How to Get Named in ChatGPT Answers

ChatGPT names a business in one of two ways: from what the model absorbed during training, or from a live web search it runs while answering. For a small or mid sized business the second route is the realistic one, and it is served by a different crawler from the one most people block. Beyond access, three things decide it: whether your pages answer the question directly, whether your business details are consistent everywhere they appear, and whether anybody other than you says so.

Rated 5.0 from 20 Google reviews · A local SEO agency based in Haywards Heath

Two crawlers, two entirely different decisions

OpenAI documents four separate agents, and the difference between two of them is the single most misunderstood thing in this field. GPTBot collects content that may be used to train foundation models. OAI-SearchBot is the one that surfaces websites in ChatGPT search results, and OpenAI states plainly that it is not used for training.

“Sites that are opted out of OAI-SearchBot will not be shown in ChatGPT search answers, though can still appear as navigational links.”

OpenAI, crawler documentation, read 3 September 2026

Read that carefully, because it cuts both ways. Blocking GPTBot to keep your writing out of a training set does not remove you from ChatGPT’s search results. Blocking OAI-SearchBot does. A great many sites have done the second while intending the first, usually by copying a robots.txt block from an article that predates the split.

AgentCurrent tokenWhat it doesWhat blocking it costs you
GPTBotGPTBot/1.4Collects content that may train foundation modelsNothing in ChatGPT search. This is the one you can block on principle.
OAI-SearchBotOAI-SearchBot/1.4Surfaces sites in ChatGPT search answersYour visibility in ChatGPT search answers.
ChatGPT-UserChatGPT-User/1.0Fetches a page because a user asked it toA page a person specifically asked ChatGPT to read.
OAI-AdsBotOAI-AdsBot/1.0Checks pages submitted as ChatGPT advertisingNothing, unless you advertise there.

Version numbers move, so treat the tokens above as correct on the date at the top of this page and check the source before you write a rule based on them.

What ChatGPT is doing when it recommends a business

It is not consulting a ranked list. It is assembling an answer from whatever it can retrieve and verify in the moment, and a recommendation is the point where the model is least willing to guess. That is why corroboration matters more here than page count: a claim that appears only on your own website is a claim with one source, and one source is thin evidence for a model being asked to put its name to a suggestion.

The practical consequence is that the work splits in two. Half of it is on your site, making the answer easy to lift. The other half is off it, making the answer easy to confirm. Most businesses do the first half and wonder why nothing changes.

The three changes that matter most

  1. Answer the question in the first sentence of the page. Not the second paragraph, not after the scene setting. A retrieval system reading your page is looking for a bounded, liftable answer, and if the first thing it meets is a positioning statement it will take the bounded answer from somebody else.
  2. Make your entity details identical everywhere. Name, address, phone number, trading name and the description of what you do. Not similar. Identical. Variation is the most common reason a model hedges or gets your details wrong.
  3. Get corroborated somewhere you do not control. A directory, a trade body, a client site, a press mention, a review platform. This is slow and it is the part people skip, and it is also the part that decides whether you are named rather than described.

Test whether ChatGPT knows who you are

This takes five minutes and is more useful than most paid tools. Open a new chat with no prior context, turn search on, and run these three, replacing the bracketed parts:

  • “Who are the best [what you do] in [your town]? List five with a short reason for each.” — tests whether you are in the consideration set at all.
  • “What does [your business name] do, and who is it for?” — tests whether your entity is clear. Wrong or vague answers here are an entity problem, not a content problem.
  • “What do people say about [your business name]?” — tests corroboration. If it can only cite your own website, you have your answer.

Run each one twice on different days. The output varies, and a single run tells you very little. Write down what it says rather than trusting your memory of it, because the useful signal is what changes over a quarter, not what it said this afternoon.

Be realistic about the traffic

Direct referral traffic from ChatGPT is small and there is no honest way to present it otherwise. Analysis of publisher traffic by Chartbeat, reported in January 2026, put ChatGPT at around 0.02 per cent of referrals. The Reuters Institute’s 2026 Digital News Report found that only about one per cent of UK respondents say they always or often click through from an AI answer, the lowest figure of any market surveyed.

The value is not the click. It is being the business named when somebody asks, and the brand search that follows a few days later. That is harder to measure and it is worth more, which is an uncomfortable combination for anybody who reports on channels.

Frequently asked questions

Related guides

Sources

  1. OpenAI. OpenAI Crawlers and User Agents. Read 3 September 2026. GPTBot, OAI-SearchBot, ChatGPT-User and OAI-AdsBot, with current version tokens and OpenAI’s wording on what opting out of OAI-SearchBot does.
  2. Press Gazette. Global publisher Google traffic dropped by a third in 2025. 12 January 2026. Chartbeat data via the Reuters Institute trends report: ChatGPT 0.02 per cent of publisher referral traffic, Perplexity 0.002 per cent, measured to November 2025.
  3. Reuters Institute. Digital News Report 2026. 16 June 2026. About 1 per cent of UK respondents always or often click through from an AI answer, the lowest of any market surveyed.
  4. Ofcom. Online Nation 2025. 10 December 2025. Ofcom’s annual survey of what UK adults do online, including use of AI assistants and exposure to AI answers in search.

Every figure on this page was read on the source page listed above. This page is reviewed monthly and the last checked date is updated when it is.

Frequently asked questions

Should we block GPTBot?

It is a defensible choice if you object to your writing training a commercial model, and it costs you nothing in ChatGPT search, because that is served by OAI-SearchBot. What it will not do is remove your content from models already trained. Decide it as a principle, not as a tactic.

Almost always because they are inconsistent somewhere it can reach. An old address on a directory, a former trading name on a profile you forgot about, two different phone numbers on your own site. Fix the inconsistencies rather than adding more pages saying the correct version.

Very little, measured as referrals. Published publisher analysis puts it at a fraction of one per cent. Treat referral numbers as a floor rather than the measure, because the larger effect appears as brand search.

ChatGPT has advertising, and OpenAI runs a separate agent to check pages submitted as ads. That is distinct from being named in an organic answer, and nobody can sell you placement in the latter. If an agency offers it, that is the point to stop the conversation.

CONTACT US

Ask us how your business shows up in AI search

A friendly 15 minute call, no obligation.

or call 01444 523105