How do you get your business recommended by ChatGPT?
By being findable, readable and unambiguous to a machine. Three small files and one habit decide most of it, none of them need a developer, and you can check every one of them free in about ten minutes.
The rest of this page is that list, in order, with the free check for each step. No sign-up and no email field anywhere on it.
Anyone whose customers have started asking AI instead of searching.
When somebody asks an assistant who to hire in your trade, the answer names two or three businesses. There is no list underneath it, no second page, and nothing to appeal. You are either in that answer or you are invisible for that question.
The good news is that the first half of this is housekeeping, it is free, and most of your competitors have not done it either.
Six steps. The first two are the ones audits skip.
Let the right robots in
Each assistant reads the web through its own crawler, and they are not interchangeable. GPTBot is training only — block it or allow it, your visibility in ChatGPT search does not change. OAI-SearchBot is the one that decides. OpenAI's own wording: sites opted out of OAI-SearchBot will not be shown in ChatGPT search answers. Check yours with the robots.txt checker.
Check nothing is refusing them anyway
robots.txt is a request. A firewall or bot-management rule is enforcement, and it runs before your website does, so a crawler refused there never reads your robots.txt at all. This arrives switched on with some CDN and security plans and nobody emails you about it. The crawler block checker is the only way to see it.
Write an llms.txt
A plain text file at yoursite.com/llms.txt saying what your company does, who it serves and where the important pages are. Five minutes. An engine that cannot work out what you do writes its own guess, and its guess is worse than your sentence. Draft one from your sitemap here.
Publish your prices as a file
yoursite.com/pricing.md, plain markdown. This is the one almost nobody does. Assistants increasingly compare options on a buyer's behalf, and one that cannot find a price cannot put you in the comparison. A price behind a contact form is a price the machine treats as absent. If the real answer is “it depends”, publish the ranges and what moves them.
Label your pages for machines
Structured data tells an engine what it is reading. Four types carry the weight: Organization on the site, FAQPage on your FAQ sections, Article on posts, HowTo on step-by-step guides. Most site builders have a plugin that does it in an afternoon. Validate what you have.
Make the page agree with itself
Your <title>, your <h1> and your og:title are three promises about one page, and nothing warns you when they drift apart. A template copied to a new page keeps the old heading; every tag stays valid and every validator passes. See all three side by side.
It is not ranking. It is assembly.
A search engine returns a list and lets you choose. An assistant writes an answer and names a handful of businesses inside it. To do that it gathers what it can find about a category, decides what it trusts, and composes.
So the question is not “where do I rank”. It is “when a machine assembles an answer about my trade, is there enough unambiguous material about me to include?” If what it finds is thin, or says two different things, it names somebody else. It does not tell you it did that, and there is nobody to ask.
That is why the order on this page matters. A crawler that cannot reach you contributes nothing, however good your content is. A page that describes itself three different ways contributes confusion. The steps above are, in order, can it reach you, can it read you, and does what it reads agree with itself.
What this does not do.
None of this makes you the best answer
It makes you a findable, readable, unambiguous one. Which business an assistant actually names depends on what third parties say about you, and that is earned over months rather than configured in an afternoon.
There is no switch and no appeal
No ranking to climb, no second page to be on, nobody to email. An engine assembles an answer from what it can find, and if what it finds is thin or contradicts itself it names somebody else and never tells you.
Expect weeks, not days
Engines re-read on their own schedule. A fix made today shows up when they next look, which is why anyone promising movement inside a fortnight is describing something they do not control.
Some of it genuinely will not matter to you
If your buyers do not ask an assistant for a recommendation in your trade, this is housekeeping rather than growth. Ask one three questions your customers ask and you will know within ten minutes.
Five moments worth ten minutes.
You are invisible and nothing explains it
Pages indexed, robots.txt clean, schema valid, and still never named. That combination points at the two checks most audits never run: the crawler names, and whether anything is refusing them at the edge.
After a CDN, host or security change
The commonest cause of a sudden disappearance. The day the DNS moved can be the day the answers stopped, and nothing announces it.
Before paying anyone for AI visibility work
Content and authority cannot help a crawler that never gets a response. Check the door opens before paying someone to decorate the room.
When an assistant describes you wrongly
Being described as the wrong business is worse than being absent, and the usual cause is a shared address or a near-identical name rather than anything on your site.
Once a quarter after that
Put a visible “Last updated” date on your key pages and actually update them. Engines weight recency hard: a decent page from last month beats a great page from two years ago for most commercial queries.
Every step above has a tool on this site.
We built these because we needed them. They take a url, run a real check against the live page, and show you what came back. No sign-up, no email field, and we do not store the pages people check.
| Check | What it answers |
|---|---|
| Robots.txt checker | Which of the 18 AI crawlers your site allows, and what each one costs you. |
| AI crawler block checker | Whether something at your edge is refusing those crawlers regardless of robots.txt. |
| llms.txt generator | Drafts one from your own sitemap, and says plainly what is unproven about the file. |
| Structured data validator | Whether your markup parses and whether its references actually resolve. |
| Title tag checker | Your title, h1 and og:title side by side, so a disagreement is visible. |
| Canonical tag checker | Whether your canonical points where you think, following it to see. |
| Sitemap validator | Finds the sitemap your robots.txt declares and reports what is unusable in it. |
Findable is the floor. Being named is the job.
Everything above makes you legible to a machine. Which business it names depends on what other sites say about you, directories, reviews, mentions, the pages an engine already trusts in your category. That is earned over months and it is the part people hire out.
We publish how we test it, including the questions where we are not named ourselves, on our methodology page. Prices are on the pricing page.
What owners ask about this.
How do I get my business recommended by ChatGPT?
Make sure the crawler that feeds ChatGPT search can reach you, then give it something unambiguous to read. In order: confirm OAI-SearchBot is not disallowed in your robots.txt, confirm nothing at your firewall is refusing it regardless of what robots.txt says, publish an llms.txt describing your business, publish your prices as a plain file, add Organization and FAQPage structured data, and make your title and heading describe the same page. That is the findable-and-readable half, and it is the half you control. Which business gets named also depends on third-party evidence about you, which takes months.
Is this the same as SEO?
It overlaps and it is not the same job. SEO competes for a position in a list of links. This competes to be named inside an answer where there is no list, no second page and no appeal. Good SEO helps because the same crawlers are involved, but a page can rank well and still never be quoted, and that is the gap this work closes.
Does blocking GPTBot hide me from ChatGPT?
No, and this is the most common mix-up there is. GPTBot is training only. OAI-SearchBot is what decides whether you appear in ChatGPT search answers, and ChatGPT-User is what fetches a page when somebody asks ChatGPT to go and look at it. Anthropic splits theirs the same way with ClaudeBot, Claude-SearchBot and Claude-User. Blocking the training crawlers while allowing the search ones is a perfectly reasonable position. Blocking the search ones by accident is just lost business.
Will Google-Extended hurt my Google ranking?
No. Google-Extended controls Gemini and AI training only and does not affect Google Search ranking at all, which is the opposite of what most people assume when they block it. Blocking Googlebot is the one that costs you Search, and it also removes you from Brave Search, whose crawler follows Googlebot's rules and has none of its own.
Do assistants really read llms.txt?
Honestly: no major assistant has confirmed that it does. It costs five minutes, it cannot hurt, and it gives any engine that does look a clean summary instead of a guess. Anyone selling you llms.txt as the answer is overselling a file. We say the same thing on the page where we generate one.
Why would publishing prices help?
Because an assistant comparing options on a buyer's behalf can only include what it can read. A price behind a contact form is, to a machine, a business with no price, and a comparison that cannot include you names somebody it can. If your pricing genuinely varies, publishing the range and what moves it is still more useful than nothing.
How long before any of this shows up?
Weeks for the technical half, because engines re-read on their own schedule rather than yours. Months for the part that depends on other sites mentioning you. Our own measured experience: fixing an entity problem removed a wrong answer within three days and did not by itself produce a right one.
What if an assistant describes my business as something else entirely?
That is usually an entity problem, not a content problem, and the usual cause is an address shared with a similarly named company. A machine resolving who you are leans heavily on address, and two businesses at one address with similar names are one business to it until something separates them. It is fixable, but the fix is every public record agreeing at once rather than a single edit.
Related: Which AI crawlers can see my site? · AI crawler block checker · Robots.txt checker · Our methodology
By Steven Allbright, founder, Ultimate Elite Solutions · Last updated: 2026-10-09