When buyers search local, Google's AI backs off

01 · The headline number

Local intent is the best AI Overview insulation there is

We split every query in the dataset into general intent ("emergency plumber", "family lawyer", "best accounting software") and local intent (anything carrying a suburb, a city, or "near me"). Then we measured the AI Overview trigger rate for each, separately, across all 18 industries.

The result is one of the cleanest signals in the entire study. General queries trigger an AIO 40.1% of the time. Local queries trigger one just 24.7% of the time, a 15.4-point, or 38% relative, collapse. And it isn't an average hiding a messy split: local intent suppressed AIO in 16 of the 18 industries we measured. The moment a searcher signals "I want a business near me", Google's AI tends to get out of the way and hand the result back to the map, the reviews, and the organic listings.

For local businesses this reframes the whole AI-search panic. The headlines about AI Overviews eating organic clicks are real, but they're describing the informational web, the "how does", "what is", "best way to" queries. The local-service web, the queries that actually end in a phone call or a booking, is the most insulated surface left. The trick is knowing exactly how insulated your category is, and what replaced the 3-pack while you weren't looking.

02 · The local penalty by industry

Every service vertical is insulated. Two e-commerce-shaped ones aren't.

The size of the local-intent discount varies enormously by category. Below is every industry in the study, ranked by how much AI Overviews retreat when location intent is added. A negative gap (blue) means local queries are safer, the AI fires less. A positive gap (green) means the opposite.

The high-consideration service categories are the most insulated. Real Estate (−37.5pp), IT & Tech Services (−36.0pp) and B2B Services (−33.6pp) see AI Overviews more than halve when a location is attached. Google appears to read "buyer's agent Bondi" or "managed IT Parramatta" as a find-a-provider decision, not a summarisable question, and defers to the local results.

Even the categories with the smallest discount still move the right way. Legal (−10.1pp) and Healthcare (−10.0pp) retain high AIO rates even locally, because so much of their query mix is genuinely informational ("is a will valid without a witness", "symptoms of…"). But the direction holds: adding a location never increases exposure for a service business.

The two exceptions prove the rule. E-commerce is the only category that inverts hard (+17.6pp), because "buy [product] online" is a shopping query Google's AI loves to summarise, and there's no local provider to defer to. Hospitality (+1.3pp) is essentially flat, and unusual for a different reason, its general AIO rate is already so low (8.6%) that there's almost nothing left to suppress. If you sell a product or a meal, local intent won't save you. If you sell a local service, it's your strongest card.

03 · AI vs the local pack

AI Overviews and local widgets compete for the same slot

When Google does show a local-business widget, an AI Overview is markedly less likely to appear with it, and vice versa. Across all 116,918 SERPs, a local widget appeared on 17.4% of pages where there was no AIO, but only 10.3% of pages where an AIO fired. They're fighting over the same prime real estate above the organic results, and Google mostly picks one.

Split the whole dataset four ways by what actually shows up above the organic results, and the competition becomes obvious, the two surfaces almost never appear together:

Only 3.9% of Australian SERPs show an AI Overview and a local widget at the same time. More than half (51.4%) show neither, that's the clean organic battleground where classic SEO still wins the click outright. A third (33.9%) are AIO-dominated. And one in ten (10.8%) are local-widget territory with no AI in sight.

The practical read: for a local-service business, the AIO is not usually your problem, the entity widget is. When Google decides a query is local, it tends to answer it with a reviews block or a knowledge panel rather than an AI summary. Winning local search in 2026 is less about out-writing the AI and more about owning the entity surfaces that show up in that 10.8%-plus-3.9% slice. Which raises the obvious question: what are those surfaces now?

04 · The vanished 3-pack

The three-business map pack you optimised for appeared zero times

For fifteen years, the three-business map widget at the top of a local SERP was the object of local SEO. In our snapshot, the discrete "Local Pack" widget, as the data source labels it, appeared zero times. Not once.

That's a labelling story, not an absence story, and the distinction matters. Widen the lens to every local-business surface Google now uses, knowledge-graph entity panels, Google Reviews widgets, hotel packs, local services blocks, top-sights carousels, and they appear on 17,210 SERPs (14.7% of the dataset). Local presentation hasn't disappeared. It has fragmented. Google Reviews blocks (13,556 SERPs) and knowledge-graph cards (15,181 SERPs) now carry most of the load the 3-pack used to.

Why this matters more than it sounds: if your local SEO strategy is still framed as "rank in the 3-pack", you're optimising for a target that no longer reliably renders as a discrete widget. Searching "plumber Sydney" doesn't dependably surface a labelled three-business box any more, it surfaces a reviews block, an entity card, sometimes a services unit, and increasingly an AI Overview. The win condition expanded from one widget to a portfolio of entity surfaces: Google Business Profile completeness, review velocity, and knowledge-graph eligibility through structured data and consistent brand mentions across the web.

The target moved; the goal didn't. Local visibility is still won above the organic results, it's just spread across at least five named feature types instead of converging on a single map box. The businesses that win local search now are the ones tracking, and feeding, all of them.

05 · Geography barely matters

It doesn't matter which city your customers search from

We ran the dataset across six Australian markets, three metro capitals and three regional centres, to see whether AI exposure changes by location. It barely does. Averaged across all 18 industries, every city lands within 1.5 percentage points of every other.

Group those six markets into metro versus regional and the two are statistically indistinguishable, regional Australia actually sees a hair more AI exposure than the capitals:

The practical implication: there's no "regional discount" on AI search. A trades business in Newcastle or a clinic on the Sunshine Coast faces the same AI exposure as one in Sydney, and the same fragmented local-widget landscape. You can't outrun AI Overviews by operating outside the capitals, and you don't need a different playbook for regional markets. The local-intent insulation from Section 1 is what protects you, and it works the same everywhere in the country.

06 · What to do about it

Five moves for local businesses in the AI search era

1. Build a real page for every suburb and service you cover, and lean into location intent. Local queries trigger AI Overviews 38% less than general ones, the strongest insulation in the study. That makes "[service] [suburb]" pages your most defensible organic surface. Don't settle for one thin paragraph per area: build genuinely useful pages with local pricing, recent jobs, service-area maps and FAQs for every postcode you actually work in. This is the half of search the AI mostly leaves alone.
2. Stop optimising for "the 3-pack", it doesn't render any more. The discrete three-business map widget appeared zero times in 116,918 SERPs. Local presentation has fragmented across knowledge-graph cards, Google Reviews blocks, local-services units and entity panels. Reframe the goal from "rank in the pack" to "own the entity surfaces": a complete, active Google Business Profile, structured data on every location page, and consistent NAP and brand mentions that make you knowledge-graph eligible.
3. Treat reviews as a ranking surface, not a vanity metric. Google Reviews blocks were the single most common local widget in the dataset (13,556 SERPs). When Google chooses to answer a local query, it reaches for reviews more than almost anything else. A steady velocity of recent, responded-to reviews now directly influences whether you appear in the surface that's replacing the 3-pack, not just your perceived trust.
4. Win the clean organic SERP, it's bigger than you think. 51.4% of SERPs show neither an AI Overview nor a local widget. On more than half of all searches the old rules still apply: the business with the best page and the strongest signals takes the click outright. AI panic shouldn't pull your whole budget toward chasing the AI. The biggest single slice of local search is still won the traditional way.
5. Run the same playbook everywhere, metro and regional alike. AI exposure varies by less than 1.5 points across six markets, and regional centres see marginally more of it, not less. There's no safe harbour outside the capitals and no need for a separate regional strategy. Audit your category's specific local-intent discount, build the location pages, own the entity surfaces, and apply it consistently across every area you serve.