Yelp is now inside ChatGPT. Most of you can relax
The news in one line: Yelp licensed its reviews, ratings, photos, and business details to OpenAI. ChatGPT will surface them in local answers with Yelp branding, and Request a Quote is coming, so a lead can start and finish inside a chat window.
Here’s my read.
This is a patch for ChatGPT, not a power move for Yelp
Local has been ChatGPT’s weakest surface. It invents restaurants and gets hours wrong. Licensed real-time data fixes that. OpenAI needed this more than Yelp did, which tells you who had leverage in the room.
For Yelp, the trade is simple: get paid to sit inside the answer instead of being summarized for free. The bet underneath it is shakier. Branded links inside AI answers are supposed to send value back to Yelp. Publishers heard the same promise from Google’s featured snippets & AI Overview, and Google kept most of the clicks. Yelp is wagering it plays out differently in chat because an AI answer cites two or three sources instead of ten blue links. Maybe. Or maybe it just teaches users they never need to open Yelp again.
Why you shouldn’t reprioritize anything yet
The deal is not exclusive. ChatGPT keeps reading Google Business Profiles, websites, and everything else. Yelp is one more voice, not the voice.
Early testers are seeing Yelp content mostly in big US metros where Yelp already has review density. So the practical rule is boring: if Yelp matters in your market, you already have a profile and reviews there, carry on. If it doesn’t, this deal won’t make it matter.
Anyone telling you to drop everything and build a Yelp strategy this week is selling something.
The part that actually matters
LLMs have been chewing on third-party reviews for a while, so that’s not the news. What changed is the plumbing. Scraped review data is stale, partial, and legally shaky. Licensed data is real-time, structured, and blessed. That means ChatGPT can now lean on Yelp content harder and more often, cite it by name, and eventually transact on top of it. The difference between “the model probably saw your reviews in training” and “the model queries your review feed at answer time” is the difference between background noise and a ranking system.
And once there’s a licensed pipe, there’s a feedback loop worth optimizing. Fresh reviews, response rates, and structured profile data now plausibly influence what gets pulled into an answer today, not what got baked into a model eight months ago.
Three things worth doing:
Keep your review engine running, but know the rules per platform. Google lets you ask. Yelp explicitly forbids soliciting reviews and filters the ones it suspects were prompted. Which creates an awkward setup: this deal just raised the value of Yelp reviews, and Yelp’s own policy blocks the main way businesses generate them. On Yelp the only compliant play is passive. Make the profile visible, respond to everything, and let volume come slowly. That policy is also why Yelp density only exists in big metros, and why this deal’s impact will stay geographically lumpy.
Ask ChatGPT the questions your customers would ask and check what comes back. Wrong hours and stale complaints in an AI answer are the new bad citation.
Watch Request a Quote. A lead born inside ChatGPT looks like nothing in your analytics. Whoever solves attribution for that channel first gets a real head start.
Models reading your reviews is old news. Models querying them live, with a contract behind it, is the new thing. Plan for the second one.
