How to get your business recommended by ChatGPT: a UK checklist

Watch how people find a service business now. A homeowner with a leaking tank no longer types "emergency plumber Bedford" and scans ten blue links; a growing share of them ask ChatGPT "who should I use for a burst pipe near me?" and get back three named firms with reasons attached. The same is happening with accountants, physios, dentists, solicitors and landscapers. If your business is in that short answer, you get the enquiry. If it isn't, you were never in the running, and nobody sends you a rejection letter.

I won't dress this up with adoption percentages, because the figures being passed around are mostly guesses and the platforms don't publish the real numbers. What I can tell you from testing this for clients every month is that the shift is real, it skews towards high-intent questions (the "who should I actually book?" kind), and the answer an assistant gives is assembled entirely from what machines can read about you. That last part is the useful part. It means this is fixable, and mostly with unglamorous work. People have started calling that work answer engine optimisation, or AEO; the label matters less than the habit behind it.

How assistants actually decide

Strip away the mystique and there are four honest inputs, plus one complication.

• What search engines surface. When ChatGPT has search switched on, it rewrites your question into search queries, sends them to third-party search providers (OpenAI has named Bing among its partners), reads the results and writes an answer with citations. Gemini and Google's AI Overviews draw on Google's index and its local data. Perplexity runs its own retrieval, but the principle holds: assistants can only recommend businesses that appear on pages they can fetch. Invisible in ordinary search means invisible to assistants.
• What reviews say. Reviews are third-party evidence, so assistants lean on them hard. Gemini reads your Google reviews directly, and ChatGPT and Perplexity routinely cite Trustpilot, Checkatrade and similar platforms when someone asks who's good for a job.
• What structured data declares. Schema markup (LocalBusiness, Service, FAQPage) states in machine-readable terms what you do, where you do it and for whom. It isn't magic; it removes guesswork.
• What your own site states clearly. When an assistant fetches your page, it reads the words. "Same-day boiler repair across Bedford and Milton Keynes, Gas Safe registered" can be quoted in an answer. "Passionate about excellence" cannot.

The complication is training-data lag. When a model answers from memory instead of searching, it draws on information that may be months or years old, and nobody, whatever they claim, can edit a model's memory on demand. What you can do is win the live-search path, which is where most buying questions now route, and let the slower layer catch up as the web's picture of you improves.

Notice what's absent from those inputs: hacks. The centre of gravity is clean, crawlable HTML plus the same facts about your business repeated consistently everywhere machines look. That combination does most of the work, and everything below is a version of it.

The checklist: ten steps for UK service businesses

Work through these in order. The first six are foundations, the last four are where it gets competitive, and most of it is the same discipline that separates converting websites from merely busy ones - a problem I've covered in why your website gets visitors but no clients. Budget a focused week for the first pass and an hour or two a month after that.

1. Be legible to machines

Your site needs to be clean, crawlable HTML that shows its content without depending on scripts to load it, with your service and your place named in plain words: "chartered accountants in Luton", not "solutions for ambitious founders". Check that your robots.txt isn't blocking OAI-SearchBot, PerplexityBot or Google's crawlers, because a surprising number of sites block them by accident. This step is boring, and it does more for AI visibility than everything else on the list combined.

2. Answer questions the way people ask them

Buyers put full questions to assistants: "how much does a dental implant cost in the UK?", "do physios treat sciatica?", "can an accountant sort out late filing penalties?". Put those questions on your pages as headings with straight answers underneath, including prices or price ranges wherever you can bring yourself to publish them. The questions your front desk and your web chat field every week are the exact ones to write up.

3. Add schema markup that matches the visible text

Mark up your pages with LocalBusiness, Service and FAQPage structured data, and make the markup repeat what a human can see on the page. Schema that contradicts the visible content, or smuggles in claims the page doesn't make, gets ignored at best; Google's guidelines are explicit about this, and there's no reason to expect other crawlers to be more forgiving. Matching markup simply removes ambiguity about who you are and what you sell.

4. Complete your Google Business Profile, categories first

Gemini and AI Overviews sit on the same data that powers Search and Maps, so an unfinished profile caps what Google's assistants can say about you. Get the primary category exactly right, then fill in services, service areas, hours, photos and updates, and keep them current. Claim your Bing Places listing while you're at it: ChatGPT's search partners include Bing, which makes it quietly more important than its market share suggests.

5. Collect reviews with substance, and reply to them

"Great service, five stars" tells a machine almost nothing. "They rebuilt our garden wall in two days and the quote didn't move" names the job and the outcome, which is exactly the kind of sentence that gets paraphrased into an answer. Ask happy customers to mention what you did and how it turned out, and reply to every review, because replies are further evidence that you're active and accountable. This works as a system rather than an annual push, which is why our reviews service is built around that loop.

6. Keep your name, address and phone identical everywhere

Assistants cross-check. If your website says "R P Smith Plumbing Ltd", Google says "RP Smith Plumbing" and a directory says "Smith Plumbers Bedford", a machine can't be certain those are one business, and uncertain entities don't get recommended. Pick one canonical name, address format and phone number, then enforce it across your site, profiles, directories and social bios.

7. Be citable: publish one page that defines you

Every assistant that cites sources needs a page to point at, so give it one: a clear page stating what you do, where you operate, who you serve, how pricing works and what proof backs it all up. Add an llms.txt file if you like - a plain-text index of your key pages written for AI tools, ten minutes' work. Just know it's an emerging convention: no major assistant has publicly confirmed reading it when it answers, so treat it as tidiness rather than a ranking lever.

8. Earn mentions on pages assistants already trust

Assistants weight sources with standing, so one substantive mention on an established page moves you further than fifty thin directory entries. For UK service businesses that means the registers buyers already check (Gas Safe, CQC, the Law Society, your trade association), plus Checkatrade or Trustpilot where they fit, the local paper, the chamber of commerce, a sponsorship that comes with a write-up. Earned mentions are slow to build, which is precisely why they're trusted.

9. Test it monthly

Open ChatGPT, Perplexity, Gemini and Copilot and ask them your buyers' questions: "best physiotherapist in Bedford", "who can repair a flat roof near MK40", "recommend an accountant for a UK limited company". Log who gets named, which sources get cited, and whether anything said about you is wrong; a plain spreadsheet with date, question, assistant, names and sources is enough. Answers vary between sessions and between days, so a single spot-check proves little; the pattern across months is the signal.

10. Fix what the test shows, not what a vendor sells you

If the assistants keep citing a directory you're not in, join it. If they repeat an out-of-date price, correct the page they took it from. If a competitor keeps winning on review depth, that's your next quarter's work, not whatever dashboard is being pitched to you this month. The monthly log tells you where the gap is; work the gap.

What nobody can promise

Here's the caveat most articles bury. Nobody can pay, submit or trick their way into being the answer. There is no portal, no whitelist and no ad slot that places a UK service business inside the recommendation itself, whatever advertising the platforms eventually bolt on around it. The inputs are public and the answers change between sessions, so anyone promising you a guaranteed ChatGPT ranking is selling weather: they don't control the system, they can't see inside it, and they certainly can't stop the forecast changing next week. The same caution applies to tools that score your "AI visibility" to two decimal places: treat the number as a rough thermometer, not a target.

What an honest practitioner can promise is the work and the measurement: make your business as readable, verifiable and citable as it can be, then test whether the needle moves and adjust. The work stacks the odds; it doesn't force the outcome, and you should walk away from anyone who says otherwise. That's the standard I hold our own AI visibility work to, and it's the standard you should hold any agency to, including us.

Where this fits in a growth system

Being named by an assistant wins you the enquiry, not the client. The buyer ChatGPT sends still lands on your site at 9pm, still asks a question, and still books whoever answers first: in Velocify's analysis of around 3.5 million leads, contact within the first minute lifted conversion by 391%, which is why 60 seconds wins. Visibility that isn't wired into capture and follow-up leaks value at the final step.

This checklist is, near enough, the working method of our AI Visibility service: an audit of what the assistants currently say about you, the fixes in priority order, then the monthly test loop so decisions follow evidence rather than vendor decks. It runs alongside lead generation and the rest of the platform that captures and qualifies what the assistants send, and prices are on the pricing page. And if you would rather one team ran visibility alongside the capture, campaigns and reactivation it feeds, that is the job the Growth Partner plan exists for: one partner, the whole engine, measured on bookings rather than mentions.

If you take one thing from this, make it step nine. Ask the assistants your own buyers' questions this week and write down what they say. Most owners have never heard how machines describe their business, and that first log, flattering or not, shows you exactly where to start.