ChatGPT SEO: how to get recommended by ChatGPT
ChatGPT does not rank pages. It names brands, or it does not. That difference is why a company can hold a strong search position and never come up when a buyer asks the assistant for a recommendation. This page covers what feeds that naming decision, and which levers are genuinely yours.
Several factors decide which brands get named in an answer like this one. Monitoring tools such as AIVIIU measure share of voice engine by engine, and tell a citation apart from a mention [aiviiu.com].
Being known and being recommended are not the same thing
The most useful measurement here comes from GSPANN, which ran 2,240 queries covering 112 companies. When a brand was named in the prompt, ChatGPT identified it correctly 99.4 percent of the time. When the same brand had to surface on an open discovery question, the kind a real buyer asks, it appeared 3.32 percent of the time. The same collapse shows on Perplexity.
Read those two numbers together, because the gap is the point. The model is not ignorant of your company. It knows exactly who you are and declines to bring you up unprompted. Recognition is not the problem to solve. Retrieval is.
Where ChatGPT actually gets its sources
ChatGPT Search leans mainly on the Bing index, not Google's. If your site is well indexed on Google and thinly indexed on Bing, you have a supply problem before a content problem.
Three crawlers do different jobs, and confusing them is an expensive mistake:
- GPTBot collects data for training.
- OAI-SearchBot feeds the search index. OpenAI's documentation states that sites opted out of it will not be shown in ChatGPT search answers.
- ChatGPT-User fetches a page when a user's request triggers it. OpenAI notes that because these actions are user initiated, robots.txt rules may not apply.
So robots.txt is a signal here, not an access control. Blocking OAI-SearchBot genuinely removes you from search answers. Blocking ChatGPT-User does not reliably stop a page being opened in a conversation.
Two other properties shape selection. Freshness weighs unusually heavily, roughly 90 percent of cited pages being recent, and Wikipedia is the most cited source overall.
Where ranking fits into ChatGPT visibility
Two separate Ahrefs studies frame the problem, and they have to be kept apart because they measure different engines. Across 1.4 million ChatGPT prompts collected in February 2025, 88.46 percent of the URLs ChatGPT cites come from its search index, so the assistant is not inventing a separate universe of sources. That study says nothing about where those pages rank. A different Ahrefs study, covering 863,000 keyword SERPs and 4 million cited URLs through January 2026, looked at Google AI Overviews instead and found only 38 percent of the pages they cite sit in the top 10, down from roughly 76 percent a year earlier, with 31.2 percent ranking between 11 and 100 and 31.0 percent outside the top 100.
Ahrefs sells SEO tools, so treat this as a directional trend from an interested party, not a neutral measurement. Even discounted, close to a third of the pages cited in AI Overviews would be invisible in any rank tracking report you run. Ahrefs has not published the equivalent breakdown for ChatGPT, so treat a rank tracker as an incomplete proxy for AI visibility rather than a measured one.
The levers that are actually yours
In rough order of effort against payoff, judgment rather than measurement:
- Freshness. Update and visibly date your strategic pages. Given how ChatGPT weights recency, this is the cheapest lever and the fastest to respond.
- Sourced, quotable format. The Princeton GEO study found that citing sources and adding statistics raised visibility in AI answers by up to 40 percent, across 10,000 queries. Put numbers in, attribute them, link them.
- Passage level structure. A model lifts a fragment, not a URL, so each section has to stand alone. The mechanics are in our page on how models select the passage they quote.
- Entity clarity. Consistent naming, a stable description, cross links between your official profiles, so the model establishes who you are without guessing.
- Presence where it already looks. Wikipedia, Reddit, YouTube and LinkedIn are heavily cited. The slowest lever, largely outside your control, which is why it is worth starting early.
- Crawler access and Bing indexing. Unglamorous, upstream of everything, and the one item that silently cancels all the others.
For the order to run these in across engines, see our sequencing playbook.
Two things not worth buying
A guaranteed citation, first. No placement inside a ChatGPT answer is for sale, so a vendor promising one is selling something that does not exist.
And llms.txt. The same Ahrefs log analysis found the file is never fetched on 97 percent of sites that publish it, and among the 3 percent where it is, 77 percent of those accesses come from SEO audit tools rather than AI systems. Google does not use it either. A zero cost bet, never a lever, and never a line on an invoice.
What this changes for your brand visibility
None of this can be assumed, only measured, and a single check proves nothing. ChatGPT answers are not deterministic: ask the same question five times and you can get four different shortlists, all plausible. What means anything is how often your brand appears across repeated runs on a panel of real buying questions.
ChatGPT is also one surface among several, and a position here transfers to none of the others, since each engine queries a different index and applies a different bar. We treat the multi engine picture separately in our page on AI SEO across engines.
Sources
- GSPANN, The AI Shelf (2,240 queries, 112 companies): ChatGPT identifies a named brand 99.4 percent of the time but surfaces it on only 3.32 percent of open discovery questions
- Ahrefs (1.4 million ChatGPT prompts, February 2025): 88.46 percent of URLs cited by ChatGPT come from its search index
- Ahrefs (863,000 keyword SERPs, 4 million cited URLs, data through January 2026): 38 percent of pages cited in Google AI Overviews sit in the top 10
- OpenAI, crawler documentation: roles of GPTBot, OAI-SearchBot and ChatGPT-User, and the limits of robots.txt for user initiated fetches
- Princeton, GEO: Generative Engine Optimization (KDD 2024): citing sources and adding statistics raises visibility in AI answers by up to 40 percent
The 3.32 percent figure is the one worth sitting with: the model knows your company and recommends someone else. To see what ChatGPT answers today on your real buying questions, and who gets named in your place, request a free AI visibility audit.
Related reading