Why ChatGPT Local Recommendations Run Through Foursquare, Not Yelp

When AI visibility advice covers directories, it defaults to the same stack: Google Business Profile, Yelp, BBB. These are the right answers for Perplexity. They're approximately right for Gemini, which draws heavily from Google's knowledge graph.

ChatGPT works differently. For local service recommendations, it doesn't retrieve pages the way a search engine does. It pulls from structured business databases integrated at the API level -- and the dominant provider of that structured data is Foursquare.

The 60-70% finding

In our April 22, 2026 investigation into ChatGPT's local data sourcing (`knowledge/foursquare-chatgpt-signal.md`, session 3), we documented a Yext study of 6.8 million ChatGPT citations that found 60-70% of ChatGPT local business results trace back to the Foursquare Places API. That makes Foursquare the single largest data source for ChatGPT local recommendations -- larger than Yelp, larger than Google Maps, larger than any other individual directory.

This is not a minor signal in a long list. If Foursquare doesn't have accurate, current data on your business, ChatGPT is working from incomplete or outdated information when it generates local recommendations. That gap shows up directly in ChatGPT citation scores.

The practical problem: Foursquare's consumer app is effectively inactive. The company pivoted to B2B data and API services years ago. Most business owners haven't logged into Foursquare since the app was relevant. The profile created back then -- if one was created at all -- likely has outdated contact information, wrong categories, and no service description.

Why this is different from how Perplexity uses directories

The distinction between the API layer and the page retrieval layer matters for understanding why Foursquare and Yelp are not interchangeable.

When Perplexity cites a business, it typically retrieves a page through live web search. It crawls to a directory listing, extracts content from that page, and surfaces a citation. Our June 26, 2026 investigation (`knowledge/platform-divergence-data.md`, session 59 update) confirmed that Perplexity retains approximately 88% of the original prompt words when running retrieval -- it searches for pages that match the query language directly.

ChatGPT's mechanism for local recommendations is structured data lookup, not page crawl. It queries place databases integrated at the API level. Foursquare's Places API is the primary provider of that data. This is why a business can have a well-optimized Yelp page that Perplexity cites consistently and still be absent from ChatGPT local recommendations: the Yelp page is not relevant to the API layer ChatGPT uses for local queries.

This explains the 11% domain overlap finding from the same investigation (Averi 680M citation study, confirmed by two independent studies). The overlap is low not because the platforms prefer different pages, but because they pull from different infrastructure entirely for local results.

The citation hierarchy for local queries

Our June 23, 2026 investigation on Reddit's relevance to local SMB fix plans (`methodology-recs/2026-06-23-reddit-not-local-smb-fix-plan.md`, session 56) documented the Q4 2025 citation hierarchy for local home services queries, combining ChatGPT and Perplexity:

Yelp led with 512,680 citations. BBB had 149,710. Angi had 145,633. Thumbtack, HomeAdvisor, and Nextdoor followed in decreasing order. Reddit did not appear in the top six despite dominating ChatGPT citation share in other categories.

Yelp's lead is explained by the fact that it serves both mechanisms -- Perplexity retrieves Yelp pages directly via web search, and Yelp's structured data feeds into the API layer that ChatGPT queries. It's the directory that works across both platforms simultaneously.

Foursquare doesn't appear in this hierarchy because it's no longer a web destination. Its contribution is upstream: it determines whether ChatGPT has your business in its local data set at all. A business that generates thousands of combined citations via Yelp could still be absent from ChatGPT local recommendations if Foursquare's record for that business is empty or stale. The two problems are independent.

What your Foursquare listing needs

The Foursquare business tools are separate from the consumer app. The fields that matter for what the Places API surfaces downstream:

**Business name** must match Google Business Profile and Yelp exactly. The Places API record becomes one of the structured data points used when AI models aggregate information about your business across sources. Name variants create NAP inconsistency at the API level, not just the page level.

**Address and phone** must be current and matching across all sources.

**Categories** -- Foursquare has its own taxonomy. Select the most specific applicable category. "Plumber" not "Home Services." The parent category is too broad to surface correctly on specific service queries.

**Business description** -- write this for machine extraction: specific services, specific cities, plain language. The description needs to answer "what does this business do and where?" directly, because that's the question the API answers when ChatGPT generates a local recommendation. Marketing language doesn't help here.

**Hours** must be complete and current. Hours appear directly in ChatGPT responses to availability queries and stale hours are a data quality flag.

Where this fits in fix plan priority

The Foursquare signal is specific to ChatGPT. If a Sourcepull audit shows both a low Perplexity score and a low ChatGPT score on local queries, the fix path differs by platform.

Perplexity is a page retrieval problem. Confirm that your directory pages are in Perplexity's source pool (directory presence first, content quality second). Foursquare doesn't affect Perplexity scores.

ChatGPT local is a structured data presence problem. Foursquare presence, combined with a strong Yelp profile that bridges both the API layer and the web retrieval layer, is the correct fix for low ChatGPT scores on local service queries. The broader platform architecture context from our June 9, 2026 investigation (`knowledge/platform-divergence-data.md`, session 42 major update): 49% of ChatGPT citations come from third-party directories and structured databases versus 12% from brand sites. ChatGPT's local citation mechanism is fundamentally database-driven, and Foursquare is the largest single database in that system.

The priority sequence for a business starting from zero: claim and complete the Foursquare listing before spending time on secondary directories like Thumbtack or Nextdoor. The volume of citations those platforms generate is lower, and their data doesn't feed the API layer the way Foursquare does.

Why this didn't make most guides

Foursquare stopped being visible to consumers years ago. SEO practitioners who test directory impact -- people who would notice if a listing change moved a score -- weren't testing Foursquare because consumers weren't on Foursquare. The signal accumulated invisibly while attention stayed elsewhere.

The Yext finding surfaced it because Yext operates at API integration scale. They can observe which data sources populate ChatGPT's local results in ways that standard URL citation tracking can't. A citation tracker can tell you which pages ChatGPT linked to. It can't directly observe which database query populated ChatGPT's structured local results. The 6.8M citation study was able to connect the two.

A Signal Check at sourcepull.ca audits your business across ChatGPT, Perplexity, Claude, and Gemini. For local businesses with weak ChatGPT scores specifically, Foursquare presence is consistently one of the first infrastructure gaps on the list -- and one of the most commonly missing.