Can AI find your business if you're only on Facebook?

I put a short version of this on Facebook, which felt like the right place for it. If you ask ChatGPT to "find me a cheap local painter with great reviews", it's never going to see a Facebook-only business. I thought I'd test it properly and see how true it is rather than just something web designers say (myself included!).

The short answer is that AI assistants mostly can't use your Facebook Page, for three separate reasons stacked on top of each other.

Also this isn't an argument against Facebook. It's still one of the best ways to reach real people locally, and a recommendation in a local group still leads to jobs. The problem isn't being on Facebook. It's being only on Facebook.

How AI assistants actually find a business

Two things happen when you ask ChatGPT, Claude or Gemini a question. The first is checking what it already knows from training. Essentially a snapshot of the web from some point in the past, and generally they never include local small businesses.

The second is live search and it's that one that matters. When you ask for "a plumber in Bolton", the assistant runs a search, picks a handful of promising pages, fetches them, reads them, and writes its answer out of what it found.

Both routes are controlled by a file called robots.txt. In short it's a file on every website listing which 'bot' visitors are allowed to view it.

Being missed by a bot is no longer a small concern. ChatGPT has over 900 million weekly users from their report in February 2026 and it's only increasing.

Facebook asks AI crawlers not to look

Facebook publishes this robots.txt at facebook.com/robots.txt.

On there is a list of crawlers that are allowed in, each with its own long list of strictly off-limits paths: Googlebot, Bingbot, DuckDuckBot, LinkedInBot, Applebot, GPTBot, ClaudeBot, PerplexityBot. That list looks generous.

The interesting part is which crawlers are missing, because AI companies don't use a single bot for everything. OpenAI runs GPTBot to collect training data and OAI-SearchBot to build the results that appear in ChatGPT's search answers. Facebook permits the first and prohibits the second. That means OAI-SearchBot is blocked. OpenAI's say "Sites that are opted out of OAI-SearchBot will not be shown in ChatGPT search answers." In short Facebook results can't show as an answer on ChatGPT.

The same pattern repeats. ChatGPT-User, the agent that fetches a page when you ask ChatGPT to go and look at something, isn't listed. Neither is Claude-User, nor Claude-SearchBot, nor Perplexity-User.

So what does a crawler actually receives from a Page?

Policy is one thing. Sometimes bots go outside of 'policy'. I wanted to see what a bot physically gets back, so I fetched Costa Coffee's Facebook Page.

Requested in an ordinary bot browser, it returned an error page. Requested with a crawler it returned more errors. The title was just "Facebook". The phrase "Costa Coffee" appeared zero times. No tags, no opening hours, no reviews, no structured data of any kind.

A Page has nothing worth quoting

Suppose a bot does get through the door? There's still very little on a Page it can use.

There's no structured data, no block saying "this is a business, it's called this, it's in Bury, here's the phone number, here are the opening hours." Posts are varied, so your prices and your patch sit buried in something you wrote four months ago rather than on a stable page. And the reviews, the most persuasive thing you own, are behind a login wall a bot can't get around. Facebook's robots.txt disallows /login* and its business search paths even for the crawlers it permits.

So even a bot allowed in finds nothing.

I tested it

On 28 August 2026 I asked three assistants the same question, signed in to my own accounts: "Find me a cheap local painter and decorator in Bury with great reviews."

ChatGPT searched the web in front of me: "Searched 2 websites", Checkatrade and MyBuilder. It drew a map, named four decorators with star ratings and review counts, and cited four sources: two profiles on the review aggregator Ratings Plus, and two on MyBuilder. No Facebook.

Claude didn't search the web at all. It ran a places lookup and answered straight out of Google listing data, with a map and four names. Zero sources cited. No Facebook.

Gemini knew from my account that I was in Bury. No search step, no sources panel, just Google Business Profile cards: photos, star ratings, street names, live opening hours. No Facebook.

Six cited sources across the whole test, none from facebook.com. Three businesses were named by all three assistants.

Then I pointed two of them straight at a Facebook Page, URL in the prompt, and asked for its opening hours and what the reviews said. Nothing to find, just read.

ChatGPT searched six other websites instead and answered from Costa's own site, telling me: "Facebook also throttled direct access to the page from my side, so I can't reliably read its current Facebook Reviews/Recommendations feed."

Claude tried to fetch it and failed on screen, a red block reading Failed to fetch https://www.facebook.com/Costa/. Its answer said "Facebook blocks automated access, so I can't read the page directly".

I'd also run the painter question twice through a raw web search index, the layer these assistants pull from, logging all 18 results. Every one was a trade directory (Checkatrade, MyBuilder, MyJobQuote, TrustATrader, Bark, Thomson Local, Rated People, TheBestOf) or one decorator's own website. No Facebook Pages, and no Google reviews, Trustpilot or Yell either.

The clearest result was an accident.

One Bury decorator I checked has both a Facebook Page and a website.

Across those two searches their website surfaced five times and their Facebook Page surfaced zero times.

Fetch both without logging in and you can see why. The website hands over a phone number, an email address, a postal address, a service area covering seven towns, more than thirty years in the trade, a City and Guilds NVQ, insurance, and testimonials you could quote directly. The Facebook Page gave up the name of the Page with no opening hours, contact details or customer reviews.

Same business, same trade, same town. One version is findable by an AI chatbot, and one isn't.

What to actually do about it

Keep the Facebook Page. It's where your customers already are, and it's how word travels online between people. Just don't treat it as your business home. A Page is a channel you rent on someone else's platform, on terms you don't set, and that platform has decided that the chatbots that fuel AI answers aren't allowed in.

From the tests there are three really easy recommendations to help you get ahead of every other Facebook Business out there.

1. Claim and fill in your Google Business Profile. It's free, and two of the three assistants answered me entirely out of that data. No search, no links, just names, ratings, addresses and opening hours from Google's listings. If yours is unclaimed or showing last year's hours, you're invisible down that whole route. It's the cheapest thing you can do.

2. Get onto a trade directory or two. ChatGPT's answer came out of Checkatrade and MyBuilder, and every source it cited was a directory profile. One or two you'll keep up to date. No need to go full steam into 20 different directories, one or two up to date profiles beats that as you'll never keep on top of them all.

3. Put a website at the centre of it. Not a single Facebook page was returned by an AI agent when asked for a local decorator. They all returned business profiles and told me if they had a website. A website, for a lot of people, proves legitimacy and has instant information about every aspect of your business. If you want to be the first choice then you need to have a website. You can bet the decorator with the website looks the most professional of all of the options.

The UK Business Data Survey 2026 found 76% of sole traders have a website, that means roughly one in four people working for themselves still don't have one. Ten years ago that meant being invisible to Google. Now it means being invisible to both Google and the new wave of search, AI chatbots.

All you need is one page, on a domain you own, with the information in plain text: business name, what you do, the areas you cover, phone, email, opening hours, and three or four real testimonials that aren't behind a login. That will set you apart from the competition.

Starter websites begin at £400, and business websites start at £600 if you want room to grow. Full details are on the pricing page.

If you only do one thing this week, claim the Google Business Profile.

If you do two, buy the domain.

If you want a second opinion on your own setup, get in touch and I'll tell you what I think.