Generative engine optimization · Original data
ChatGPT's local answers don't run on Foursquare. A 4,607-run test found zero citations, and so does our data
If you've been following local SEO advice lately, you've likely seen some variation of this: OpenAI teamed up with Foursquare in January, Foursquare powers ChatGPT's local recommendations, so now you need to manage your Foursquare listing. It pops up in so many posts it might as well be gospel.
A test from August 21 didn't back that up. Ben Fisher ran 2,880 prompts across twelve categories, restaurants, bars, coffee shops, hotels, museums, live-music venues, bakeries, hair salons, plumbers, auto repair, gyms, and one more, in twelve cities, all within a single 72-hour stretch. He added an independent cross-check of another 1,728 prompts on a different API surface, totaling 4,607 successful runs. Foursquare didn't appear once in the citations on the main surface, 0.00%, and turned up in exactly one run out of 1,728 in the cross-check, a 0.06% rate. Yelp, by contrast, showed up in the structured data behind ChatGPT's business cards for 95.83% of runs. Fisher also tested the behind-the-scenes influence another way, comparing which businesses ChatGPT actually named against Google's and Foursquare's own local results. The overlap with Google averaged 7.5%; the overlap with Foursquare was 5.7%, basically background noise, and different enough from what you'd see if Foursquare were quietly steering the answers that he calls the gap statistically real (p = 0.008), not just a rounding blip.
Our own numbers agree, on Foursquare
We pulled our own numbers on this before we saw Fisher's study, for a different reason. Back in August we tested a content brief that assumed Yelp, YouTube, and LinkedIn were the go-to sources AI trusts for local businesses. Across 1,423 citations from ten local-service categories in five cities, Yelp didn't show up once. We didn't check Foursquare then. We checked it now, running the same dataset again: also zero. Two separate tests, done by two different people, on two different sets of prompts, land on the same answer for Foursquare.
Where the two datasets actually split
Where the two datasets split is Yelp, and that split is the real story here. Fisher found Yelp in the grounding behind 95.83% of runs. We found it in zero. Both numbers are true. What's different is the industries behind them. His twelve categories tilt toward restaurants, bars, hotels, and bakeries, exactly the kinds of businesses Yelp was built to cover, alongside a few service categories like plumbers and gyms. Ours are ten local-service and medical categories: med spas, dentists, chiropractors, optometrists, veterinarians, personal injury lawyers, plus plumbers, auto repair, gyms, and hair salons. Yelp's review base has always revolved around food and nightlife. It makes sense that a model grounding its answers in Yelp data would lean on it hard for restaurants and barely touch it for a chiropractor.
We checked what fills that gap in our own data, since we already had it. Reddit accounts for more citations than the next two domains put together. Directory and review-aggregator sites make up about a third of everything cited. None of it is Yelp, and none of it is Foursquare, but it's also not empty space. The sources are just industry-specific ones most local SEO advice doesn't mention, because most local SEO advice is written for a generic local business rather than a plumber or a dentist.
Four of our ten categories, plumbers, auto repair, gyms, and hair salons, overlap directly with Fisher's list, so we checked those specifically instead of leaning on the overall number. Yelp still doesn't lead any of them in our data. Reddit tops the directory-type sources for auto repair, gyms, and hair salons; a service-lead platform called ServiceAgent.ai tops plumbers. That argues against a clean food-versus-everything-else split and points toward something closer to a per-category source list that happens to include Yelp for some businesses and skip it for others, restaurants and bars included in his numbers, plumbers and gyms apparently not, at least in our sample.
What this means for your checklist
The practical takeaway if you're in a local-service or medical field: dropping Foursquare from your to-do list looks safe based on two independent datasets now. Dropping Yelp is a different decision, and it hinges entirely on what you sell. If you run a restaurant or a bar, Fisher's numbers say Yelp still carries real weight in what ChatGPT cites. If you run a med spa or a law practice, our numbers say it doesn't show up at all, and the platforms actually feeding your category's answers are worth checking directly instead of assuming they match a guide written for a different industry.
What we can't say from either dataset is that this splits neatly along a food versus non-food line. Fisher's categories mix hospitality with plumbers, auto repair, and gyms, and we don't have his per-category breakdown to see whether Yelp showed up for those service categories specifically or only carried the restaurants and bars. The two collection windows sit about a week apart in August, his single-surface primary run against our mix of ChatGPT and Perplexity citations, so this is two snapshots pointing the same direction rather than a controlled comparison of the same categories under the same method.
You can check which sources actually feed your own category in about thirty seconds.
Source: Ben Fisher, "ChatGPT's Local Results Aren't Coming From Foursquare (And Probably Never Really Were)," Steady Demand (August 21, 2026): 2,880 primary prompts plus a 1,728-prompt independent cross-check, 4,607 successful runs, 12 categories, 12 metro areas, single 72-hour collection window. Our own figures come from askchat.studio's GEO probe database: 1,423 citations across 10 local-service categories and 5 cities, collected August 26-27, 2026, and our earlier check of the same dataset for Yelp, YouTube, and LinkedIn presence, published August 28, 2026.