GEO, AEO and SEO versus paid ads in ChatGPT

Published 17 September 2026 · Updated 17 September 2026 · Based on OpenAI's published documentation

A brand can show up in ChatGPT in two different ways, and they run on two different systems. One is being cited inside an answer, for free, because a crawler found and matched your page. The other is a paid, labelled ad that you built, bid on and got approved. Search engine optimisation, answer engine optimisation and generative engine optimisation all aim at the first. This guide sets out what is actually documented about each side, where the line between them sits, and why most brands need both rather than one instead of the other.

What do GEO, AEO and SEO mean for ChatGPT?

SEO earns a ranking in a list of results, AEO earns a direct answer to a specific question, and GEO, generative engine optimisation, earns a citation inside an AI system's generated answer, such as one of the sources ChatGPT names when it searches the web to answer you. They are not three different technologies; they are the same discipline, aimed at three different surfaces.

In practice the overlap is large: clear headings, one self-contained answer per section, structured data, a site that loads without a login wall, and content that is actually correct all help a page qualify for a citation the same way they help it qualify for a ranking. What changes with GEO is the goal. You are not trying to win a click position; you are trying to be the source a language model reaches for when it assembles an answer, whether or not the reader ever visits your page afterwards. This guide's own key-facts box and FAQ, and every guide on this site, are written the same way for that reason: self-contained statements a model can quote directly.

How does ChatGPT decide what to cite in an answer?

ChatGPT's search features are powered by a dedicated crawler, OAI-SearchBot, which fetches public pages so they can be surfaced and cited in a search-style answer; OpenAI documents the crawler and its robots.txt controls, but has not published the ranking logic that decides which pages actually get cited.

OAI-SearchBot is one of several crawlers OpenAI operates, and each is controlled separately in robots.txt. GPTBot is used to crawl content for training OpenAI's generative models; disallowing GPTBot opts a site out of that training use only, and does not remove it from ChatGPT search citations, because OAI-SearchBot is a different bot with a different rule. ChatGPT-User handles pages a person asks ChatGPT to fetch in the moment, and OpenAI notes that robots.txt rules may not apply to it the way they do to automatic crawling, since the fetch is user-triggered rather than a background crawl. Getting these three right, on purpose, is the first and cheapest GEO decision most sites never make deliberately.

What is llms.txt, and does OpenAI require it?

llms.txt is a voluntary convention, proposed in 2024 by Answer.AI, that puts a short summary and a set of links to a site's most important pages in one plain-text file at the site root; it is not an OpenAI standard, not a W3C standard, and there is no public confirmation that ChatGPT's citation system reads it or requires it.

The idea is straightforward: a normal HTML page carries navigation, styling and scripts that a language model has to work around, while an llms.txt file (and its longer sibling, llms-full.txt) gives a clean, structured version instead. Adoption is growing among documentation sites and content-heavy sites, this one included: adsonomy.com publishes an llms.txt, an llms-full.txt, and a plain-text sidecar for every guide, in the format the convention describes. None of that is a guarantee of being cited; it is a low-cost, low-risk way to make a site easy to parse for any model that chooses to read it, OpenAI's or otherwise.

How is that different from a paid ad in ChatGPT?

A paid ad in ChatGPT runs on a separate system: it is built as a campaign with context hints and creative, reviewed before it can serve, billed per impression or click, and always shown labelled and apart from the answer, while an organic citation is not bought, not put through an ads review process, and not labelled as an ad.

The crawlers are separate too. OpenAI documents OAI-AdsBot as the crawler used to validate the safety of web pages submitted as ads: when an ad is submitted, OpenAI may visit its landing page to check it complies with OpenAI's policies. Its documented job is ad review, not surfacing pages for citations, which is what OAI-SearchBot does. So a page can be well optimised for citation and never touched by the ads review process, and a page built purely to serve an ad campaign never needs to be cite-worthy at all, though the same basics, being reachable and genuinely relevant, help both. The mechanics of building and running the ad side, hints instead of keywords, two creative formats, bidding and review, are covered in how ads in ChatGPT differ from search and social ads and Ads Manager versus the Ads API.

Does buying ads change whether a brand gets cited organically?

No: OpenAI states publicly that advertising does not influence the answers ChatGPT gives, and the two systems, ad review through OAI-AdsBot and citation crawling through OAI-SearchBot, are separate; running an ad is not a documented way to earn or improve an organic citation.

That separation cuts both ways. It means a competitor cannot simply outspend you into a better-cited answer, which is good news if your organic content is strong. It also means there is no shortcut: paying for a campaign does not substitute for the work of being a source worth citing. The two workstreams are genuinely independent, and budget spent on one does not buy progress on the other.

Why would a brand need both organic visibility and paid ads?

Because they solve different problems: organic citation has no guaranteed placement, no fixed timeline, and no bid you can raise to get it, while a paid ad gives a guaranteed, labelled placement on a schedule you control, so a brand that wants to be reliably visible in ChatGPT conversations today, not eventually, needs the paid channel alongside the organic work.

An organic-only bet is genuinely uncertain: there is no query report, no ranking control, and no way to know in advance whether a given conversation will surface your page. A paid-only bet misses the value of being a trusted source when someone is researching a category without clicking anything yet, and it never compounds the way a body of well-structured, genuinely useful content can. The two also reinforce each other in a smaller way: a page written clearly enough to be citation-worthy, with headings phrased as questions and answers stated plainly, tends to also be a page that is easy for OAI-AdsBot to review and easy for a reader who does click an ad to act on.

What should a GEO and AEO checklist look like?

Six things, in order of how cheap they are to get right: separate robots.txt decisions for each crawler, full crawlability of the pages you want cited, plainly stated answers, an optional llms.txt, current content, and a way to watch for the effect since there is no citation report.

1. Decide the robots.txt policy for each OpenAI crawler on purpose: OAI-SearchBot for citations, GPTBot for training, ChatGPT-User for user-triggered fetches. Allowing or blocking each is a separate choice.
2. Keep the pages you want cited fully crawlable: no login wall, no cookie-consent gate before the content loads, no answer that only appears after client-side JavaScript runs.
3. Write plain, self-contained statements that each answer one question, the way a key-facts box or an FAQ answer does, rather than paragraphs that only make sense in context.
4. Publish an llms.txt if it suits your site: a short summary plus links to the pages that matter, understanding it is a convention, not a confirmed ranking input.
5. Keep the content current. There is no confirmation that OpenAI treats stale pages differently, but a wrong answer traced back to your page is a bigger risk than simply being absent from one.
6. Track brand mentions and referral traffic from ChatGPT as your GEO signal, since neither OpenAI's search product nor its Ads API publishes a query-level or citation report.

How do I measure the organic side, since there's no citation report?

OpenAI has not published a query-level or citation report for ChatGPT search, so the organic signal has to be read indirectly: referral traffic and conversions arriving from chat surfaces in your analytics, movement in direct and brand-name search, and manual spot checks of how ChatGPT answers questions in your category, while the paid side reports impressions, clicks, spend and conversions directly through Ads Manager and the Ads API.

That asymmetry is itself a reason to run both. Paid gives you a number you can act on this week: cost per click, cost per conversion, spend against a plan. Organic gives you a slower, fuzzier signal that still matters for the same reason brand search and word of mouth matter, it just cannot be optimised with the same weekly discipline. For the paid side's own measurement setup, including the pixel, UTM tagging and reconciling platform numbers against analytics, see measuring ChatGPT ads with GA4, UTMs and incrementality tests.

Frequently asked questions