What the research says about being found on Google, in AI answers and by voice assistants. Every figure carries the study it came from, and says so where the only data is American.
Five years ago this was simple. Someone’s boiler packed in, they typed “emergency plumber near me”, they got a list of links, and they clicked two or three of them. Your website was the front door and search was the road to it.
That road has been rebuilt.
The results page now answers the question itself. A map with three businesses sits at the top, each with a star rating and a call button, and most people never scroll past it. The local map pack has always worked this way — customers ring you, tap for directions, or read your reviews without ever visiting a website. That is not new.
What is new is how much of the rest of search now behaves the same way.
In the first four months of 2026, 68.01% of Google searches ended without a click on anything. That figure is American and covers browser-based searches only — there is no published UK equivalent yet.
American research also found that when Google shows an AI summary, people click a normal result 8% of the time rather than 15%. Google has publicly disputed how that study was done.
And a third route has appeared that did not meaningfully exist two years ago: people asking an AI assistant directly. The best available data on this is American, and it is striking.
AI tools are now the third most-used way American consumers find local businesses, behind only Google and Facebook, used by 45% of them — up from 6% the year before. Take-up is highest among 30 to 44-year-olds, at 64%, which is the age group that owns houses and pays for work on them.
There is no British version of that survey yet. We are not going to pretend those are UK numbers. But the machinery underneath does not change at the border: the same assistants, reading the same kinds of sources, are being asked the same questions in Bolton as in Boston.
None of that means your rankings are about to vanish. The searches that actually bring you work are among the least disrupted on the internet, and the figures behind that are on our search page.
SEO means making sure that when somebody near you searches for the work you do, your business is one of the names that comes up.
Google decides the local map pack on three things: how well you match what was searched (relevance), how close you are to the person searching (distance), and how well known and well regarded you are (prominence).
Two of those you cannot buy your way out of. You are where you are. But relevance and prominence are built, and there is a clear pecking order for how.
The single biggest individual lever in the local map pack is the primary category on your Google Business Profile — the one-line answer to “what is this business”. Proximity comes second. Reviews are a top-three factor.
That first point catches people out constantly. A business listed as a “contractor” and a business listed as an “emergency plumber” are not competing on equal terms for someone typing “emergency plumber”, no matter how good the rest of the site is. The narrowest accurate category wins.
On reviews, practitioners consistently report the same thing: a steady trickle of recent reviews does more for you than a large pile of old ones, and answering them promptly matters too. These are agency observations rather than anything Google has confirmed, so treat them as sensible practice rather than a rule.
Early reports also suggest Google’s newer AI-written summaries on Maps draw on your reviews, posts and web content, so a thin or neglected profile can read poorly even with a good star average.
We build real pages for the real work you do, in the places you actually cover. Not a template with the town name swapped out twenty times.
That distinction is not stylistic. Google’s March 2026 core update reportedly hit templated location pages in home services particularly hard, and Google now expects the website link on a business profile to point at a specific, relevant page rather than a generic homepage.
Two things follow from that, and both are measurable. The first is LocalBusiness structured data: your name, address, phone, hours and service area, written in the format Google reads directly. It is one of the schema types that still produces rich results in 2026, and for a local business it is the one that counts.
The second is pages that load fast and do not jump about. Google’s targets are a largest content paint under 2.5 seconds, interaction latency under 200 milliseconds and near-zero layout shift, measured on real visitors rather than in a lab.
On that second point, we will be straight with you, because plenty of agencies are not. Page speed is a ranking signal, not a ranking system in its own right — it acts as a tiebreaker between pages that are otherwise comparable, and it will not on its own lift you above a better-established competitor. Its real value is that people do not abandon a fast site. Only around 48% of mobile pages pass all three of those measures, so clearing them still puts you ahead of about half the web.
AEO means making sure that when the search engine answers the question itself, on the page or out loud, the answer it gives is yours.
Because a growing number of your customers are not looking at a screen when they ask. And the assistant they ask determines which listing gets read out.
Google Assistant and Gemini read your Google Business Profile and Google Maps.
Siri reads Apple Maps. It does not use Google Maps. Apple Maps takes its business information from Apple Business Connect, which is a separate and free listing.
That second line is worth reading twice. You can have an immaculate Google Business Profile, five years of reviews and a top spot in the map pack, and still be completely invisible to every customer who holds down the button and says “find me an emergency electrician”. Siri is not looking at Google. It never was. Claiming an Apple Business Connect listing is free and takes an afternoon, and it is one of the few genuinely underused wins left in local search.
For years the standard advice was to add FAQ markup to your pages to win a bigger slot in the results. That is over.
FAQ rich results stopped appearing in Google Search on 7 May 2026. The Search Console report and Rich Results Test support were withdrawn in June 2026, and API support ends in August 2026. Google has confirmed rankings are unaffected — this changed how results look, not how they rank. Existing markup does no harm and can stay.
If somebody is still selling you FAQ schema as a visibility tactic, they stopped reading in 2023.
The markup trick is gone. The structure it was imitating still works, and it works for three audiences at once.
Under a heading that asks a real question, we put a direct answer of roughly 40 to 60 words, then expand underneath. That length is not arbitrary: it is about what a voice assistant reads aloud in 15 to 20 seconds, it is the shape search engines lift for a featured answer, and it happens to be how an actual customer wants a question answered. This is well-replicated craft rather than a published Google rule, but it costs nothing to do properly.
GEO means making sure that when somebody asks ChatGPT or Gemini for a builder in their area, your business is one of the ones it names.
Most of what is sold as GEO is ordinary SEO. Google has said so itself, in writing, and we quote it in full, with what the research puts the proportion at, on our search page. What follows here is the genuinely different remainder.
There is a real remaining layer, and it is worth having. It is just smaller than the pitch.
Traditional SEO is a race for position. AI answers are a question of whether you get cited at all, and a page outside the top ten can still be the source a model draws on. Different target, different tactics.
Ahrefs’ large-scale study found that the strongest correlations with appearing in AI answers were off-site signals — being mentioned around the web — running roughly three times stronger than backlinks. Two caveats come from Ahrefs themselves: correlation is not causation, and their sample was established brands rather than sole traders, so it may not transfer cleanly to a one-van operation.
Separately, one analysis found around 82% of AI citations come from earned media, though that is a single source and we would not build a plan on it alone.
The founding academic work on this found that adding citations, direct quotations and specific figures measurably improved how often a page was drawn on, and that keyword stuffing did nothing at all.
As BrightLocal’s chief executive put it: “If your reputation only exists on Google, you are effectively invisible to the millions of people using ChatGPT to find local services.”
Early evidence suggests the engines overlap surprisingly little in which sites they draw on. Being found in one is not being found in the others.
Put that mechanism together and it becomes the case for having a website at all, rather than a listing alone. That is the case we make, in short, on our search page.