Skip to content

Insights

How do I get my business recommended by ChatGPT?

There is no submission form and no way to guarantee it. What you can do is make sure ChatGPT can reach your site, that your business is described consistently everywhere it appears, and that independent sources corroborate it — then measure whether any of it changed the answers.

Updated September 22, 2026 · 11 minute read

Short answer

There is no way to submit your business to ChatGPT and no way to guarantee a recommendation — there is no ranking lever and no paid placement in an organic answer. What you can influence is the information ChatGPT draws on: whether its crawlers can reach your site, whether your business is described the same way everywhere it appears, and whether enough independent sources corroborate what you say about yourself. Then you measure the answers before and after, because without that you are guessing.

The honest answer first

Most pages answering this question sell a service by implying control that nobody has. So, plainly: you cannot make ChatGPT recommend your business. There is no directory to join, no feed to submit, no field to fill in, and no advertising product that places a business inside an organic recommendation. Anyone promising a guaranteed AI recommendation is either confused about how these systems work or counting on you being confused.

What is true is more useful, and quite a lot less comfortable: ChatGPT's answer about local businesses is assembled from information that already exists about them. Some of that information you control. Most of it you do not, but you can influence it. The work is to make the true version of your business easy to find, easy to read and corroborated in enough places that it survives the trip into an answer — and then to check whether it did.

Two different mechanisms, two different kinds of work

ChatGPT can answer a question about local businesses in two quite different ways, and confusing them is the reason most advice on this topic is unreliable.

  1. It searches, then answers. For anything current or local, ChatGPT typically runs a search, reads what it finds and writes an answer from those pages, usually citing them. This is the path you can affect on a timescale of weeks. It depends on your pages being fetchable, indexed and clear.
  2. It answers from what it already absorbed. Asked something general, it may answer from the model itself — no live lookup, no citations. That impression was formed during training, from how the web described your business at the time. You cannot edit it. You can only add to the record and wait for a future model to be trained on a better one.

This is why timelines vary so much

Work aimed at the first mechanism can show up in a month. Work aimed at the second is a long game measured in model generations, and nobody can tell you the date. Any timeline offered for “getting into ChatGPT” that does not distinguish between these two is fiction.

Step one: check you have not accidentally opted out

This is the single most common own goal, and it is free to fix. Many sites block AI crawlers — sometimes deliberately during the 2023–24 wave of blanket blocking, often because a plugin, a host or a security product did it without anyone deciding. A site that blocks OpenAI's search crawler cannot be retrieved and quoted, no matter how good it is.

OpenAI runs separate crawlers for separate jobs, and this is the part people get wrong: blocking training does not have to mean blocking search.

OpenAI's crawlers and what each one controls
User agentWhat it doesBlocking it means
OAI-SearchBotIndexes pages so they can appear in ChatGPT's search resultsYou are not retrievable in ChatGPT search. Almost always a mistake.
GPTBotCollects content used to train foundation modelsYou are excluded from future training data. A legitimate business decision either way.
ChatGPT-UserFetches a page when a user's question requires itAffects live, user-initiated fetches of your pages.
OAI-AdsBotChecks advertising landing pages; not used for trainingOnly relevant if you advertise on the platform.

OpenAI's crawlers and what each one controls

Other assistants follow the same pattern of separating the jobs. Anthropic runs ClaudeBot for training, Claude-SearchBot for search indexing and Claude-User for user-initiated fetches. Perplexity runs PerplexityBot for indexing and Perplexity-User for fetches prompted by a specific question. Google separates Google-Extended, which governs AI training and grounding in some of its systems, from ordinary Search crawling.

So the configuration below is a legitimate position — present in search, absent from training — and it is a choice, not a default. Fetch your own robots.txt and find out which one you are currently making.

# Allowed to index so the business can appear in AI search answers
User-agent: OAI-SearchBot
Allow: /

User-agent: Claude-SearchBot
Allow: /

User-agent: PerplexityBot
Allow: /

# Declining model training. Reasonable businesses choose differently here.
User-agent: GPTBot
Disallow: /

User-agent: ClaudeBot
Disallow: /

User-agent: Google-Extended
Disallow: /
robots.txt permitting AI search indexing while declining training. Copy the intent, not blindly the file.

Declining training has a cost worth naming

Excluding your site from training data protects it from being absorbed, and also keeps your business out of the impression a future model forms of your market. If recall matters to you more than control does, the trade runs the other way. We do not think there is one right answer — but it should be a decision someone made on purpose, not something a plugin did.

What Google says about its own AI answers

Google is the only major platform that has documented this clearly, and what it says is a useful corrective to a lot of paid advice.

Google states there are no additional technical requirements for appearing in AI Overviews or AI Mode beyond ordinary Search eligibility: the page needs to be indexed, meet the usual technical requirements, and be eligible for snippets. It also notes that the snippet controls you already know — nosnippet, data-nosnippet and max-snippet — limit what its AI features can use, and that its AI features run multiple related searches rather than a single query.

Two practical consequences. First, if someone is selling you a special technical treatment to get into AI Overviews, the platform's own documentation contradicts them. Second, a max-snippet directive set years ago to protect content from being summarised is now also limiting what Google's AI can quote from you. That is worth checking.

Step two: be one business, described one way

Before anything can recommend your business, it has to be confident about what your business is. That sounds trivial. In practice it is where most local businesses are quietly losing.

A tree company might appear as “DEL Trees” on its site, “DEL Tree Service LLC” on its Google profile, “D.E.L. Landscaping — Tree Division” on a directory, with three phone numbers across the set and two different service areas. A person reconciles that in a second. A system deciding whether these are one business or three does not necessarily reconcile it at all — and a business it cannot resolve is a business it is unlikely to name.

What to reconcile, in the order that tends to matter:

  • Legal and trading name. One name, used identically everywhere, including the punctuation. Pick the one customers say out loud.
  • Address and service area. The same address everywhere, and an explicit statement of where you actually work. “Serving Montgomery County and surrounding areas” is a description; a named list of towns is an answer.
  • Phone. One primary number. Call-tracking numbers that differ per listing are a known cause of entity confusion.
  • Services. The same service vocabulary across your site, your profiles and your directory listings. If your site says “crown reduction” and everywhere else says “tree trimming”, you have split your own evidence.
  • Structured data. LocalBusiness and Service markup that matches the visible page. Schema that contradicts the page is worse than no schema.

Step three: get corroborated by somebody who is not you

Your website asserts. Everything else corroborates. A model's impression of your market is built from many independent sources describing the same businesses, and a business that exists only on its own domain has one source — the one with the most obvious motive.

This is usually the real difference when a competitor is named and you are not. Not a better website. A longer list of places that independently confirm they exist and do this work.

Where corroboration tends to come from, roughly in order of effort:

  • Your own profiles, completed properly. Google Business Profile, Apple Business Connect, Bing Places. Complete, categorised, consistent with the site.
  • Industry bodies and certifications. Trade association membership, licensing registers, manufacturer certifications. These carry weight because not everyone can claim them.
  • Local institutions. Chambers of commerce, business associations, municipal supplier lists, sponsorship pages. Geographically specific and hard to fake.
  • Review platforms, for their text. The rating is a summary; the reviews are content describing specific jobs in specific places. That text is quotable in a way a star count is not.
  • Genuine local coverage. Local press, community newsletters, neighbourhood forums. Slow, unreliable to plan for, disproportionately valuable when it happens.

Not link building with a new coat of paint

Buying placements on low-quality directories is the reflex this description invites, and it does not work here for the same reason it stopped working in search: those sources carry no authority and describe nothing. What matters is whether a source is independently trusted and actually says something specific about your business.

Step four: answer the question a customer actually asked

People talk to assistants differently from how they type into a search box. They give context, they describe a situation, and they ask for a judgement. “I have a maple that has lost a big limb over my driveway after the storm — is that an emergency, and what does removal usually run?”

Pages that answer that kind of question directly — stating what the work involves, what it typically costs, what affects the price, when it is urgent — are material an assistant can use. Pages of adjectives about commitment to excellence are not.

The test we use is simple: could a stranger read one page of your site and correctly answer a customer's question about your business? If not, neither can a machine.

  • Say what things cost, in ranges, with the variables named. Most local service sites refuse to. It is the single most common thing buyers ask an assistant, and the businesses that answer it are the ones with something to quote.
  • One page per service, per place you actually serve. Not doorway pages for towns you have never worked in. Real pages for real coverage, with detail that proves it.
  • Publish evidence, not claims. Dated jobs, locations, before-and-after specifics, named certifications. “Trusted by homeowners since 1998” asserts; “214 removals in Montgomery County in 2025” can be cited.

Step five: measure, or you are just decorating

Everything above is reasonable. None of it is proof. The only way to know whether any of it changed ChatGPT's behaviour towards your business is to have recorded what ChatGPT said before you started.

That means a fixed set of questions phrased the way your buyers speak, run repeatedly, across platforms, with every answer stored. Then the same questions again after the work, compared like for like. Change the questions between rounds and you have thrown away the comparison — which is, not coincidentally, how a lot of reporting in this field manages to look successful.

Expect the first measurement to be uncomfortable and expect it to be surprising in a specific way: the most common finding is not absence, it is inaccuracy. Assistants confidently describing service areas you left, services you dropped, or hours that changed two years ago.

Common questions

Can I submit my business to ChatGPT?

No. There is no submission process, no directory to join and no paid placement inside an organic recommendation. The only route is the information about your business that already exists on the web.

How long does it take to appear in ChatGPT's answers?

It depends which mechanism you mean. Changes affecting what ChatGPT retrieves when it searches can register within weeks. Changing what the model recalls without searching depends on future training runs, and nobody outside OpenAI can give you a date.

Should I block GPTBot?

That is a genuine business decision and reasonable owners go both ways. What matters is knowing that training and search indexing are controlled by different user agents, so you can decline training without making yourself unretrievable in ChatGPT search. Check what your robots.txt currently says — many sites are blocking both without having chosen to.

Does schema markup make ChatGPT recommend me?

No single technical change does that. Structured data makes your claims machine-readable and consistent with your visible page, which helps a system resolve who you are. Schema that contradicts the page is worse than none.

Is this different from ranking in Google?

Related but not the same. Indexing and snippet eligibility are preconditions for AI answers that search the web. But an answer names two or three businesses rather than ten, and some answers involve no search at all.

Sources

Primary documentation from the platforms themselves, cited where this page makes a claim about how they behave. Where something is not documented publicly, this page says so rather than asserting it.

More resources

  • What is AI visibility?

    AI visibility is whether AI assistants name your business when someone asks them who to hire. It is a different question from search ranking, it is measured differently, and it can be tested.

  • Why does AI recommend my competitors instead of me?

    Usually not because they are better. Because more independent sources describe them, describe them consistently, and describe them in the specific terms a customer's question used.

Find out what AI currently says about your business

A baseline records which businesses are named across your buyer questions, on which platforms, and which sources the assistants relied on. Nothing on this page is a substitute for measuring your own.