AI Overviews and local search: the split nobody tells local businesses about
Local Falcon found summaries on 15% of simple local queries and 92% of informational ones phrased locally. The split is by intent, not geography, and it changes what you measure.
Local businesses have been told two contradictory things this year. One is that AI summaries are coming for their traffic like everyone else's. The other is that local is safe because a generated paragraph cannot hand out a phone number. Both are wrong in the same way: they treat "local" as one category, and it is two. A study by Local Falcon found AI Overviews on 15% of simple local-intent queries and on 92% of informational queries phrased locally. That is not a margin of error. It is a different situation depending on which half of your traffic you are looking at.
Quick answer
- "Tacos San Francisco" almost never gets a summary. 15% in the Local Falcon sample.
- "How long does an eye exam take near me" almost always does. 92% in the same sample.
- The split is by intent, not by geography. Wanting a place is safe; wanting to know is not.
- Which half you live on decides everything, and most local businesses have never separated them.
| Query shape | AI Overview | What it means for you |
|---|---|---|
| "dentist near me", "tacos San Francisco" | Rare | The map decides; nothing much changed |
| "how much does a crown cost" | Common | The answer is served; the click is not |
| "how long does X take near me" | Very common | Local wording, informational intent |
| Your business name | Almost never | Tells you nothing about the rest |
What kind of change this is
The mental model that gets local businesses to the wrong conclusion is that Google decides by place: if the query mentions a city or says "near me", it shows a map and skips the summary.
It decides by what the person wants. Someone who wants a place gets places, because no paragraph can be visited or called. Someone who wants to understand something gets an explanation, and it makes no difference that they typed "near me" at the end. The local phrasing is a habit, not an intent.
That is why a dental practice can be simultaneously untouched and heavily hit. Its "dentist near me" traffic is fine. Its "how much does an implant cost" traffic, which is the part that fills the calendar with high-value appointments, is exactly the part where the answer now appears above the results.
Do AI Overviews affect local search?
Unevenly, and the split is by intent rather than location. Local Falcon found AI Overviews on 15% of simple local-intent queries and 92% of informational queries with local phrasing. A business that only checks "near me" searches concludes nothing changed, while the half that brings considered customers is the exposed one.
Why the safe half is safe
The reasons matter more than the percentage, because they predict cases the study did not cover.
A generated paragraph cannot complete the transaction. It cannot hand over a phone number that connects, show whether you are open right now, or take a booking. When the intent is to go somewhere, the format that resolves it is the map, so the summary adds nothing and takes space.
And the answer changes by the hour. Opening times, availability, current wait. That is data pulled from a listing, not text worth generating, which is why the listing carries more weight here than any page you write: how to rank in Google Maps.
This is the half where the old playbook still works exactly as it did, and where the panic being sold to local businesses is misplaced.
Why the other half is not
Every question a customer asks before choosing is informational. How much does it cost, how long does it take, does it hurt, what is the difference between the two options, do I need an appointment. Those queries have always brought the most valuable traffic to a local business, because whoever asks them is deciding and not browsing.
They are also, word for word, the queries a generated summary answers best. So the exposure is concentrated in precisely the content a practice or a shop spent years writing, and not in the "near me" searches everyone monitors.
The consequence is a measurement trap. A business whose reporting is built around map visibility and "near me" rankings will see a flat line and conclude it is unaffected, while its informational pages quietly stop bringing anyone. The two live in the same Search Console property and the total hides both.
What to actually do about it
Separate the two halves before deciding anything. Export your queries and split them: those that would trigger a map, and those that ask a question. Compare impressions against clicks in each group, month over month, on a fixed list. If the informational group shows impressions holding while clicks fall, that is the summary, and no amount of listing work touches it. Setting that comparison up properly starts with the tool itself: how to set up Google Search Console.
On the safe half, keep doing what worked. Listing complete, categories right, reviews collected steadily, hours accurate. Nothing on that side of the split was disrupted by any of this, and the temptation to change it because the industry is anxious is the most expensive reaction available to a local business right now: local SEO for small businesses.
On the exposed half, change the metric before changing the content. An informational page that loses clicks but keeps impressions is still doing something: it is putting your name in front of someone deciding. Judging it by visits will get it deleted; judging it by whether your business is the one named is harder and closer to what it is now worth, and it needs a different set of indicators from the start: how to know if SEO is working.
And add the thing the summaries pull from. Costs, processes and common questions on your own site, and a presence on the review and community platforms these systems draw on. Collecting reviews steadily is the cheapest half of that and the one most businesses postpone: how to get Google reviews. A business that only ever fed its Google listing has nothing for a summary to cite.
When this is not your problem
When your traffic is mostly your own brand name. Summaries barely appear there, and any conclusion drawn from that segment will be wrong for the rest.
When you have no informational content at all. There is nothing to lose yet, and the question is whether to build it knowing what it now returns.
And when the listing is not verified. Nothing above applies until the business appears at all: why your business does not appear on Google Maps.
Mistakes that repeat
- Treating "local" as one category. It is two, split by intent, and they behave in opposite ways.
- Checking only "near me" rankings. That is the half that did not change; it hides the half that did.
- Concluding from brand-name traffic. Summaries barely appear there and it represents nothing.
- Deleting informational pages that lost clicks. They may still be putting your name in front of buyers.
- Reacting to a single percentage. Presence measurements range from 25% to 60% depending on the study.
- Assuming the listing protects everything. It protects the map half and nothing else.
Data and transparency
The 15% of simple local-intent queries and 92% of informational queries with local phrasing come from a Local Falcon whitepaper on AI Overviews and local business visibility, checked on 13 August 2026, and are cited with that attribution because the split, not the exact figure, is what the article rests on. The Seer Interactive study measured organic CTR falling from 1.76% to 0.61% on queries with an AI Overview, alongside 2.72% to 1.62% on queries without one; the second series is quoted with the first because part of the decline was not caused by the summary. General presence measurements range from 25.11% (Conductor, 21.9 million queries, Q1 2026) to 60.32% (Xponent21, US, April 2026), are not comparable with each other and are not averaged here. No figure is given for how much revenue a local business loses: that depends on the mix between the two halves, which is exactly what this article says you have to measure yourself. Verified as of August 2026.
Primary sources, opened on 13 August 2026: the Seer Interactive study.
What this changes
The most useful thing in the Local Falcon numbers is not either percentage. It is that they are so far apart.
A fifteen and a ninety-two in the same category means the category was the wrong unit all along. Every conversation a local business has had this year about whether AI is coming for its traffic has been conducted at a level of description too coarse to answer the question, which is why the answers have been contradictory and everyone has been able to find support for whichever one they preferred.
The version worth having is narrower and duller: which of my queries are people asking to go somewhere, and which are they asking to find something out? One list is fine. The other one needs a different plan and a different way of being counted, and nobody was going to notice the difference by watching a total.
Frequently asked questions
Are local businesses safe from AI Overviews?
Half of their traffic is, and half is not. Queries where someone wants a place rarely trigger a summary, because a paragraph cannot be visited or called. Queries where someone wants to understand something trigger one very often, even when the phrasing includes "near me". Since the second group is usually where the high-value customers come from, "safe" is too strong.
Does "near me" protect a query from getting a summary?
No. The phrasing is a habit; the intent is what decides. "How long does an eye exam take near me" is an informational question with local wording, and in the Local Falcon sample those got a summary 92% of the time. The protection comes from wanting a place, not from mentioning one.
Should I stop writing informational content for my local business?
Not on that evidence alone. What changes is what you expect from it and how you judge it. A page that loses clicks but holds impressions is still placing your name in front of someone who is deciding, which has value that visit counts do not capture. Deleting it because traffic fell is the most common overcorrection in this whole subject.
How do I know which half my traffic is on?
Export your queries from Search Console and split them into two lists: those that would show a map and those that ask a question. Then compare impressions against clicks in each list across two comparable months. The pattern to look for in the informational list is impressions holding steady while clicks fall.
Does optimizing my Google listing help me appear in summaries?
Not directly. Summaries draw on web content, review platforms and community sites rather than on listing fields. A business that has only ever worked on its listing has nothing for a summary to pull from, which is why costs, processes and common questions on your own site matter more here than any listing setting.
Most sites do not have a ranking problem
They have a what-happens-next problem. You can rank first and still sell nothing. The diagnostic looks at both and tells you which one is costing you money.
See the diagnostic