AI Overviews are changing who gets found locally

If you've ever paid someone to "get you ranking on Google", the thing you were buying has quietly changed shape. Google now answers a lot of local searches itself, in a block above everything else, and that block names far fewer businesses than the old map pack did. This post goes through what the research actually found, what it means for a plumber or a clinic in a single town, and what's worth doing about it — honestly, without pretending anyone has this fully solved.

What an AI Overview is, in plain terms

When you search for something now, you'll often get a few paragraphs of generated text at the top of the page. It answers the question directly, pulls from a handful of sources, and cites them with small links. Below that, you still get the map pack and the blue links — but they've been pushed down, and a decent share of people never scroll that far because they've already got their answer.

For an informational search ("why is my boiler losing pressure") that's obviously useful to the searcher. For a commercial local search ("emergency plumber Leeds") it changes who gets seen. Instead of a map with three businesses and a list of ten more, you get a summary that may name two or three firms, or sometimes none at all — just general advice about what to look for in a plumber.

The numbers, and where they come from

There are three pieces of research worth knowing. I'll be specific about what each one measured, because the headlines get mangled.

Clicks on queries with an AI Overview

Ahrefs' February 2026 study found AI Overviews reduce organic click-through by 58% on the queries where they appear. The important bit isn't the 58% on its own — it's the trend. Eight months earlier, Ahrefs measured the same effect at 34.5%. So the drag is getting stronger, not settling down. Anyone who told you in mid-2025 that this would blow over was wrong, and the data says so.

Seer Interactive looked at 5.47 million queries and found organic click-through rate on AI-Overview queries fell from 1.76% to 0.61% at one point — then partially recovered to 2.4% by February 2026 as users got used to the format. That recovery matters. It suggests people are learning when to trust the summary and when to go look for themselves, and the floor wasn't permanent. But "partially recovered" is doing real work in that sentence. The pattern is volatile, not stable.

What's happening specifically in local search

This is the part most relevant if you serve one town. A Whitespark case study found AI Overviews now appear on roughly 68% of local business queries, compared with the traditional local 3-pack appearing on about 39%. Read that again: the AI answer is now more common on local searches than the map pack is.

And across 322 local markets studied, the AI-generated local results showed 68% fewer businesses than a traditional map-pack listing would, with most of those markets showing fewer distinct businesses represented at all. Fewer names, repeated more often.

The upside, for whoever gets named

The same research found that businesses cited inside an AI Overview earn roughly 120% more organic clicks per impression than uncited businesses on the same query. Visibility hasn't disappeared. It's concentrated. Being in the summary is now worth more than it used to be worth being on page one, and being absent from it costs more than it used to cost being fourth in the map pack.

Why fewer businesses get shown

It helps to understand the mechanism rather than treating it as Google being arbitrary. An AI Overview is generated from sources the model can read and summarise. To name you in an answer about emergency plumbers in Leeds, it needs something to draw on that clearly says who you are, where you work, what you do and what evidence backs that up.

A map-pack listing had three slots and an algorithm that filled them by proximity, prominence and relevance. It was a ranking. An Overview isn't a ranking — it's a summary, and summaries are shorter than lists. If three businesses make the point adequately, there's no editorial reason to add a fourth. That's why 68% fewer businesses appear. Nothing against you; the format just doesn't need you.

The practical consequence is that "being findable" and "being summarisable" are now two different things. A thin one-page site with a phone number was always findable. It gives a model almost nothing to summarise.

What this actually means for a small business

1. A Google Business Profile is still necessary, and no longer sufficient

The profile still feeds the map pack, which still shows on about 39% of local queries. Keep it accurate, keep collecting reviews, keep the opening hours right. But if the AI answer appears nearly twice as often as the map pack, the profile alone now covers less of the ground than it did.

2. Your own website matters more than it did two years ago

This is the uncomfortable irony. The same shift that reduces clicks to websites also increases the value of having one with real content on it. An Overview has to cite something. Directory listings, your Business Profile, review sites and your own pages are all candidates — and your own pages are the only ones you control.

What tends to be citable: specific service pages that name the service and the area in plain language, straightforward answers to the questions customers actually ask, prices or price ranges if you're willing to publish them, and evidence of real work. What tends not to be: a homepage with a stock photo, a three-word tagline and a contact form.

If you don't have a site worth citing, building one is no longer a vanity project. RedAgents' website builder exists for exactly that gap, and the walkthrough of how it works covers what you get and where it stops. Worth reading alongside why AI-generated websites tend to look identical — because the sameness problem is real, and it matters more when the thing reading your site is looking for something distinctive to say about you.

3. Impressions without clicks are still worth something

If someone reads an Overview that names you, doesn't click, but then searches your business name directly or just rings you — that's a conversion your analytics will never attribute properly. The Seer finding that CTR partially recovered to 2.4% suggests people do go looking once they've got a shortlist. Don't assume a flat click graph means a flat business.

Which is also an argument for making sure the phone gets answered when they do ring, because that call may be the only trace of the whole journey. The honest arithmetic on missed calls is a separate problem, but it compounds with this one.

4. Stop buying "ranking" as a product

If someone sells you position one on a keyword, ask what's above position one on that query. If it's an AI Overview that names three competitors, you've bought a seat behind a wall. The question to ask an SEO supplier now is: "On which of my target queries does an AI Overview appear, and who does it cite?" If they can't answer that, they're selling you 2019.

What's genuinely uncertain here

I'd rather be straight about the limits of this research than present it as settled.

• The numbers move fast. Ahrefs' own measurement nearly doubled in eight months. Seer's CTR figure fell and then partly recovered. Any figure in this post is a snapshot, not a constant.
• Averages hide enormous variation. A 58% average click reduction across all queries tells you very little about your query in your town. Some local searches still show no Overview at all.
• "How to get cited" is not a solved discipline. Nobody can sell you a reliable method for appearing in an AI Overview, and you should be wary of anyone claiming otherwise. What we know is that models cite sources with clear, specific, verifiable content. That's a direction, not a recipe.
• The 322-market study is one study. It's good evidence. It isn't the final word on every sector and every country.

Where this doesn't matter much

Not every business should reorganise around this. Some genuinely shouldn't bother.

If most of your work comes from repeat customers and referrals — a lot of established trades, a lot of salons with a full book — search is a small slice of your pipeline and a shrinking slice is still small. Fixing your quote follow-up or your rebooking habit will do more for you than any amount of content. Why customers go quiet after a quote is probably the more profitable read.

If you sell to other businesses rather than consumers, local search was never the main channel anyway. Direct outreach does the work there, and contact-form outreach versus cold email covers that ground properly.

And if you're at capacity — booked out six weeks, turning work away — more visibility is a cost, not a benefit. Put the effort into pricing instead. How to price a job you won't resent is the better use of an evening.

A short, honest to-do list

If you've decided this does affect you, here's the order I'd work in. No tools required for the first two.

1. Search your own top five queries, from a phone, logged out. Write down whether an AI Overview appears, and which businesses it names. That's your actual baseline. It takes fifteen minutes and it's more useful than any report you'll be sold.
2. Check whether your site answers the questions in those Overviews. If Google's summary explains what to look for in an emergency plumber, and your site doesn't address a single one of those points, you've found your content gap.
3. Write the specific pages. One per service, one per area you genuinely cover, in plain language, with real detail. Specifics get cited; adjectives don't. If you're using AI to draft them, the usual complaints about AI writing tools are worth knowing first — generic output is exactly what won't get cited.
4. Keep the Business Profile tidy. It still covers roughly 39% of local queries and costs nothing but attention.
5. Measure phone calls and direct traffic, not just clicks. If Overviews shift people from clicking to calling, clicks will understate you. Direct searches for your business name are the signal to watch.

The honest summary of the shift is this: Google is doing more of the answering, naming fewer businesses, and rewarding the ones it names more heavily than before. That's harder if you're invisible and better if you're specific. The businesses most exposed are the ones with no website worth reading and no reason for anyone — human or model — to pick them out of a list of twelve identical firms. That was always a weak position. It's just become an expensive one.