Why ChatGPT doesn't recommend your business

Someone in your city needs an electrician tonight. Ten years ago they opened Yellow Pages. Five years ago they searched Google and picked from the map pack. Increasingly, they open ChatGPT and type "who's a reliable electrician in Bristol?" — and they get a short list of three or four names with a sentence about each.

You are either on that list or you do not exist for that customer. There is no page two.

This is a different game from SEO, and most of the advice you'll read conflates the two. Ranking is about being findable. This is about being quotable — an answer engine has to be able to say something specific and confident about you without hedging. Here is what actually decides that.

You might be blocking the crawlers

Start here, because it is the one that takes ten minutes and silently undoes everything else.

Answer engines send their own crawlers, and they are not Googlebot. If your robots.txt blocks them — or your host, firewall or bot-protection blocks them before they reach it — then nothing else on this list matters. You have opted out of being cited without ever deciding to.

The agents worth knowing about:

• GPTBot and OAI-SearchBot — OpenAI's crawler and its search grounding
• PerplexityBot — Perplexity's index
• Google-Extended — controls Gemini and AI Overviews separately from search
• ClaudeBot — Anthropic
• Bingbot — Copilot's answers ride on Bing's index

That Google-Extended line catches people out. It is not a search setting. You can rank perfectly well in Google and still be excluded from AI Overviews because a plugin or a well-meaning consultant blocked it to "stop AI stealing our content."

Check yourdomain.com/robots.txt right now. If you see any of those names next to Disallow: /, that is your afternoon's work.

The machine can't tell what you do

Most local business sites describe themselves in a way that is perfectly clear to a human and nearly useless to a model.

"Quality workmanship since 1998." "Your trusted local partner." "We go the extra mile." A person reads that and fills in the gaps from the photos, the van, the tone. A model reads it and extracts approximately nothing, because nothing in it is specific — no trade, no city, no service, no proof.

Compare it to what an engine actually needs in order to name you:

• What you do, in the words a customer would use — "emergency electrician," not "electrical solutions"
• Where you do it, by name — the towns and postcodes, written out
• What you don't do — the jobs you turn down are as clarifying as the ones you take
• Why someone picks you over the other four, stated plainly enough that it could be quoted back

A model generating an answer is looking for a sentence it can lift. Give it one. If your site does not contain a sentence you would be happy to see quoted verbatim in a recommendation, write it today.

Nobody else says it about you

This is the uncomfortable one, and it's why this work takes months rather than an afternoon.

Models weight what other people say about you far more heavily than what you say about yourself. Your own site establishes that you exist and what you claim. Directories, local press, trade bodies, forum threads, supplier pages, review platforms and community sites establish whether it's true.

If you are absent everywhere except your own domain, an engine has one source with an obvious motive, and it will hedge — "you may want to check local directories" — rather than name you. That hedge is the sound of you losing the job.

You cannot shortcut this, and you should not try. Fake reviews and paid mention farms are detectable, they poison the well, and the platforms that matter are getting better at spotting them every quarter. The legitimate version is slower and it works: be genuinely listed where your trade is genuinely listed, and be consistent everywhere you appear.

Your details disagree with each other

Name, address, phone. If your unit number is on the Google listing but not on Yell, if you've got an old mobile on one directory and a landline on another, if the business is "Smith & Sons" here and "Smith and Sons Ltd" there — every inconsistency is a reason for a model to lower its confidence and recommend someone whose details agree with themselves.

It is boring, unglamorous work. It is also the highest ratio of result to effort on this entire page.

You're invisible to the parser

Schema markup — the structured data in your page's source — is how you tell a machine this is a business, this is the service, this is the area, these are the opening hours without it having to infer any of it from your layout.

For a local service business the ones that earn their keep are LocalBusiness, Service, and FAQPage. That last one matters more than people expect: answer engines are, structurally, in the business of matching questions to answers. A page that is literally formatted as questions and answers is the path of least resistance.

Write the ten questions you actually get asked on the phone. Answer them in plain language. Mark them up. That single page will do more for your AI visibility than a year of blog posts about industry trends.

What to do this week

In order, because the order matters:

1. Read your robots.txt. Unblock the agents above if they're blocked.
2. Write one sentence that says what you do, where, and why you. Put it near the top of your homepage.
3. Make your name, address and phone identical everywhere they appear.
4. Build a real FAQ page from real questions and mark it up.
5. Then — and only then — start thinking about content and mentions.

Then go and ask ChatGPT, Gemini and Perplexity who the best in your trade is in your town. Ask three times each, in a fresh session, because these systems are non-deterministic and one run tells you nothing. Write down who gets named.

That list is your actual competition now. Not who ranks above you — who gets said out loud when a customer asks.

---

This is the diagnostic half of what we do. If you'd rather not run it yourself, get in touch — we do it across five engines and hand you the fix list.