Tag: Google AI Overviews

When someone searches for your business by name on their phone, the panel that appears is often the first and only thing they read. For years, that panel has been built from your Google Business Profile: hours, phone number, photos, reviews and a map pin. In September 2026, marketers began reporting something new at the top of it, an “AI Overview” label above a short, AI-written description of the business.

It is a test, not a launch, and Google has not announced it. But it points in a direction every local business should take seriously, because the description is no longer only what you wrote. This piece covers what was reported, what is and is not known, and a practical checklist of what to check now.

What was reported

On September 17, 2026, Search Engine Roundtable reported that some local knowledge panels, which are Google Business Profiles, now show an “AI Overview” label. A Show more button below the description expands the panel into a chat-style follow-up experience that resembles AI Mode. Ben Fisher, co-founder of the local search agency Steady Demand, posted about it first, and Barry Schwartz of Search Engine Roundtable said he was able to reproduce it himself.

On September 23, the UK agency Marwick Marketing described the same behavior. Its write-up says the change has so far been spotted only on mobile and that UK marketers have replicated it. It also says Google has not announced anything, and it describes the change as “not confirmed as a permanent change yet”.

That is the full set of established facts. We found no official Google statement or documentation describing the feature.

A test, not a rollout

The distinction matters because it should shape your response. A rollout is a settled change that you plan around. A test is something Google may expand, modify or withdraw, and may be showing to only some users, in some places, on some devices.

There is also precedent for this kind of experiment. Search Engine Roundtable reported in June 2024 that Google had shown AI-generated text as the description in some local panels, text that business owners had not written themselves. Schwartz’s September report says Google has been testing variations of this for a while. So the sensible reading is not that local search changed overnight. It is that Google keeps experimenting with letting AI describe a business, and the current version is the most visible so far.

The practical consequence is that some customers will see the classic panel and some may see the AI version. Both need to be right.

What the one detailed example shows

The most detailed public example is Schwartz’s own. He found that most of the AI-written description of his business was correct. One part, a section on what he calls niche solutions, was somewhat out of date. Importantly, he did not blame the AI for that. He noted that the content on his own company site was old, which suggests where the description drew its detail.

That small example carries a useful lesson. An error in an AI description may not be an AI invention. It may be an accurate summary of outdated information you published, or left in place, somewhere. Schwartz’s own reaction was mixed. He wrote that he was unsure he wanted AI describing his business “when I can” do it himself.

For a local business, the lesson is that every public statement about you, on your site, your profile and elsewhere, is a potential input. If a service you stopped offering is still listed, a summary can repeat it.

What a larger snapshot adds, with caution

On September 26, Fisher’s company published a study of how AI Overviews merge with business panels on mobile. It should be read with that context in mind: the author sells local search services, and the study is a single-day snapshot, which the author acknowledges.

The sample was 100 local brands across ten service industries, searched on mobile from a New York anchor point, with four query variations per brand. The study reported that 26% of brands showed an AI Overview on a plain brand-name search on mobile, against 6% on desktop. Of the overviews that appeared, about 80% were merged into the business panel. In four of 68 phone searches, the result surfaced a different, similarly named business. And in a three-day repeat test, one of 20 brands changed layout from one day to the next.

We would not treat those figures as a benchmark. One day, one city and an agency author make the findings suggestive, not conclusive. They support three cautious observations. The experience is not uniform. It appears more common on mobile. And a business can be confused with another of similar name, which is a reason to check your own results rather than assume.

What may feed the description, and what is unproven

The obvious question is which of your information the AI uses. Marwick Marketing says local panels draw on Google Business Profile data and on Google’s Knowledge Graph. Beyond that, advice circulating in the industry gets more speculative.

Some commentators say the services list matters more than website copy, or that review wording shapes the description, or that posts and photos act as freshness signals. Those are interpretations. We found no published test showing which fields drive the text, and one widely cited guide rests on a single reported example with no sample size or method.

So the honest position is this: nobody outside Google can say exactly how the description is built. What you can do is make every plausible source accurate, current and consistent. That costs little, and it is the same work that supports good local search engine optimization whether or not the test continues.

What to check now

Start with what customers see.

Search for your business by name on a phone while logged out of your Google account. Note whether an AI Overview label appears, read the description line by line, and take a dated screenshot. Do the same for a search like your name plus reviews. Results vary, so repeat the check on different days.

Look for errors. List anything wrong, outdated or missing, such as retired services, old locations or wrong hours, then trace each one to its source: your website, your profile, a directory or a review.

Then fix the sources.

Review your Business Profile against Google’s own guidelines. Choose categories carefully: Google tells businesses to use “the fewest number of categories” that describe the core business, and to pick the most specific ones. Make sure the services and products you list are real and current. Keep the business description honest and relevant, and remember that the guidelines do not allow special promotions, prices or links in it.

Check hours, including holiday and seasonal hours, and make sure your address and phone number match exactly across your website, your profile and your main directories. A mismatch gives any summarizer two versions to choose from.

Update your website. If the AI reads your site, an outdated services page or an old about page is a liability. Remove or revise anything that no longer applies.

Look at your reviews. Respond to recent ones, and read how customers describe you. Their language will appear in the public record about your business. Do not ask for scripted reviews or write fake ones.

Test for confusion. If your name is similar to another business, search from a different location or on a different device and see whether the wrong business appears. If it does, keep your name, address and categories distinct and accurate.

Finally, set a routine. Pick a day each month, repeat the checks and keep the dated screenshots in one place.

What not to do

Do not stuff keywords into your business name or description to influence the AI. Google’s guidelines say the name should reflect how customers know the business, and violations can lead to changes in your information or removal of your listing.

Do not assume you can opt out. Marwick Marketing reports that the only option it found was filtering AI results with the Web results toggle, which is a viewer’s choice, not a business’s.

Why this matters beyond Google

This is part of a broader shift in how people find local providers. Generative engine optimization and answer engine optimization are about being described correctly and recommended in AI-generated answers, and the AI Overview in a business panel is one of those answers, attached to your name. The same discipline applies on every platform: clear, accurate, consistent information published in places machines can read.

The Steady Demand snapshot also found that most AI citations in its sample pointed to third-party platforms, such as review sites and forums, rather than official brand sites. As a single-day sample, that is only an indication. But it is a reminder that what other people say about you may shape the answer as much as what you say yourself. Strong local search engine optimization, generative engine optimization and answer engine optimization now overlap, and a profile that is accurate everywhere is the common denominator.

Where this leaves you

The confirmed facts are these. Search Engine Roundtable reported on September 17, 2026 that some local knowledge panels show an AI Overview label, with a Show more button that opens a follow-up chat. The change was reproduced by Schwartz and by UK marketers, appears so far to be mobile only, and has not been announced by Google. It is a test.

The response is a routine, not a rebuild. An AI Visibility Audit can show where your information is wrong across AI platforms. Search for yourself, read what the panel says, trace any error to its source, and correct your profile, your website and your listings. Treat your public information as the raw material for every summary of your business, human or machine.

Frequently Asked Questions

1. What is the AI Overview label in local knowledge panels?
It is a label that some searchers see at the top of a business’s panel, above a short AI-written description. A Show more button opens a chat-style follow-up, similar to AI Mode. Search Engine Roundtable reported it on September 17, 2026.

2. Has Google confirmed this feature?
No. Google has not announced it, and we found no official documentation. It should be treated as a test that may change or disappear.

3. Is it only on mobile?
So far, reports describe mobile. Marwick Marketing says it has only been spotted there, and a September snapshot study by Steady Demand found AI Overviews far more often on mobile than on desktop. That study was a single-day, agency-run sample, so treat it as an indication.

4. Can I opt out of AI descriptions of my business?
We found no business-side opt-out. Marwick Marketing reports that the only option it found is a viewer filtering AI results with the Web results toggle.

5. What should I check first?
Search for your business by name on a phone while logged out, read the description, and screenshot it. Then check your profile categories, services, description, hours and contact details, and your website for outdated statements.

6. Does review wording affect the AI description?
Some commentators say so, but we found no published test that proves it. What you can do is respond to reviews, keep your service information accurate and never write or script fake reviews.

7. How often should I check?
Monthly is a reasonable routine for most local businesses. Keep dated screenshots so you can see what changed and when, and check more often if you make major changes to your services or hours.

If you want to know how your business appears across Google and AI answer engines, and where the information is wrong or missing, that is exactly what we look at in a free AI Visibility Audit.

On August 27, SEO consultant Chris Long spotted something new in live search results and posted it immediately: “Holy moly SEOs: Google is now revealing the AI Mode prompt box BY DEFAULT for some queries.” His reading was blunt. “This is how Google continues to push AI Mode as the default search experience.”

Google confirmed it the next day. In a statement to Search Engine Roundtable, the company said: “For some queries, AI Overviews may dynamically expand for topics where our systems determine it’s most useful for people. Our research has shown that with this dynamic experience, users find Search more helpful and engage deeper in follow-up exploration.” Robby Stein, VP of Product for Google Search, added a detail that matters if you’re mid-read: “If you’ve already started scrolling, the expansion stops to keep your place.”

What actually changed is this. An AI Overview used to load as a short snippet, with a manual “Show More” click needed to see the full answer. For some queries now, it loads fully expanded from the start, with an open “Ask anything” box sitting underneath it, the same interface AI Mode uses. Google has not said which share of queries trigger this, whether that share will grow, or when the behaviour might change again. Stein was explicit that it’s still being tuned: “We’re always testing and tuning these experiences, so we’ll continue to refine based on user feedback.”

A related change, one week earlier

A week before dynamic expansion, on August 19, Google rolled out a related but separate change: generative UI moving from AI Mode into AI Overviews. This is the feature that builds a custom interactive tool for a query instead of describing an answer in text. Google’s own example is searching “ph scale” and getting an interactive visual built on the spot inside the overview, not a paragraph explaining one. Practice quizzes for exam prep, custom document generation, and a homework-helper feature using Lens all shipped alongside it.

These two changes are easy to conflate because they’re pushing the same direction, but they were confirmed on different dates for different reasons, and it’s worth keeping them separate when you’re explaining this to a client. Generative UI changes what the answer looks like. Dynamic expansion changes how much of it you see before doing anything at all.

Neither is really a standalone event either. Both sit inside a longer arc that started at Google’s I/O in May 2026, when AI Overviews and AI Mode formally merged into one integrated search experience running on Gemini 3.5 Flash. Everything since then, including this, has been Google steadily erasing the line between a summary above the results and a conversation instead of the results.

The 48% and the 58% are not the same statistic

Coverage of this topic has been sloppy about one thing, and it’s worth being precise here because it’s an easy stat to misquote. BrightEdge found, in March 2026, that AI Overviews appear on roughly 48% of tracked queries, with the rate far higher in specific verticals: 88% for healthcare, 83% for education, 82% for B2B technology. Separately, Ahrefs found in February 2026 that when an AI Overview appears, the top organic result’s click-through rate falls by 58%. One is a prevalence number. The other is a click-loss number. They come from different studies, measuring different things, and citing them together as one connected trend, the way they often get repeated, overstates what either study actually claims on its own.

What both numbers agree on, independently, is the direction. AI Overviews show up on a large and growing share of searches, and when they do, fewer people click through afterward. Two separate, peer-reviewed studies this year support the second half of that. Pew Research found clicks on 8% of visits where an AI Overview appeared, against 15% where it didn’t, a gap that held even after controlling for query length and individual searcher habits. A randomized field experiment in July found that hiding AI Overviews entirely lifted outbound clicks by roughly 40% per search. A second randomized experiment in August found a smaller but still real 8.8 percentage point lift, and a genuinely interesting null result alongside it: removing the AI Overview didn’t measurably change how satisfied or trusting people felt about their search. The benefit Google says it delivers isn’t showing up in what users report either way.

Why “position” barely means what it used to

Here’s the detail most content advice skips, and it’s the one that should change how you think about your own reporting. Google’s own Search Console documentation states that an AI Overview occupies a single position in search results, and every link inside that overview is assigned that same shared position. There is no internal ranking. You cannot see whether your link sat first or fourth inside the answer, only that the whole block sat wherever it sat. This is exactly why treating an AI Overview citation as a traditional search engine optimization ranking problem misses the point, there is no position left to optimize toward once you’re inside the block.

The impression rule is stricter too. A normal search result counts as an impression once it’s served on the page. A link inside an AI Overview only counts once it’s actually scrolled or expanded into view. Two links from the same site inside one overview count as a single impression on the chart-level view, though they’re counted separately on the page-level table, which means the two numbers in your own Search Console won’t always add up to each other, and that’s not a bug, it’s two different ways of aggregating the same event.

There’s a real attribution problem sitting underneath all of this too. One publisher who tracked this properly, using a custom GA4 dimension that catches the text-fragment links Google sometimes appends to AI Overview citations, found that 22.4% of that traffic was landing in their analytics as Direct, not Organic Search. Nothing in a standard analytics setup tells you this is happening. If you get meaningful AI Overview citations, some real share of that traffic is currently invisible to whoever reports on your channels.

The surface nobody’s talking about

There’s one more wrinkle worth knowing if you’re trying to explain the scale of this to a client. A separate but related surface, People Also Ask, has effectively been taken over by the same shift. Independent tracking in early September found 97% of People Also Ask answers are now AI Overviews, up from just 12.6% a year earlier. A second, independently run dataset checking the old snippet endpoint directly found it had stopped returning results at all from August onward. Whatever you think dynamic expansion means for search results generally, the box that used to sit quietly beneath a query with a handful of related questions has already, almost entirely, become the same AI-generated surface this whole piece is about.

What to actually check on your own site

Pull Search Console and look at click-through rate by query, not just average position. A page holding position one or two with a falling CTR on an informational query is a strong sign an AI Overview is intercepting the click before your listing is ever really seen, whatever the position column says.

Separate your query types honestly. Local, transactional, and highly specific searches remain far harder for an AI Overview to fully resolve on its own, and that’s still where organic clicks convert. Simple factual and broad comparison queries are exactly the ones dynamic expansion is built for, and they’re the ones where you should expect this to get worse before it gets better.

Check whether your content is actually being cited inside the overview for your target terms, separate from whether you rank underneath it. Under dynamic expansion, being the source Google pulls from now matters more than the position of the blue link sitting below a fully expanded answer nobody has to click to read. This is the whole premise behind generative engine optimization and answer engine optimization as disciplines distinct from traditional ranking work, earning the citation itself rather than the position beneath it.

And don’t chase this as if it were a ranking algorithm problem, because it isn’t one. This is an interface change, not a core update. Rewriting content to win back a click that an interface decision is now suppressing by design treats the wrong layer of the problem.

What this means going forward

Google’s Search Console rolled its generative AI performance reports out to every property worldwide on August 31, closing the gap for site owners who previously had no visibility into any of this at all. It’s worth using, even with its limits: impressions only, no clicks, no queries, and the counting quirks above. It’s still the closest thing to a first-party signal you’ll get for how often your content actually surfaces inside these expanding answers.

The broader shift is straightforward to state and harder to act on. A “Show More” click used to be a small piece of friction that gave your listing a chance to be seen before someone gave up and left. That friction is being removed, query by query, and Google has told you plainly that it intends to keep tuning the removal rather than reverse it. Structuring content so it earns the citation inside the answer, not just a ranking beneath it, through deliberate answer engine optimization and generative engine optimization work rather than traditional keyword targeting alone, is no longer a forward-looking recommendation. It’s the baseline for anything you publish from here.

Frequently asked questions

1. Is “dynamic expansion” a new Google algorithm update?
Answer: No. It’s an interface change to how AI Overviews display, confirmed by Google on August 28, not a change to the ranking algorithm. It affects how much of an AI-generated answer a searcher sees automatically, not which pages rank where beneath it.

2. Are the 48% and 58% AI Overview statistics from the same study?
Answer: No, and they measure different things. BrightEdge’s 48% (March 2026) is how often AI Overviews appear across tracked queries. Ahrefs’ 58% (February 2026) is how much the top organic result’s click-through rate falls when an AI Overview appears. They’re often quoted together as if connected, but they come from separate research.

3. Does average position still mean anything if I’m cited inside an AI Overview?
Answer: Less than it used to. Google’s documentation confirms every link inside an AI Overview shares one single position value, with no visibility into internal ordering. Whether you’re cited at all now matters more than where within the overview that citation sits.

4. Why is my AI Overview traffic showing up wrong in my analytics?
Answer: A documented case found 22.4% of one site’s AI Overview traffic was misattributed as Direct rather than Organic Search, due to how citation links are sometimes formatted. Standard analytics setups don’t catch this automatically.

5. What should I actually change first?
Answer: Start with Search Console’s generative AI performance report, live worldwide since August 31, and cross-check it against click-through rate by query in your normal performance report. Prioritise fixing content for the query types AI Overviews answer most fully rather than reacting site-wide.

If you want to know whether your content is actually earning citations inside AI Overviews and AI Mode, or just ranking beneath answers nobody clicks through, that is exactly what we check in a free AI Visibility Audit.