A growing share of local buyers now asks an AI assistant to name a provider instead of scanning a page of results, and very little published data describes how those recommendations behave over time. Spearleaf has released a study of 216 U.S. markets, built from 12 service categories across 18 cities, measuring how ChatGPT's local business panel and Google's AI Overview select businesses and how long those selections last. Captures ran in August 2026 and asked one plain question in every market, the kind a real buyer types.
The central finding is churn. When the identical question was asked again in the same 214 paired markets 4-9 days later, only 48.2 percent of the businesses ChatGPT had displayed the first time were displayed again (469 of 974 displayed businesses, August 2026). The top recommendation held on even less often, with the same business in the first position both times in 34.1 percent of markets (73 of 214). Nothing announced the turnover, and no interface change accompanied it. Asking the same question again was enough to produce a different set of names.
The study measured that instability side by side across surfaces. Google's traditional local pack in the same 216 U.S. markets barely moved over the window, with a wave-to-wave overlap of 0.82 on the Jaccard scale (198 market pairs). ChatGPT's displayed panel scored 0.36 (214 market pairs), and the set of businesses named inside Google's AI Overview scored 0.26 (21 market pairs). The AI Overview itself appeared on 25.9 percent of market captures in the same August 2026 window (216 markets, two query forms, two waves), and when it named businesses, 87.6 percent of those businesses already sat in the same page's local pack (127 markets).
The signals that separated shown businesses from unshown ones were consistent, and the study reports them descriptively rather than causally. Businesses with an address in the asked-about city were far more likely to be displayed, a descriptive odds ratio of 14.4, with an in-city address on 92.0 percent of displayed cards versus 73.9 percent of observed candidates that were not displayed (1,790 of 1,946 versus 3,020 of 4,086). Review depth tracked survival as well. Businesses still shown a week later carried a median of 226 reviews, against 164 for those that dropped, while the median displayed rating was 4.9 for both groups. Across every city tier the median minimum rating on a panel was 4.8, which frames a near-perfect rating as the price of entry and review volume as the separator.
The ratings on those cards did not all come from Google. Across the first wave's 799 resolved rating cards, 17.4 percent displayed Yelp ratings rather than Google ratings (139 of 799). The visual tell proved dependable, since every card carrying a Yelp logo validated against Yelp's own listing (138 of 138). The findings come from the ChatGPT web app's own responses and city-targeted Google SERP data, with every displayed rating validated against same-day Yelp and Google snapshots and a 50-card hand audit in which 49 of 50 cards matched.
"You can't hold a rank in AI search. You can only be worth recommending every time it looks, and be everywhere it looks," said Joshua Albanese, founder of Spearleaf. "The steadiest thing in the data was the boring part. A deep bench of reviews at a near-perfect rating kept getting a business picked back up, wave after wave, and that is the part an owner actually controls."
The full report, every chart, and a methodology FAQ are published at Spearleaf.com, including reading-list data on which sites ChatGPT pulled while forming answers; reviews.birdeye.com was the most retrieved domain, appearing in 37.7 percent of markets (81 of 215). For owners who want the findings turned into a working plan for their market, the Spearleaf team applies the same measurements to individual businesses.
The study window covers one week in August 2026, on one pinned model and one phrasing family, and the findings describe that window rather than all time. Aggregate tables behind each figure are available to researchers and reporters on request, so the numbers can be checked and compared against future waves.
About Spearleaf Spearleaf is a local SEO and AI search optimization agency in Fort Myers, FL, founded by Joshua Albanese. The agency helps local service businesses rank in the Google Map Pack and earn citations from AI assistants including ChatGPT, Perplexity, Gemini, and Claude, through organic content, Google Business Profile optimization, reputation management, web design, and customer reactivation campaigns. Engagements run month to month after an initial term.
###
For more information about Spearleaf, contact the company here:
Spearleaf
Joshua Albanese
(239) 920-7750
joshua@spearleaf.com
11000 Metro Pkwy #13, Fort Myers, FL 33966
