What Data Does ChatGPT Use to Recommend Local Businesses? (And Why Foursquare Controls Most of It)

Most business owners assume that if they rank on Google, they appear in ChatGPT. The data says otherwise. Here is what actually powers AI local recommendations, and why the gap between Google visibility and AI visibility is wider than most people realize.

When a customer asks ChatGPT "best Italian restaurant near me" or "find a property manager in Amsterdam," ChatGPT does not search Google. It does not check your Google Business Profile. It does not look at your star rating on Maps. It pulls from a completely different data infrastructure, and the company most responsible for that infrastructure is not one you would expect: it is Foursquare.

Understanding where AI platforms get their local data is the first step to appearing in AI-generated recommendations. And right now, most businesses are operating on entirely wrong assumptions about how this works.

The Numbers That Should Change How You Think About AI Search

Consumer adoption of AI for local business discovery has moved faster than almost any technology shift in recent memory. In 2025, 6% of consumers used AI assistants to find local business recommendations. By early 2026, that number is 45%.

That is not a gradual adoption curve. That is a near-vertical jump in one year.

At the same time, only 1.2% of local businesses actually appear in AI-generated recommendations. The gap between the number of consumers using AI to find businesses and the number of businesses visible to those consumers is enormous, and it is growing.

If your business is not structured for AI visibility, you are invisible to nearly half your potential customers at the exact moment they are looking for what you offer.

Where ChatGPT Actually Gets Its Local Data

Independent research analyzing how major AI platforms source local business information produced results that surprised most marketers and developers.

Foursquare is the primary data source for ChatGPT local recommendations, appearing in roughly 70% of local recommendation queries analyzed. This is not a partnership that gets announced in press releases. It is a data pipeline relationship that most businesses are completely unaware of.

Yelp is a secondary source, appearing in about 33% of searches across AI platforms. Perplexity, notably, uses Yelp as its primary local data source in every single industry category studied.

Google Business Profile data does not flow directly into ChatGPT recommendations the way most people assume. Your Google ranking and your AI visibility operate largely independently.

The Foursquare Situation in 2025-2026

If you are not familiar with Foursquare's recent history, here is the short version.

Foursquare was originally a consumer check-in app that became a massive database of place data. In 2025, the company shut down its consumer-facing city guide to focus entirely on B2B location intelligence. It now sells foot traffic analytics, place data, and location APIs to enterprises, mapping platforms, and AI companies.

The consumer shutdown did not affect the data pipeline. Foursquare's place database, built from years of check-ins, venue data, and business listings, continues to be one of the most comprehensive sources of real-world place information available. AI companies use it because it is accurate, structured, and regularly updated.

Foursquare has also partnered with Reprompt, a startup that uses AI agents to scan the web for real-time updates, enriching Foursquare listings with current data and providing transparent provenance for each data point. The pipeline is actively maintained, not static.

This matters for your business because if your Foursquare listing is incomplete, inaccurate, or missing, you are invisible to 70% of the data pipeline that ChatGPT uses to generate local recommendations.

Why Google Rankings Don't Predict AI Visibility

This is the assumption that trips up most businesses and marketers: if I rank on page one of Google, I must be visible in AI search.

Research shows the opposite is often true. Roughly 89% of ChatGPT citations come from websites ranked at position 21 or lower on Google. The overlap between traditional local search ranking and AI recommendation selection is approximately 45%, meaning more than half of AI recommendations are businesses that would not appear in a standard Google local pack result.

There are structural reasons for this. Traditional local SEO weights proximity heavily. If you are 300 meters from someone's location, you rank higher than a business 3 kilometers away, all else being equal. AI platforms do not work this way. They weight data quality, entity consistency, and structured completeness over physical proximity.

A business with a well-structured listing, accurate geo data, consistent NAP information across directories, and rich schema markup on its website can appear in AI recommendations ahead of a nearby competitor whose data is thin, inconsistent, or missing from key sources.

This is a fundamental shift in how discovery works. The rules have changed, and most businesses have not updated their approach.

The Three Data Layers AI Platforms Use

Understanding the full picture means understanding that AI local recommendations are not built from a single source. They layer multiple signals:

Layer 1: Third-party place databases (Foursquare, Yelp)

This is the primary layer and the one most businesses neglect. Foursquare and Yelp act as pre-curated, structured repositories of business information. AI models trust these sources because they have been maintained and validated over years.

What Foursquare evaluates for completeness:

• Business name, address, phone number (exact and consistent)
• Hours of operation (updated for holidays and seasonal changes)
• Business category (specific, not generic)
• Photos (minimum 10, recent)
• Description (150+ words, specific to the location)
• Price range and attributes

An incomplete listing on Foursquare is not just a missed opportunity. It actively reduces your likelihood of appearing in ChatGPT recommendations because incomplete entries are weighted lower in the data pipeline.

Layer 2: Structured data on your own website

AI crawlers read your website looking for machine-readable signals. JSON-LD schema is the most effective format. A LocalBusiness schema with complete fields, including GeoCoordinates, tells AI systems exactly what your business is, where it is, and what it offers without requiring the AI to interpret unstructured text.

{
  "@context": "https://schema.org",
  "@type": "LocalBusiness",
  "name": "Your Business Name",
  "address": {
    "@type": "PostalAddress",
    "streetAddress": "123 Example Street",
    "addressLocality": "Amsterdam",
    "postalCode": "1000 AA",
    "addressCountry": "NL"
  },
  "geo": {
    "@type": "GeoCoordinates",
    "latitude": 52.3676,
    "longitude": 4.9041
  },
  "telephone": "+31-20-555-0100",
  "openingHours": "Mo-Fr 09:00-18:00",
  "url": "https://yourbusiness.com"
}

The geo field with precise coordinates is one of the most consistently underused signals in local AEO. AI platforms extract GeoCoordinates specifically because address strings are ambiguous and require parsing. Coordinates are unambiguous.

Layer 3: Contextual geo signals

This is the layer most guides skip entirely. Beyond your own listing data, AI platforms evaluate geographic context: what is near your business, what transit options exist, what neighborhood or district you are in, what service area you cover.

A vacation rental listing that includes nearby landmarks, walkability context, and public transit distances is significantly more likely to appear in AI recommendations than a listing that contains only address and nightly rate. This is not because the extra information is marketing copy. It is because AI systems use geographic context to verify and enrich the entity they are evaluating.

For businesses using mapping APIs, this context can be generated systematically: nearby points of interest, isochrone-derived service areas, transit proximity from routing data. The geo data layer is not a manual writing task. It is an infrastructure task.

The Platform-by-Platform Breakdown

Different AI platforms weight these layers differently:

ChatGPT Primary source: Foursquare (70% of local recommendations) Secondary: Yelp, business websites, web crawl Key action: Complete Foursquare for Business listing

Perplexity Primary source: Yelp (used in every industry category) Secondary: Direct web crawl, structured data on business sites Key action: Complete and accurate Yelp listing, plus JSON-LD on website

Google AI Overviews Primary source: Google's own index (Google Business Profile, structured data, reviews) Appears in 68% of local searches; 92% of informational local queries Key action: Google Business Profile optimization, LocalBusiness schema, FAQ schema

Gemini Similar to Google AI Overviews but has a much lower local recommendation rate (11% vs ChatGPT's 1.2%) Key action: Same as Google AI Overviews

The practical implication: no single platform dominates, and optimizing for one at the expense of others leaves you exposed. A business that has only worked on Google Business Profile will appear in Google AI Overviews but be invisible to the 31% of consumers using ChatGPT for local recommendations.

What Entity Consistency Actually Means

The phrase "NAP consistency" (Name, Address, Phone) appears in almost every local SEO guide, but it deserves more specific treatment in the context of AI search.

Research on AI recommendation selection shows that businesses with consistent entity data across 60 or more directories grow AI search visibility 74% faster than businesses managing listings manually or inconsistently. This is not because 60 directories is a magic number. It is because AI systems use cross-platform consistency as a signal of entity legitimacy.

When ChatGPT encounters your business name in multiple data sources, all with the same address, phone number, and business category, it treats that entity as verified. When it finds inconsistencies (old phone numbers, multiple address formats, slightly different business names), it weights the entity as less reliable and surfaces it less often.

The specific inconsistencies that cause the most damage:

• Business name variations (abbreviations, punctuation differences, missing LLC or Ltd)
• Address formatting differences (St vs Street, apartment vs unit designations)
• Old phone numbers that still appear on directories
• Outdated hours that contradict your Google listing
• Category mismatches between platforms

For location-heavy businesses like property portals, hotel groups, or multi-location food and beverage companies, this consistency problem multiplies across every location in your inventory.

The Gap Most Businesses Are Missing

Standard AEO guides cover domain authority, schema markup, and NAP consistency. These are necessary but not sufficient for location-based businesses.

The gap is geo data: precise coordinates, nearby context, transit information, service area boundaries. This is the layer that AI systems use to understand the physical reality of a location, and it is the layer that almost no existing SEO or AEO guide addresses.

For businesses built on mapping APIs, this layer is a byproduct of normal integration. Geocoding APIs produce coordinates. Routing APIs produce travel time data. Points-of-interest APIs produce nearby context. The data is there. The question is whether it is being structured and exposed in a way that AI crawlers can use.

For businesses that do not have this infrastructure, the gap is real and growing. As AI becomes the primary discovery channel for location queries, the businesses with the best geo data layer will accumulate a compounding advantage over those without it.

What to Check First

Before taking on a full AEO overhaul, start with a diagnostic of where you currently stand:

1. Search for your business in ChatGPT using the exact queries your customers would use
2. Do the same in Perplexity
3. Check your Foursquare listing is claimed and complete
4. Check your Yelp listing for accuracy
5. Validate your LocalBusiness schema using Google's Rich Results Test
6. Confirm your GeoCoordinates are present and accurate

If you want a structured audit of where your location pages stand against AI visibility criteria, the MapAtlas AEO Checker runs through the specific signals AI platforms evaluate, including the geo data layer that most tools miss.

Frequently Asked Questions

Does ranking on Google guarantee appearing in ChatGPT recommendations?

No. Research shows that 89% of ChatGPT citations come from websites ranked at position 21 or lower on Google. Traditional Google SEO and AI search visibility are largely independent systems with only about 45% overlap.

What data does Perplexity use for local business recommendations?

Perplexity uses Yelp as its primary source for local recommendations, appearing in every industry category studied. It also crawls the web directly and weights structured data on business websites.

Why did Foursquare shut down its consumer app?

Foursquare shut down its consumer-facing city guide in 2025 to focus entirely on its B2B data business. The location intelligence database remains active and continues to power AI recommendation systems through enterprise data partnerships.

How do I get my business into ChatGPT recommendations?

Claim and fully complete your Foursquare for Business listing, ensure NAP consistency across major directories, add LocalBusiness JSON-LD schema with GeoCoordinates to your website, and make sure your Yelp listing is accurate. Structured geo data, including nearby context and service area information, also improves your position in the AI data layer.

Does my business need to be on Foursquare to appear in ChatGPT?

Not exclusively, but it helps significantly. Foursquare accounts for roughly 70% of ChatGPT local recommendation data. A complete, consistent presence across Foursquare, Yelp, and your own website's structured data gives you the best coverage across all major AI platforms.

Conclusion

The assumption that Google ranking translates to AI visibility is the single most common and most costly mistake businesses make as AI becomes a primary discovery channel. ChatGPT pulls from Foursquare, not Google. Perplexity pulls from Yelp. AI Overviews pull from Google's own index with heavy weight on structured data.

The businesses that will win AI local discovery are not necessarily the ones spending the most on SEO. They are the ones that understand the data pipeline, maintain complete and consistent listings across the right platforms, and structure their location data in the formats AI systems can parse and trust.

That is an infrastructure problem as much as a marketing problem. And for businesses operating location-heavy platforms, it is one of the highest-leverage investments available right now.

Related reading:

• Why your hotel is invisible on ChatGPT
• NAP consistency and AI search: why it matters more than you think
• JSON-LD schema for local business AI citations
• The complete AEO guide for local businesses
• Check your AI search visibility for free