ChatGPT's local map is Foursquare data — claim your listing tonight

Ask ChatGPT for funeral homes near a town and the answer often arrives in two parts: a paragraph, and a small map — pins, names, addresses, hours. The paragraph is assembled from the open web. The map is not. The map runs on a places database licensed from Foursquare, and somewhere in it is an entry for your funeral home that you almost certainly never wrote.

Where the map card actually comes from

On December 16, 2024 — the day ChatGPT’s search rolled out worldwide with maps attached — Foursquare announced it had partnered with OpenAI, in its own words: “Our Places API — with 100M+ POI across 200+ countries — delivers relevant real-world results to user queries.” The claim is still on Foursquare’s business-listings page today, which names OpenAI alongside Apple Maps as apps its data powers. OpenAI, for its part, has never itemized its local stack — analysts were left guessing at launch whether it would build its own listings system or license one — so the confirmation stands on Foursquare’s side of the deal. But it stands.

Keep the two pipes separate in your head, because they take different work. ChatGPT’s cited text answers run on Bing’s web index and the open web — that lane is your website, your pages, the record you publish. The structured map card — the thing a family taps to call you — is the places-data lane, and that lane is Foursquare. You may have read that ChatGPT “falls back to Bing Places,” and it’s a common vendor conflation: the web index and the listings platform are different systems. Fixing your website does not fix a wrong phone number on the map card. Only the places data fixes the places data.

The app died; the data is the product now

If Foursquare rings a bell as the app where people competed to be “mayor” of the coffee shop, you’re remembering correctly — and that app is gone. Foursquare shut down its consumer City Guide app in December 2024 and rebuilt the company around the data those years of check-ins produced: 100 million-plus points of interest across 200-plus countries, built on 16 billion check-ins and a billion photos, tips, and reviews. A month before the OpenAI deal, it open-sourced the core dataset for free commercial use, which means the record it keeps about your funeral home doesn’t just reach ChatGPT — it can propagate to any app or model that ingests the open data. The company stopped being an app and became plumbing. Plumbing is exactly where a funeral home’s details now live or die.

Your listing exists whether you made it or not

Foursquare auto-generates listings from public data, so your home is very likely already in the database — with whatever name, hours, and phone number the public record happened to carry. Unclaimed means nobody at your funeral home has ever checked it. And a wrong entry doesn’t present itself as uncertain: a family asking ChatGPT at midnight gets the stale phone number or the wrong hours rendered with the same confidence as a right one. If the entry is missing outright, you’re absent from the first step of the lookup — not ranked low; not present.

That’s the stake, stated plainly. The family reading the map card is under the worst time pressure of their year. The number on the pin either rings your night line or it doesn’t.

Claiming it — the ten-minute version

Go to business.foursquare.com and search your home’s name and city. The flow, as the major listings platforms document it: find the auto-generated listing (create one only if the search truly comes up empty), claim it, and verify — by phone in most cases, a PIN read to the number on file, near-instant. Claiming and editing are free; the roughly $20 charge Foursquare offers only speeds up verification, and you don’t need it. Foursquare adjusts this flow from time to time, so trust the live portal over any walkthrough, including this one.

Then fill every field, in this order of consequence: exact name, address, and phone — written character-for-character the way they appear on your Google Business Profile and website; primary category “Funeral Home,” plus a secondary if cremation or pre-planning is core to what you do; hours, including your after-hours phone reality; website; a plain description that names your town and service area; photographs of the actual building. Blanks are where a model guesses, and the whole point of claiming is to leave it nothing to guess about. The character-for-character part matters more than it looks: mismatched name-address-phone across Foursquare, Google, Yelp, and Apple is precisely how an AI ends up stating your details wrong — the trust signal machines read first is consistency.

What claiming buys — and what it doesn’t

Honesty about the ceiling: you’ll see it claimed that 60 to 70% of ChatGPT’s top local results come from Foursquare data. That figure traces to one Spanish-market analysis of fifty local prompts — a real test, but a small one, and OpenAI has never published the actual split. Treat the percentage as an estimate and the mechanism as confirmed. What claiming buys is correctness and eligibility: your home present on the card, with the right phone number, the right hours, the right pin. What it does not buy is the paragraph — the prose recommendation above the map is assembled from a different internet entirely, and earning a place in it is the documentation work of publishing answers machines can verify.

Ten minutes, free, once — with a quarterly glance to confirm nothing drifted. Most of what decides how AI describes your funeral home lives on servers you will never touch. This is the rare piece of it with an edit button, and the deadline argument writes itself: the next family to ask ChatGPT for help in your town will read whatever the database holds tonight.