serkankaya.
Türkçe sürüm
Let’s work together ↗

What Is GEO? Getting Cited as a Source in AI Search

What is GEO, how is it different from SEO, and how do ChatGPT, Perplexity and AI Overviews pick sources? A practical guide to setup, content and measurement.

GEO (Generative Engine Optimization) is the work of getting your content selected and cited as a source in AI answers from ChatGPT, Perplexity, Gemini and Google AI Overviews. The short answer to “what is GEO” is this: SEO gets your page into the list, GEO gets the information on your page into the answer.

In this article I treat GEO as an engineering problem rather than a marketing buzzword: how does a generative engine break down a question, which pages does it read, which passage does it pick, and which link in that chain can we influence? Having moved from software development into SEO and then performance marketing, I will also try to show where those three fields meet in GEO.

Where does the term GEO come from?

The term spread with an academic paper titled “GEO: Generative Engine Optimization”, published in late 2023.[1] The researchers defined search systems that gather information from several sources and summarise it with a language model as “generative engines”, and pointed out that content creators have almost no control over when and how they appear in these answers.

What makes the paper important is that it defines visibility as a measurable goal and compares different content strategies on a test set.

Visibility increase in generative engine answers with GEO strategies
Up to 40%
The effect of each strategy differs from field to field
Varies by domain
Source: Aggarwal et al., GEO: Generative Engine Optimization (arXiv:2311.09735)

This number should not be read as a guarantee. The study ran in a controlled test environment, and the same paper clearly says the effect varies by domain, which means each field needs its own approach. My takeaway: there is no single universal GEO tactic, but the structure of the content and the level of evidence really do make a difference.

Alongside GEO you will also see AEO (Answer Engine Optimization), AI SEO and LLM SEO. They all point to roughly the same goal. I use GEO on this site because it is the only one of these terms with academic roots and a defined measurement framework.

How do generative engines pick source pages?

To understand GEO, you first need to see how the answer is produced. The details differ between platforms, but the overall flow is similar:

User question Sub-queries (fan-out) Page and passage selection Answer with sources
Most generative engines split the question into sub-queries, fetch pages, pick passages and present the answer with source links.

1. Breaking down the question

Google officially says that AI Overviews and AI Mode may use a “query fan-out” technique when generating answers:[2] many related searches are run across subtopics for a single question. In practice, the answer given to someone who asks “what is GEO” is also fed by the results of sub-queries such as “GEO vs SEO” and “how to measure GEO”. Pages that answer a topic’s whole question tree, not just one keyword, get more chances.

2. Fetching pages

The pages used for the answer come either from a search index or from live crawling. Here it is important to tell two separate doors apart:

  • Crawling for search and citation: According to OpenAI’s documentation, OAI-SearchBot is used to show websites in ChatGPT’s search features. On the Perplexity side, PerplexityBot plays the same role.
  • Crawling for model training: Bots like GPTBot collect content as training data. OpenAI explicitly states that a site can allow OAI-SearchBot and block GPTBot and still appear in search.

3. Passage selection and citation

The model does not use the whole page, only the section that best answers the question. Paragraphs that make sense on their own, contain a clear claim and carry their context with them are better candidates. Passages that start with “as mentioned above” and cannot be understood without the previous section are easily dropped.

The difference between GEO and SEO

Short answer

What is the difference between GEO and SEO?

SEO gets a page ranked in search results; GEO aims to get the information on that page into AI answers from ChatGPT, Perplexity and AI Overviews as a cited source. GEO is not an alternative to SEO but a layer built on top of it: a page that can’t be crawled or ranked has little chance of being cited in an AI answer.

Almost every competing article makes this comparison, but most present the two disciplines as alternatives. I think the right model is a “layer”: GEO is built on top of a solid SEO foundation.

Classic SEO

  • Goal: a top position on the results page
  • Unit: page and URL
  • Success: impressions, clicks, rankings
  • Optimisation: keywords, technical health, backlinks

GEO

  • Goal: being a source inside the answer
  • Unit: passage and claim
  • Success: citations, brand mentions, AI referral traffic
  • Optimisation: question tree, evidence, entity consistency
Both disciplines share the same infrastructure; what differs is how success is measured and the unit content is optimised for.

Google’s own documentation supports this layer model: to be shown as a supporting link in AI Overviews or AI Mode, a page only needs to be indexed and eligible to be shown with a snippet in Google Search, and Google states that there are no additional technical requirements or special optimisations. So you cannot start GEO by skipping technical SEO. I explain how I build that foundation on my SEO services page.

Technical GEO: what to do on the infrastructure side

This is where competing content is weakest. Most articles say “use structured data” and move on. Looking at it as a developer, there is a concrete list to check.

  1. Make a deliberate decision about bot access

    In robots.txt, handle search and citation bots (OAI-SearchBot, PerplexityBot) separately from training bots (such as GPTBot). Make sure you are not accidentally blocking the search bot of a platform you want to appear on.

  2. Check your CDN and WAF rules

    Bot protection settings in layers like Cloudflare can block these bots even when robots.txt allows them. Google also reminds site owners that crawl permission has to be granted in the CDN and hosting setup, not just in robots.txt.

  3. Serve content as HTML

    Main content that only loads through client-side JavaScript may not be processed the same way by every bot. Static generation or server-side rendering makes the content available in the first response.

  4. Keep structured data accurate

    Schema types such as Article, FAQPage, Organization and Person make it machine-readable who wrote a page and what it is about. The schema must match the information visible on the page exactly.

  5. Tie entity information to one source

    Brand name, person name, job title and contact details should be written the same way across the site, social profiles and directories. A model can read conflicting information from different sources as a trust problem.

Technical GEO checklist0/6

Do you need llms.txt?

llms.txt is a proposed file for presenting a site’s summary and key pages to language models in Markdown. It is a proposal, not yet a standard, and no major AI search product has officially said it uses the file as a ranking signal. Still, because the cost is low, I add it: this site’s /llms.txt file is generated automatically at build time from published articles and service pages, so I never have to update it by hand.

How do you write GEO-friendly content?

Technical infrastructure makes sure the page can be read; the content decides whether it gets selected. These are the rules I follow:

Put the answer first

The first paragraph of every page should answer the question in the title directly, in two or three sentences. The same rule applies to ## headings: if a heading is a question, the first sentence under it should be the answer. This helps with both Google’s featured snippets and passage selection in generative engines.

Write passages that stand on their own

If you cut a paragraph out of the article and read it alone, does it still make sense? Using the term itself instead of a pronoun the first time it appears, spelling out abbreviations and giving each paragraph a single claim make a passage quotable.

Cover the question tree

Because of the fan-out logic, you need to answer the cluster of questions around a question, not just the question itself. Gathering the main topic in a pillar article, going deeper on sub-questions in separate articles and connecting them with internal links is a structure that works for SEO and GEO at the same time. I built the article plan for this site with the same logic.

Show your evidence

If you use numbers, dates and claims, state the source, and link to official documentation or the primary source where possible. As the academic GEO study also points out, sourced and specific statements are worth more than vague generalisations. Made-up statistics may get attention in the short term, but they damage trust in an environment where the model cross-checks against other sources.

Add first-hand experience

Generative engines already know the general information. What gives them something new is work done in the field: an error you hit during a setup, a platform’s undocumented behaviour, a constraint specific to one industry. I deliberately add notes like these from the background I describe on my About page.

How do you measure GEO?

Measurement is the hardest part of GEO, because there is no standard report like Search Console that covers every platform. I use a three-layer approach:

  1. Question-list test: I build a list of 20 to 30 questions the target audience would ask and ask them at regular intervals in ChatGPT, Perplexity, Gemini and Google AI Mode. I record in a table whether the brand is mentioned, which page is cited and which competitors appear. Because answers can change with personalisation and model updates, I draw conclusions from the trend, not from a single test.
  2. Referral traffic: In GA4 I group sessions from sources such as chatgpt.com, perplexity.ai and gemini.google.com into a separate channel group. This traffic may be small in volume, but its intent is usually high.
  3. Conversion link: I track whether AI-referred sessions turn into form submissions, appointments or sales. Visibility alone means little; to see whether it turns into business, the tracking setup has to be right. This part is the same tracking infrastructure I build on the performance marketing side.

What to watch out for when doing GEO in Turkey

Some situations for Turkish-language content are not covered in English guides:

  • The pool of Turkish sources is smaller. On many topics there are few high-quality Turkish sources, so a well-structured, well-sourced Turkish page can compete against fewer rivals than an English page on the same topic.
  • Terminology is messy. Users search for the same thing with different phrases such as “AI SEO”, “yapay zeka SEO’su” or “showing up in ChatGPT”. Covering these variations naturally in headings and the FAQ increases the chance of a match during fan-out.
  • KVKK and consent management. Measuring AI-referred traffic still has to follow cookie consent rules under KVKK, Turkey’s data protection law. Because data from users who decline consent is missing, GA4 numbers are better read as a trend than as exact values.
  • Local entity information. For businesses serving Turkey, keeping information on Google Business Profile, local directories and industry platforms consistent with the website helps the model recognise the brand correctly.

What GEO is not

A “quick fix” market has grown up around GEO very quickly, so it is worth drawing clear boundaries:

  • It is not hidden text or instructions aimed at the model. Adding text that users cannot see and that speaks to AI conflicts with search engines’ spam policies and can damage trust permanently.
  • It is not a one-off project. Models, indexes and answer formats keep changing. A page cited as a source today can drop off when its content ages or a better source appears.
  • It is not a new name for keyword stuffing. Repeating the same phrase over and over in a passage hurts readability and does not make the passage more likely to be selected. A clear answer, the right term and evidence work better.
  • It is not a guarantee. As Google’s documentation also states, even a page that meets every requirement is not guaranteed to be crawled, indexed or shown. GEO increases the probability; it does not make the outcome certain.

Where should you start?

You do not need to buy a new tool to start with GEO. First make sure your current site is crawlable, indexable and technically healthy, and check that your robots.txt and CDN settings do not block the search bots of the platforms you want to appear on. Then take your 10 most important pages, rewrite the first paragraph as a direct answer, make passages stand on their own and back your claims with sources. Finally, build a question list and take a first measurement; when you ask the same list again three months later, you will start to see what worked.

In short, GEO is not a new discipline that replaces SEO; it is a layer on the same foundation that measures success by citations as well as clicks. Structuring content so that machines can read it is the software side’s job: static generation, clean HTML and correct structured data are the foundation this site is built on too. If that side interests you, take a look at my software development page.

Key terms

GEO
GEO (Generative Engine Optimization) is the work of getting content selected and cited as a source in AI answers from platforms such as ChatGPT, Perplexity, Gemini and Google AI Overviews.
Query fan-out
Query fan-out is when a generative search system splits a single user question into subtopics, runs a separate related search for each one and builds the answer from those results. Google says it can use this technique in AI Overviews and AI Mode.
AEO
AEO (Answer Engine Optimization) is optimisation aimed at getting content used in the direct answers given by search engines and AI assistants. In practice it points to largely the same goal as GEO.
llms.txt
llms.txt is a proposed file for presenting a site’s summary and key pages to language models in Markdown. It is not an official standard yet, and no major AI search product has officially said it uses the file as a ranking signal.
OAI-SearchBot
OAI-SearchBot is the crawler OpenAI uses to surface websites in ChatGPT’s search features. It is separate from GPTBot, which collects training data; a site can block GPTBot and allow OAI-SearchBot and still appear in search.

Frequently asked questions

What is GEO?

GEO (Generative Engine Optimization) is the work of getting your content used and cited as a source in AI answers from ChatGPT, Perplexity, Gemini and Google AI Overviews. The goal is less about ranking higher and more about being the source that is named and linked inside the answer.

What is the difference between GEO and SEO?

SEO aims to get a page listed near the top of search results, while GEO aims to get the information on that page pulled into the generated answer. In practice GEO is a layer built on top of SEO: a page that cannot be crawled and indexed cannot be a source in AI answers either.

Does GEO need a special schema or file?

Google states in its official documentation that there are no additional technical requirements or special schema for AI Overviews and AI Mode. Structured data and files like llms.txt can help content be understood, but they do not guarantee visibility on their own.

How do you measure the results of GEO?

Instead of a classic rank report, you build a list of target questions, ask them regularly in ChatGPT, Perplexity, Gemini and Google, and record whether the brand is mentioned and which page is cited. On top of that, referral traffic from AI platforms is tracked as a separate channel in your analytics.

Will GEO replace SEO?

Not in the short term. Most generative engines draw on search indexes and live web crawling to build their answers, so an SEO foundation remains a prerequisite for GEO. What changes is the definition of success: alongside clicks, citations and brand mentions are now outcomes worth measuring.

Sources

  1. GEO: Generative Engine Optimization (Aggarwal et al., arXiv)
  2. Google Search Central: AI features and your website
  3. OpenAI: Overview of OpenAI Crawlers
  4. Perplexity: Perplexity Crawlers
  5. The llms.txt proposal