-SEO methods are not only largely ineffective but also frequently have a negative impact on document ranking, which is opposite to what is expected." — Puerto et al., C-SEO Bench, NeurIPS Datasets and Benchmarks 2025
Ranking higher among the sources the model reads worked better. Gains also shrank as more pages adopted the same methods: a tactic every competitor uses stops being an edge.
Where we write at volume, specifics come first. Our content factory, an n8n pipeline (n8n is a workflow automation tool), turns new videos from tracked channels into articles for a media client under NDA. Drafts written from a video's title and description missed what was actually said, so each article is built on the full transcript, with a review step before anything goes live.
Our takeaway: most AI content fails at the angle, not the prose. Google's guide gives the search version: "Don't just recycle what others on the internet have already said, or could easily be produced by a generative AI model."
AI Content FactoryClient under NDA
Freshness shows a pattern too. Across 17 million citations, AI assistants cited pages 25.7% fresher than Google's organic results (Ahrefs, July 2025), AI Overviews excepted. Update facts and dates when they change, not to look new.
What doesn't work: llms.txt, chunking and AI-only rewrites
Most "AI-ready" tweaks change nothing measurable. Google lists them as things you can ignore, and independent data agrees.
llms.txt is a Markdown summary of a site for AI tools, proposed by Jeremy Howard in 2024. Google Search ignores it: the file "will neither harm nor help your site's visibility or rankings". In server logs of 137,000 sites, Ahrefs found 97% of llms.txt files got zero requests in May 2026, and SE Ranking's domain-level study (November 2025) found no correlation with AI citations.
Chrome's Lighthouse audit for llms.txt checks the file but says providing it is "optional at the moment". Keep one if it costs nothing; don't expect citations from it.
Chunking, cutting pages into small blocks for AI, isn't needed: "There's no requirement to break your content into tiny pieces for AI to better understand it," says Google, and there's no ideal page length. Nor do you need to "write in a specific way just for generative AI search."
Schema markup, structured data that describes a page for machines, still earns rich results, but Google says there's no special schema.org markup for AI. When Ahrefs tracked 1,885 pages that added JSON-LD between August 2025 and March 2026 against 4,000 control pages, citations showed no major uplift, yet cited pages were almost three times more likely to have JSON-LD. That is correlation without causation in plain view.
How do you check whether ChatGPT recommends your business?
Ask the AI tools your buyers' questions, the same set every month, and log who gets named and linked. For Google, add the Search Console report.
Google's Generative AI performance report shows how often links to your site appeared in AI Overviews and AI Mode, by page, country and device; as of August 31, 2026, it covers all websites. Clicks from these features sit in the regular Performance report under the "Web" search type, mixed with other search clicks.
For ChatGPT and Perplexity, run your own panel:
- Write 20–30 questions buyers ask before choosing a provider like you, with your city or niche.
- Run them monthly in ChatGPT with search, Perplexity and Google AI Mode, each time in a fresh session.
- Log if you are named and linked, which of your pages, and which competitors.
- Check server logs for OAI-SearchBot and PerplexityBot visits to those pages.
GEO checklist: what to fix first
Start with access, then your pages, then everything off your site.
- Indexing. Key pages are indexed and can show with a snippet; check in Search Console.
- Crawlers. OAI-SearchBot, PerplexityBot and Googlebot are allowed in robots.txt and the CDN. GPTBot and Google-Extended are a training decision.
- AI features control. The site is set to Include in Search Console's generative AI features control, which Google requires for eligibility.
- Questions. 20–30 buyer questions, each mapped to one page that answers it in the first lines.
- Proof. Your own figures, process, named experts and cases on those pages.
- Profiles. Google Business Profile and, for stores, Merchant Center are current.
- Mentions. Reviews, directories, videos and articles on other sites, none of them bought.
- Skip list. llms.txt as a strategy, chunking, AI-only rewrites and offers of guaranteed citations.
- Measurement. The monthly question panel plus Search Console's Generative AI report.
Our SEO and GEO services work through the same list, from crawler access to the monthly question panel. If a redesign is on your roadmap, protect the pages that already rank before anything else, with our guide to redesigning a site without losing SEO.
Questions we get asked
Can I pay ChatGPT to recommend my business?
Not inside the answer, as far as OpenAI's documentation goes. OpenAI sells ads in ChatGPT, and its OAI-AdsBot checks the landing pages of submitted ads, but the docs describe no way to buy a citation in an organic answer. Those links come from ChatGPT search, which finds pages through OAI-SearchBot. Check OpenAI's ads documentation for what is on offer today.
Does blocking Google-Extended remove my site from AI Overviews?
No. Google-Extended is a robots.txt token that decides if your content is used to train Gemini and to ground answers in Gemini Apps and Vertex AI; Google says it doesn't affect inclusion in Google Search. To keep pages out of AI Overviews and AI Mode, exclude the site in Search Console's generative AI features control, which takes effect in 1–2 days and removes that traffic too, or use nosnippet or noindex.
Does Perplexity respect robots.txt?
Its search crawler does. PerplexityBot finds pages for Perplexity's search results, isn't used to train AI models and follows robots.txt; changes can take up to 24 hours. Perplexity-User, which fetches a page when a person asks about it, generally ignores robots.txt because a user requested it. To appear, allow PerplexityBot and the IP ranges Perplexity publishes.
How long does it take to show up in AI answers?
Crawler settings apply quickly: OpenAI says about 24 hours for search after a robots.txt change, Perplexity up to 24 hours. A new or rewritten page then has to be recrawled and earn rankings, which takes weeks to months, and mentions on other sites build over months. Nobody can promise a date, because the same question can return different sources from run to run.
Do YouTube and Reddit help you get cited by AI?
They are among the most cited sources. Pew found Wikipedia, YouTube and Reddit were the most cited sources in Google's AI summaries in March 2025, and Ahrefs' Brand Radar showed YouTube as the most cited domain in AI Overviews in early 2026. In Ahrefs' brand study, YouTube mentions correlated most with AI visibility, at about 0.74. That is correlation: publish useful videos and answers where buyers look, not spam.
Can an agency guarantee my business appears in ChatGPT?
No. Answers change from run to run, and Google says no third-party tool has access to its internal ranking or AI systems. A credible agency promises the work instead: crawler access, pages that answer real buyer questions, mentions on other sites and a monthly report on a fixed set of questions. Treat a guaranteed placement in AI answers as a red flag.
Sources
- Optimizing your website for generative AI features on Google Search · Google Search Central, 2026
- GEO: Generative Engine Optimization · Aggarwal et al., KDD 2024, 2024
- Update: 38% of AI Overview Citations Pull From The Top 10 · Ahrefs, 2026
- Top Brand Visibility Factors in ChatGPT, AI Mode, and AI Overviews (75k Brands Studied) · Ahrefs, 2025
- We Analyzed 137K Sites: 97% of llms.txt Files Never Get Read · Ahrefs, 2026
- We Tracked 1,885 Pages Adding Schema. AI Citations Barely Moved. · Ahrefs, 2026
- Google users are less likely to click on links when an AI summary appears in the results · Pew Research Center, 2025
- New Study: AI Assistants Prefer to Cite “Fresher” Content (17 Million Citations Analyzed) · Ahrefs, 2025
Share this article
About the author
Artsiom Yankovich
Head of Design & Strategy
Leads design and strategy, from product logic to brand. 9+ years, 150+ projects.
