A client of mine asked me something recently that stuck. His neighbour needed a plumber, and instead of Googling one, she asked ChatGPT: "who's a good plumber near Pangbourne?" It gave her three names, with reasons, and she rang the first one. No search results page, no map, no scrolling. Three names.
If you run a local business, the obvious question is: how do I become one of those three names? There's a growing industry ready to sell you a mysterious-sounding answer. The real one is less exciting and much more useful: AI assistants recommend businesses they can find and verify in ordinary search indexes. Which means most of the work is local SEO you may already half-know, plus a couple of genuinely new wrinkles that most of your competitors haven't noticed yet.
I build and rank websites for UK tradespeople and small businesses, so I spend a lot of time watching where enquiries actually come from. Here's what I've learned about the AI side, honestly, including the parts nobody can control.
01What actually happens when someone asks ChatGPT for a plumber
It helps to strip the magic out of this. When someone asks an assistant a local question, roughly this happens:
- The assistant recognises it can't answer from memory (it has no idea who's trading in Pangbourne this week), so it runs a web search against a search index.
- It reads the top results: business websites, directories, review platforms, local listings.
- It summarises what it found into a short, confident answer, usually naming two to four businesses with a one-line reason for each.
That's it. There's no secret database of approved businesses, no form to submit your company to OpenAI, and no paid placement inside the answers. The assistant is doing what a diligent friend would do: searching, skimming and reporting back.
Two consequences fall straight out of that. First, if the index can't find you, the assistant can't recommend you. A business with no website, or an unindexed one, simply doesn't exist to it. Second, if your details are inconsistent, one phone number on your site, a different one on an old directory, a third on Facebook, the assistant either garbles them or quietly skips you for someone whose story checks out. Machines summarising sources hate contradictions even more than Google does.
02The index behind the answer (and why it's probably Bing)
Here's the detail that changes what you should actually do. ChatGPT's web search is built on Microsoft's plumbing, not Google's. OpenAI's search partnership is with Microsoft, and studies through 2025 and 2026 have consistently found that the sources ChatGPT cites overlap heavily with Bing's top results, with some analyses putting the overlap at well over 80%.
I'll hedge that properly, because this space shifts fast: OpenAI has been building more of its own crawling and ranking over time, the exact blend isn't public, and the arrangement could change with one press release. Other assistants differ too: Microsoft's Copilot is Bing through and through, Google's Gemini and AI Overviews lean on Google's index, and the smaller players mix and match. Nobody outside these companies knows the recipes exactly.
But the practical takeaway survives all that uncertainty: Bing matters far more than its 4% UK browser share suggests. Bing's index feeds Bing itself, DuckDuckGo, Yahoo, Copilot and (in large part) ChatGPT's web search. Most UK small businesses have spent years optimising for Google and have literally never looked at how they appear in Bing. If a meaningful slice of your future customers is asking assistants instead of Googling, that blind spot now has a cost.
ChatGPT can only recommend what its search index can find and verify. Be findable in Bing as well as Google, keep your details consistent everywhere, and you're ahead of almost every competitor already.
03Mostly, this is local SEO wearing a new hat
Now the part that should be reassuring. About 90% of what gets you recommended by an assistant is exactly what gets you ranked in a normal local search. If you've read my piece on ranking on Google in 90 days, none of this will surprise you:
- A fast, crawlable website. Assistants read pages the same way search crawlers do. A site that loads in under a second with clean HTML gets read fully and quickly. A bloated template that takes five seconds and hides its content behind scripts may get skimmed or skipped. This is one of the quieter reasons I custom-code sites rather than using builders.
- Consistent name, address and phone. The same details, spelled the same way, on your website, Google Business Profile, Bing Places, Facebook and any directories. Boring, unglamorous, and genuinely load-bearing.
- Real service pages that name real towns. "Boiler repairs in Pangbourne, Goring and Whitchurch" on an actual page beats "quality plumbing solutions" on a generic one, both for search engines and for an assistant deciding whether you serve the area it was asked about.
- Reviews, plural, and recent. An assistant naming "a good plumber" wants evidence of good. Star ratings and review counts are the easiest evidence on the internet to summarise.
If your business currently doesn't show up well in ordinary search, fix that first; the same repairs carry over. I've written about the usual causes in why isn't my business on Google, and the Google Business Profile guide covers the single highest-impact free listing. None of that work is wasted in an AI world. It's the foundation the new stuff sits on.
04The Trustpilot wrinkle most competitors miss
Here's the concrete edge, and it's the one thing in this article I'd act on this week if I ran a trade business.
Google's map pack runs on Google reviews. Bing's local results don't. Bing pulls ratings into its local panels from third-party platforms, and in the UK that prominently means Trustpilot, alongside Facebook and Yell. Trustpilot reviews also surface through Bing and Yahoo more broadly via Microsoft's merchant ratings integration.
Follow the chain: Bing's local results lean on Trustpilot, Bing's index feeds DuckDuckGo, Yahoo, Copilot and much of ChatGPT's search, therefore a modest Trustpilot presence makes you look established across the entire non-Google internet at once. Meanwhile, the fifty Google reviews you worked hard for are largely invisible on that side of the fence.
Almost every small business I audit has this exactly backwards. They obsess over Google reviews (rightly, Google still sends most of the traffic) and have zero presence anywhere else. So in Bing, DuckDuckGo and the assistants that read them, they look like a business nobody has ever reviewed, while a competitor with eight Trustpilot reviews looks like the safe recommendation.
Here's the split, roughly, as it stands in mid-2026. Treat the right-hand column as "strongly influenced by", not gospel, because these arrangements do shift:
| Where the customer asks | Index behind it | Reviews that count most |
|---|---|---|
| Google search / Maps | Google reviews | |
| Google AI Overviews / Gemini | Google reviews | |
| Bing | Bing | Trustpilot, Facebook, Yell |
| ChatGPT web search | Largely Bing + own crawling | Whatever the sources cite, so Trustpilot again |
| Copilot | Bing | Trustpilot, Facebook, Yell |
| DuckDuckGo / Yahoo | Bing | Trustpilot, Facebook, Yell |
One column of that table runs on Google reviews. Every other row leans the same way, and most of your competitors are invisible in all of them. That's the gap.
Keep collecting Google reviews as your first priority, my guide to getting more Google reviews covers how. Then open a free Trustpilot business profile and ask your next five to ten happy customers to review you there instead. You don't need hundreds. Going from zero reviews to eight genuine ones is the single biggest visibility jump you can make on the Bing side of the internet, and it typically costs nothing but a few polite asks.
While you're at it, claim your Bing Places listing at bingplaces.com. It's Bing's version of a Google Business Profile, it takes twenty minutes (you can import your Google profile directly), and it's the closest thing that exists to registering your business with ChatGPT. Most UK tradespeople have never touched it, which is precisely why it's worth doing.
05Being quotable: write pages an assistant can lift answers from
The second genuinely new-ish habit: assistants don't rank pages, they quote them. When ChatGPT says "XYZ Plumbing covers Pangbourne and offers emergency call-outs", it lifted that from somewhere. Your job is to make your site the easiest place to lift it from.
In practice that means:
- Say plainly who you are, where you are and what you do. An about page that reads "We're a family-run plumbing firm based in Pangbourne, covering West Berkshire and South Oxfordshire since 2009" is machine-quotable. "Passionate about delivering excellence" is not. It contains no facts.
- Answer real questions in plain HTML. A short FAQ on your service pages ("Do you charge a call-out fee?", "Which areas do you cover?", "Are you Gas Safe registered?") gives an assistant ready-made sentences. Text buried in images, PDFs or script-heavy widgets is invisible to it.
- Use structured data. LocalBusiness and FAQ schema is machine-readable labelling for your name, address, opening hours, service area and reviews. Search engines have read it for years; it removes any ambiguity about your basic facts. Every site I build ships with it as standard.
- Keep one page per real question where you can. A page titled "Boiler repair costs in Berkshire" that actually answers the question will get cited. A homepage that mentions forty things briefly won't.
Notice that all of this also makes your site better for humans. That's not a coincidence. Assistants were trained on human judgements of usefulness, so the reliable way to look good to them is to genuinely be useful. There's no separate machine audience to trick.
06What doesn't work (and what's snake oil)
Because a new channel always attracts old cons, dressed in new acronyms:
- Keyword stuffing. Writing "best plumber Pangbourne" forty times has not worked on Google since roughly 2012, and it works even less on a system that reads your page like a person would. It makes you look exactly like what you are: a page trying too hard.
- Fake reviews. On Trustpilot or Google, bought reviews get filtered, and getting caught can nuke the genuine ones with them. Worse, assistants cross-reference sources; a business with fifty glowing reviews and no other footprint reads as suspicious, not established.
- "AI SEO" packages. If someone offers to "optimise your business for AI" for a chunky monthly fee, ask them for the task list. In 2026 it will be some subset of: fast site, consistent details, service pages, schema, reviews, Bing Places. All real, none proprietary, and most of it is one-time work, not a £300-a-month subscription.
- Anyone guaranteeing placement. There is no reliable way to guarantee a Google ranking, and AI answers are a step less predictable than that. A guarantee about what ChatGPT will say is a guarantee about something the seller does not control. Walk away.
07An honest note on uncertainty
I want to be straight about the limits of everything above. This corner of search is moving faster than any I've watched. The index partnerships could be renegotiated, assistants are adding their own crawlers and local data deals, and the way answers get assembled changes between model versions without anyone announcing it. Some of the specifics in this article will age; the Bing overlap numbers in particular are a snapshot, not a law of physics.
Nobody, including me, can promise you a slot in an AI answer. What I can say with confidence is that every plausible version of this future rewards the same things: a fast site that machines can read, business details that agree with each other everywhere, real reviews on more than one platform, and pages that plainly answer the questions your customers ask. Those aren't bets on one company's architecture. They're bets on "systems that summarise the web will favour businesses whose web presence is clear and credible", which is about as safe as bets get.
And the timeline logic is the same as ordinary search: this compounds, slowly, then noticeably. If you want a realistic sense of how long the underlying work takes to pay off, I've written that up in how long does SEO actually take. If your site exists but the phone isn't ringing from any channel, start with why your website isn't getting enquiries instead; AI visibility won't rescue a site with a broken foundation.
08The checklist
Everything above, as a list you can actually work through:
- Fast, crawlable website with real service pages naming the towns you cover.
- Identical name, address and phone on your site, Google Business Profile, Facebook and every directory you're on. Fix the strays.
- Google Business Profile claimed, verified and actively used.
- Bing Places claimed (import from Google, twenty minutes, free).
- Google reviews as a steady weekly habit.
- Trustpilot profile with your next five to ten happy customers pointed at it.
- Plain-English about page stating who, where and what, plus short FAQs on service pages.
- LocalBusiness and FAQ schema on the site.
- Nothing dodgy: no stuffing, no bought reviews, no "guaranteed AI rankings" invoices.
A sole trader could clear most of this in a few focused evenings. The website items are where people tend to stall, and that's the part I do for a living.
Want to know how findable you actually are?
I'll happily take a look. My free website check covers speed, structure and how your business shows up in search, and if you don't have a site yet (or have one you'd rather not admit to), I'll build you a free custom preview of your homepage before you pay anything. No contracts either way.
A full custom-coded site is £100 to build plus £20/month for hosting, domain and maintenance. If you want the visibility work done for you, the Local Growth Programme at £500 plus £60/month covers the site, the listings, the schema and ongoing local SEO. Details on the services page, and 13 live client sites with a 5.0★ average rating on the work page if you'd like to judge for yourself.
Get my free preview