Joy Hawkins at Sterling Sky ran the same set of queries through both formats and counted how many distinct businesses each surfaced. The traditional three pack returned 18,330. The AI generated local results returned 5,943. That is roughly a third of the coverage, and the AI versions typically name one or two businesses rather than three, with no call button attached.

For a business that was appearing at position three, that is the whole difference between being found and not being found.

Two numbers that disagree

Here is where the honest version of this gets complicated, and where most coverage picks a side.

Whitespark studied 540 manual queries across six local industries and found AI Overviews appearing on about 68 percent of them. Local Falcon looked at queries like plumber near me and found AI Overviews on roughly 7 percent.

Both are probably right. They are measuring different things. Somebody typing plumber near me has already decided what they want and Google still hands them a map. Somebody typing a question about their actual problem, which is how people search when they are not sure what they need yet, is considerably more likely to get a synthesised answer.

So the exposure depends on where in the decision your customers find you. If they arrive knowing your trade and wanting somebody nearby, little has changed. If they arrive describing a symptom, a lot has.

The direction is not ambiguous

At its developer conference in May, Google said AI Mode had passed a billion monthly users and that its query volume had more than doubled every quarter since launch. BrightEdge tracked AI Overviews appearing on about 48 percent of queries by February, up from 31 percent a year earlier.

The same event reported that overall search volume reached an all time high, which is worth holding onto. This is not people searching less. It is the same demand arriving through a narrower opening.

What being cited is worth

Seer Interactive found that brands cited in AI Overviews earned roughly 120 percent more organic clicks per impression than uncited brands on identical queries. That is a large gap and it points at the actual shape of the problem.

The distribution is getting less even. Previously a local search produced three visible businesses and ten more a click away, and being sixth was worth something. Now a single query might name two, and everybody else is invisible for that search. The upside for whoever gets named is proportionally larger.

Concentration cuts both ways, and for a small business it usually cuts the wrong way. But not always, which is the part worth thinking about.

What actually gets you named

Nothing on this list is new, which is either reassuring or frustrating depending on your temperament.

These systems assemble answers from your business profile, your reviews, your website, and whatever else they can corroborate about you. They favour information that appears consistently in several places, because consistency is the cheapest available proxy for reliability.

That means the work is the same work. A complete and current business profile. Reviews that are recent rather than merely numerous. Hours, address, and service descriptions matching everywhere they appear. A website that says plainly what you do, where you do it, and what it costs.

The difference is that these things used to improve your position and now they determine whether you exist in the answer at all. The threshold moved from ranking to inclusion.

Where a small business has an advantage

Specificity. An AI answer assembled from vague sources produces a vague recommendation, and vague recommendations get skipped.

A business whose website says it works on pre 1970 properties in three named neighbourhoods gives these systems something concrete to match against a specific question. A business whose website says it provides quality service to the local area gives them nothing, and will be omitted in favour of somebody who was clearer.

Most competitors in most local markets are still writing the second kind. That gap is available and it costs nothing but the willingness to be specific about what you do and refuse.

What to do about it, in order

Ask your customers how they found you and write the answers down. Six months of that record will tell you more about your own exposure than any study, because it tells you whether your customers arrive knowing what they want or describing a problem.

Then check what an AI answer actually says about your category in your area. Search the way a customer would, in an incognito window, and see whether you appear. If you do not, look at who does and what their information looks like.

Then fix the boring things. Profile completeness, review recency, consistent details, and a site that answers the questions people actually ask before they call. There is no separate optimisation for this, whatever anybody is selling.

The thing not to do

Do not abandon what works because a study frightened you. The map pack still exists, still drives calls, and still responds to the same signals it always did. Businesses that dismantled their local search work in response to the first wave of AI coverage lost the channel that was still producing customers.

The reasonable position is that one channel is narrowing while another opens, both reward the same underlying work, and neither is worth panicking about in a business that has been trading for eight months.

A note on measurement

Your analytics will not show you this cleanly. There is no report distinguishing a customer who found you through an AI answer from one who typed your name, and the tools adding this are early and inconsistent.

Which returns to the same conclusion as most measurement problems at this scale. Ask people. It costs one question at the start of a conversation you were having anyway, and it produces better information than any dashboard you could install this year.