When Google's AI Overviews name local businesses they overwhelmingly repeat the local pack sitting on the same results page.

The ChatGPT posts in this series looked at one AI surface. This one crosses the street to Google. In August 2026, alongside the ChatGPT panel study, we captured city-targeted Google SERP data for the same 216 US markets, 12 service types across 18 cities, in two phrasings and two waves, 864 desktop captures in all. The question: when Google's AI Overviews answer a local "best" search, where do the named businesses come from, and how does that surface behave?

The headline is reassuring for anyone who has invested in local SEO. Do AI Overviews affect local SEO? Yes, and mostly by pointing back at it.

The study captured every market twice in two phrasings

Each market was searched two ways: the conversational form, "who is the best {service} in {city}", and the keyword form, "best {service} in {city}". Both were captured in two waves three days apart. For each capture we recorded whether an AI Overview appeared, which businesses it named, what it cited, and what the local pack on the same page showed. The complete tables live on the study page at how ChatGPT picks a local business.

Across the 216 markets, an AI Overview appeared for 25.9% of market observations (CI 22.8% to 29.3%; 864 captures). But that single number is almost misleading, because the trigger rate swung hard on three different axes.

Question phrasing triggers the AI Overview

The conversational form, "who is the best plumber in Tampa", triggered an AI Overview 39.1% of the time. The keyword form, "best plumber in Tampa", triggered one just 12.7% of the time. Same intent, three times the AI coverage, purely from phrasing. As search behavior drifts toward asking full questions, expect the AI layer to sit on a growing share of local searches.

The rate was also restless within the study window, rising from 20.6% in the first wave to 31.2% in the second just three days later. And it varied sharply by industry:

AI Overview trigger rate by vertical (18 markets each): personal injury lawyers 43.1%, real estate agents 43.1%, window cleaning 31.9%, med spas 30.6%, pool service 27.8%, down to dentists 15.3% and therapists 13.9%.

If you are a personal injury lawyer or a real estate agent, the AI layer is already sitting on a large share of your "best" searches. If you are a therapist, it barely shows up yet.

AI Overviews mostly repeat the local pack

Now the core finding. When an AI Overview named local businesses, it named few, about 1.8 per answer on average (maximum 5, across 215 content-loaded AI Overviews). And those few were rarely discoveries: 87.6% of AI-Overview-named businesses were already in the local pack on the very same results page (CI 82.9% to 92.1%, n=127 markets).

The citations complete the picture. Of 1,683 citation links across the captured AI Overviews, 1,182 pointed at google.com itself. The runners-up were yelp.com with 54 and reddit.com with 24. Google's AI layer is, on this evidence, chiefly a summary of Google's own local results with fewer winners per page. It concentrates visibility rather than redistributing it.

The local pack holds steady while AI answers churn

The second big finding is about stability, and it puts every surface in this series on one ladder. Between capture waves, we measured how much each surface's list repeated itself, on a 0-to-1 overlap scale:

Repeat overlap (Jaccard) between waves on the conversational query arm: Google local pack 0.82 (n=198 markets), ChatGPT recommendation panel 0.36 (n=214), AI Overview named set 0.26 (n=21).

The local pack repeated at 0.82. The ChatGPT panel, as covered in the churn post, repeated at 0.36. The AI Overview's named set repeated at just 0.26, the twitchiest surface we measured, on a small sample of 21 markets where it named businesses in both waves.

Read the ladder from the bottom up and the strategy writes itself. The volatile AI surfaces sample from the stable one. A pack position endures between visits, and the AI Overview draws 87.6% of its names from that pack.

GBP work defends the stable surface

So what does an owner actually do about AI Overviews? On this data, nothing new, and that is the good news. The work that earns and holds a local pack position, a complete and accurate Google Business Profile, steady honest reviews, real local relevance, is the same work that populates the AI Overview above it. There is no separate AI Overview playbook worth buying, because the AI Overview is reading Google's own index.

The nuance worth acting on is measurement. Since the conversational phrasing triggers the AI layer three times as often, check your market with full-question searches, not just keyword searches, to see what buyers who ask questions actually see. And because the named set is so volatile, judge yourself on pack presence, the surface that repeats at 0.82, rather than on any single AI Overview appearance.

The questions owners ask most about AI Overviews and local search, answered from the study data.

Ready to hold the surface Google's AI reads from?

The pack is the durable asset and the AI layer samples from it. That makes your Google Business Profile, your review engine, and your local relevance the whole ballgame on this surface, exactly the fundamentals Spearleaf builds for service businesses every day. If you want to know whether your market's AI Overviews are already naming businesses, and whether yours is positioned to be one of them, reach out through our contact page and we will look at it with you.

This article was written by Joshua Albanese, founder of Spearleaf, a local SEO and marketing agency based in Fort Myers, Florida. Joshua has built six businesses from zero using organic search and content, and he leads Spearleaf's SEO, AI search, and local strategy for service businesses across the U.S.