What is GEO: showing up in AI search

How Google’s AI answers, ChatGPT, Perplexity and Copilot find and cite pages; what the documentation says, what it leaves out, and what to do in the first month.

· 12 min readSEO

Short answer

GEO (generative engine optimisation) is the work of making a page fit to be cited as a source in ChatGPT, Perplexity, Copilot and Google’s AI answers. Its footing is the same as SEO: a page that is crawlable, indexed and clearly written. No provider guarantees a citation.

In a meeting, your customer says: “When I was looking for a supplier I asked ChatGPT first, and it named three companies.” That evening you type the same question. The answer lists three of your competitors with source links beside them; your site is not there. For the same query on Google, though, you are on the first page. GEO, short for generative engine optimisation, deals with this gap: making a page fit to be shown as a source inside the answer in AI search. In this article we answer the question “what is GEO” with what Google, OpenAI, Perplexity, Anthropic and Bing say in their own documentation. Where the documents are silent, we say that plainly too.

What GEO and AEO mean

GEO and AEO (answer engine optimisation) are two names for the same work. The subject is the systems that respond to a question with a written answer in place of a list of links: Google’s AI Overviews and AI Mode, ChatGPT search, Perplexity, Copilot, Gemini and Claude.

The one official document that uses the term is Bing’s. According to the Bing Webmaster Guidelines, GEO focuses on content’s eligibility for grounding and reference in AI responses. The same document adds straight afterwards: SEO does not guarantee rankings or traffic, and GEO does not guarantee grounding or citations.

How an AI answer finds your page

Before writing an answer, these systems retrieve pages from the web. They retrieve them either from a search index or from pages gathered by their own crawlers. The table below shows where each system looks and what you can manage.

Answering system Where it gets the page Your control Note in the documentation
Google AI Overviews and AI Mode Google’s search index robots.txt for Googlebot; nosnippet, max-snippet, noindex The page must be indexed and eligible to be shown with a snippet; no additional technical requirements
ChatGPT search Pages crawled by OAI-SearchBot OAI-SearchBot in robots.txt A blocked site is not shown in search answers; a robots.txt change takes about 24 hours to take effect
Perplexity Pages crawled by PerplexityBot PerplexityBot in robots.txt Not used to train foundation models
Claude Claude-SearchBot and Claude-User These two names in robots.txt If blocked, the site’s visibility may be reduced
Copilot The Bing index robots.txt for Bingbot; NOINDEX, NOARCHIVE, NOCACHE NOARCHIVE prevents content from being used in Copilot responses
Google’s systems outside Search, such as Gemini Content crawled by Google Google-Extended in robots.txt Limits training and grounding; the AI features in Search are managed through Googlebot

The crawler names stand for three separate jobs, and mixing them up is the most common mistake:

  • Search crawlers: OAI-SearchBot, PerplexityBot, Claude-SearchBot. Appearing as a source in answers depends on these.
  • Fetchers acting for a user: ChatGPT-User, Perplexity-User, Claude-User. They run when someone asks the assistant a question or gives it an address. OpenAI writes that robots.txt rules may not apply to these requests, and Perplexity that they are generally ignored.
  • Training crawlers: GPTBot and ClaudeBot. They collect content that may be used in model training.

According to OpenAI’s document, these settings are independent of one another: you can disallow GPTBot and leave OAI-SearchBot allowed. The same document does not say whether ChatGPT search also draws on other search indexes; it describes only its own crawler.

There is one more detail worth knowing on the Google side. According to the document, AI Overviews and AI Mode may split the question into subtopics and run several related searches (“query fan-out”), and so show a wider and more diverse set of links than a classic search.

SEO and GEO: common ground and the difference

Google’s document is plain: SEO best practices remain relevant for AI features, and there are no additional requirements or special optimisations needed to appear in AI Overviews or AI Mode. Bing says the same: Bing and Copilot rely on the same crawling, indexing and ranking foundation as traditional search. This is why your place in classic search still matters; a page that is not in the index, or does not make it into the retrieved results, cannot be a source for the answer.

The difference lies in two places. In classic search the result is a ranked list of links, and success is measured by position, impressions and clicks. In an AI answer your page becomes a source inside the answer, and the measure shifts to citations, mentions and referrals.

The common ground of SEO and GEO and where they part: the form of the result and the measure Classic search (SEO) Result: a ranked list of links Measure: position, impressions, clicks Common ground of both Crawlable page In the index Clear, useful content Structured data AI answer (GEO) Result: a source inside the answer Measure: citations, mentions, referrals Neither guarantees a ranking or a citation The common ground of SEO and GEO and where they part: the form of the result and the measure Classic search (SEO) Result: a ranked list of links Measure: position, impressions, clicks Common ground of both Crawlable page In the index Clear, useful content Structured data AI answer (GEO) Result: a source inside the answer Measure: citations, mentions, referrals Neither guarantees a ranking or a citation
Figure: SEO and GEO rest on the same technical and content footing; they part in the form of the result and in what success is measured by.

In practice this means individual passages do the work more than the page as a whole. In Bing’s words, content should be easy to understand without external context.

What the documents say helps

  • A crawlable page. Google asks that crawling is allowed both in robots.txt and by any CDN or hosting infrastructure, and that important content is available as text. Bing says critical content should not be hidden in parts that are only drawn in the browser. We set out these checks step by step in the technical SEO checklist.
  • Allowing AI crawlers. To appear in answers, OpenAI and Perplexity recommend allowing their search crawlers in robots.txt and through the IP ranges they publish.
  • The answer first. According to Bing, essential information should sit near the top of the page, and long introductions before the main topic should be avoided.
  • Information that can be verified without outside help. Bing wants facts and definitions stated explicitly and important information visible on the page itself. Writing the source next to a number is our own way of applying this principle; the document does not require it separately.
  • Clear structure. Bing recommends a logical heading hierarchy and semantic HTML elements. Putting a comparison in a table instead of burying it in a paragraph fits this, but none of the documents mentions tables by name.
  • One topic, consistent naming. According to Bing, addresses focused on one primary topic are more likely to be selected for grounding; the names of people, organisations, products and places should be written the same way everywhere.
  • Structured data. Google writes that no special schema.org markup is needed for AI features and that markup should match the visible text. Bing says structured data may support grounding but does not guarantee visibility.
  • Being mentioned elsewhere. Bing writes that links from relevant sites support discovery and the evaluation of authority. These documents say nothing about the effect of brand mentions that carry no link.

What is unknown and out of your hands

  • There is no guarantee of a citation. Bing says so outright. Google also notes that even for a page that meets every requirement, crawling, indexing or serving is not guaranteed.
  • No selection formula has been published. There is no document that explains why a given page made it into a given answer. Google writes that AI Overviews and AI Mode may use different models and techniques, that the links they show therefore vary, and that AI Overviews often do not trigger at all.
  • Answers are not fixed. A generative model does not have to give the same answer to the same question every time. Do not draw a conclusion from a single try.
  • Training and search are separate things. These documents do not say whether allowing GPTBot or ClaudeBot has any effect on being cited in answers.
  • There is no magic measure. None of the documents we read gives an ideal paragraph length or word count.

How to measure it honestly

  1. Referrals in Analytics. Look at the session source in the traffic acquisition report; visits from AI tools appear under that tool’s domain name. Some visits may arrive without referrer information, as direct traffic, so the figure is a lower bound. For the report to work properly, the steps in the GA4 setup checklist need to be complete.
  2. Search Console. According to Google’s document, sites appearing in AI Overviews and AI Mode are included in overall search traffic and reported in the Performance report within the “Web” search type. The document does not describe a separate report or filter that splits this traffic out; in other words, you cannot read the share coming from AI answers there in isolation.
  3. Bing Webmaster Tools. The guidelines recommend looking beyond clicks: content may appear in Copilot responses as impressions, citations or grounding references, and a decline in clicks does not always mean a loss of visibility. The AI Performance report in the tool shows which of your pages are cited.
  4. Manual prompt checks. Write down ten or so questions your customers ask and put the same questions to the same tools every month. Note three things: was your brand mentioned, was your site cited as a source, and was the information given correct.

The “AI visibility score” in third-party tools is not data published by the providers; it is an estimate based on the tool’s own sample queries. It can be used to follow a trend, but it is not enough to base a decision on.

First-month checklist

  1. Open yoursite.com/robots.txt. Googlebot, Bingbot, OAI-SearchBot, PerplexityBot and Claude-SearchBot should not be blocked. Decide separately on GPTBot, ClaudeBot and Google-Extended.
  2. Check the bot-blocking setting in your CDN or firewall. Even when robots.txt allows them, this setting can stop crawlers at the door.
  3. Verify the site in Search Console and Bing Webmaster Tools; check that your important pages are in both indexes.
  4. Get the ten questions customers ask most often from the sales team. We explained how to draw up the topic list in the B2B content plan guide. Put the questions to the tools and write the result in a table; this is your baseline.
  5. Go over the five most important pages that answer these questions: the answer in the first paragraph, each page on one topic, definitions and numbers explicit, comparisons in a table.
  6. Write the company name, product names and contact details the same way across the site and on outside profiles. We gathered what the about and contact pages should contain in what a company website needs.
  7. Check that the structured data carries the same information as the text visible on the page.
  8. At the end of the month, ask the same ten questions again and compare them, together with the referrals in Analytics, against the first table.

Common mistakes

  • Blocking every AI crawler wholesale in robots.txt without separating training from search.
  • Trusting a package that sells a guaranteed citation or “first place in ChatGPT”. We explained what to look for in a proposal in how to choose an SEO agency.
  • Adding hidden text or instructions to a page to steer AI models. Bing treats this as grounds for reduced visibility or removal from search.
  • Stuffing keywords and writing in artificial language in the hope of triggering a citation.
  • Forgetting a nosnippet or NOARCHIVE tag left over from an old template.
  • Dropping SEO and turning to GEO alone. The two share the same footing.

We followed the same order on our own site: our robots.txt file is open to search and answer crawlers. Every article starts with a short answer that can be read without the rest and ends with the sources it relies on. In our SEO service the first month goes to the technical audit; there is no point discussing AI visibility before crawling and indexing problems are closed.

Common questions

Is GEO replacing SEO?

No. Both Google and Bing write that AI answers rest on the same crawling and indexing foundation as search. GEO is a layer of attention added on top of SEO: allowing AI crawlers, passages that put the answer first, explicit definitions and consistent naming.

Do I need an llms.txt file?

Google writes that you don’t need to create new machine-readable files or AI text files to appear in its AI features. The crawler documents from OpenAI, Perplexity and Anthropic also ask the site owner for nothing beyond a robots.txt setting. Adding the file does no harm, but we have not read an official document saying it has an effect.

How long does it take to appear in AI answers?

No provider gives a timescale. Two things are known: OpenAI writes that a robots.txt change can take about 24 hours to be reflected in search results; Google notes that recrawling a page can take anywhere from several days to several months. Whether your page then makes it into an answer is not in your hands.

If you do only one thing today, open your robots.txt file and search it for the names OAI-SearchBot, PerplexityBot and Claude-SearchBot; if they sit under a Disallow: / line, your first job is clear. Then type the question your customers ask most often into an assistant and note the sources under the answer. That list tells you which of your pages to go over this month.

Sources

This article is general information, not legal, financial or technical advice. Platforms, their rules and the law change over time, so some details may have changed since 10 November 2026 or contain small errors. Before you make a decision for your business, check the current official sources or write to us.

Tell us about
your project.

Fill in the form or write to us directly, and we will agree on a day to talk.

Your first meeting is with Deniz Sezer, founder of Detmore.

First meeting free · Reply within one working day · Written proposal in 2 to 3 days · NDA ready

What do you need? pick as many as you like

Name, email and phone are required.

Contact

We reply within one working day.