In July 2026 Yelp licensed its reviews and business data to OpenAI. For local and service businesses, that single deal has a direct implication: your Yelp presence, and by extension your presence on other licensed review platforms, may now feed straight into ChatGPT's answers about local businesses. Your reviews stopped being just a destination page someone might visit and became training and retrieval input for a system that answers questions on your behalf.
I want to be careful here, because the honest position includes a large caveat: we do not know exactly how OpenAI weights this data. What we can reason about is the shape of the change and what a sensible local AEO response looks like.
Key Takeaways
- Yelp licensed its reviews and business data to OpenAI in July 2026.
- The practical implication: your Yelp profile may now be a direct input into ChatGPT's answers about local businesses.
- Review platforms are becoming retrieval infrastructure, not just destinations people browse.
- Summarisation behaves differently from browsing: a model reads all your reviews at once, where a human reads the first six.
- Recency and volume likely matter differently under summarisation than under human browsing, though the exact weighting is unknown.
- Prioritise the platforms with licensing deals and structured data, starting with your Google Business Profile and Yelp.
- This is part of a broader pattern: licensing deals now shape what models can say about your business.
Your review profile is no longer only a page people visit. It is a source a model reads on your behalf.
What Actually Changed
Yelp holds two decades of structured local data: business categories, hours, locations, attributes, price bands, photos and an enormous body of written reviews. That combination is unusually hard to reconstruct from open web crawling, because much of it is structured, verified and continuously updated.
Licensing it to OpenAI does something specific. It moves that data from "possibly encountered by a crawler, possibly not" into "available as a known, structured source." That is a categorical difference, not a volume difference.
Why licensed data is different from crawled data
- Structure. Crawled reviews are text on a page. Licensed data comes with fields, rating, date, category, location, the model can reason over.
- Completeness. A crawler sees what it happens to fetch. A licensed feed sees everything.
- Reliability. Licensed data does not depend on a robots.txt file, a rate limit, or a rendering quirk.
- Permission. Licensed data can be used without the legal ambiguity currently being litigated around crawled content.
Review Platforms as Retrieval Infrastructure
This is the framing shift I would want a local business owner to take away.
For twenty years, a review platform was a destination. You optimised your profile so that a human who landed there would be persuaded. Success meant a click, a call, a booking.
Now a review platform is increasingly a source layer. A person asks an assistant "best physiotherapist near Indiranagar" and never visits Yelp, Google Maps or your website. The assistant reads the source layer and answers. Your profile influenced the outcome without being seen.
What that changes about optimisation
- The audience is a summariser first, a human second. Both still matter, but the order has flipped for a growing share of queries.
- Consistency across platforms matters more. A model reconciling conflicting hours, addresses or service lists may simply express less confidence about you, or pick a competitor whose data is coherent.
- Categories and attributes carry more weight. Structured fields are the easiest thing for a model to filter on. If you are not correctly categorised, you are not in the candidate set.
- The tail of your reviews matters. A human reads the top six. A model can read all four hundred.
How Summarisation Changes Recency and Volume
This is the part I find most interesting, and where I want to flag uncertainty most loudly.
Under human browsing
Volume is a trust signal, 400 reviews reads as more credible than 12. Recency is a freshness check, a human glances at whether the last review is from this month or 2023. Both are heuristics applied at a glance.
Under model summarisation
The dynamics plausibly differ:
- Volume may matter less as a trust badge and more as evidence density. Twelve detailed, specific reviews may support richer, more confident statements than four hundred one-liners.
- Recency may matter more, because a model asked about your current service quality can weight recent reviews and detect that older ones describe a different business.
- Specificity may matter most of all. A review that says "great place" gives a model nothing to say about you. A review that says "they fixed my laptop screen in two hours on a Sunday" gives it a fact it can repeat.
The honest caveat: we do not know how OpenAI weights any of this. These are reasoned expectations based on how retrieval and summarisation systems generally behave, not documented behaviour. Treat them as hypotheses to test against your own visibility, not as settled tactics.
Specific, recent, detailed reviews give a model something concrete to say. Generic praise gives it nothing.
Which Platforms to Prioritise
You cannot maintain twelve review profiles properly. Here is how I would rank them for a local or service business.
Tier one: non-negotiable
- Google Business Profile. Still the largest source of local intent data and deeply embedded in Google's own AI surfaces. Complete every field, keep hours accurate, post regularly, respond to reviews.
- Yelp. Newly important because of the OpenAI licensing deal. If you have been neglecting it because your Yelp traffic is low, that reasoning is now out of date, traffic is not the point any more.
Tier two: category-dependent
- Vertical platforms that dominate your category: healthcare directories, legal directories, home-services marketplaces, travel platforms. Where a vertical platform is the authoritative source in your category, it is likely to be in the retrieval mix.
- Your own site, with proper LocalBusiness structured data. This is the one source you fully control, and it is the reference point against which other sources get reconciled.
Tier three: maintain, do not invest
Everything else. Keep the listing accurate so it does not contradict your primary sources. Do not run campaigns there.
A Practical Local AEO Checklist
- Reconcile your NAP everywhere. Name, address, phone identical across every platform. Inconsistency is the cheapest own goal in local search and it is worse under summarisation.
- Complete every structured field. Categories, attributes, service areas, hours including holidays, payment methods, accessibility. These are the filters.
- Fix your Yelp profile this month. Photos, categories, description, hours. If it has been dormant, it is now load-bearing.
- Ask for specific reviews. Not "leave us a review" but "if you have a minute, mention what we helped you with." Specificity is what a model can repeat.
- Keep review flow continuous. A steady trickle beats a quarterly burst, if recency weighting works the way it plausibly does.
- Respond to reviews, especially negative ones. Your response is text a model can read, and a good response reframes a bad review.
- Add LocalBusiness schema to your site, including aggregate ratings where you legitimately have them.
- Test your own visibility. Ask ChatGPT the questions your customers ask. See what it says about you and your competitors. Do it monthly.
The Broader Pattern: Licensing Shapes What Models Can Say
The Yelp deal is not an isolated event. It is one instance of a pattern that now defines AI visibility.
Where a platform has licensed its data, the model has rich, structured, permitted access. Where it has not, the model has whatever it crawled, subject to blocks, rate limits and legal risk. That asymmetry means the distribution of what models know about the world is increasingly shaped by commercial agreements rather than by what is publicly available.
For a local business, the operational takeaway is simple: your visibility is partly determined by which platforms you are strong on, because platform strength now translates into model access. Being excellent on a platform that has no licensing relationship is worth less than being good on one that does.
That is uncomfortable, and it is not something you can change. It is something you can allocate effort around.
What I Would Not Conclude
- Not that Yelp is now the most important local platform. Google Business Profile remains dominant.
- Not that we know how the data is weighted. We do not.
- Not that review manipulation will work. Platform detection has improved, licensing agreements come with quality obligations, and the downside is severe.
- Not that your website no longer matters. It is the canonical source against which everything else is checked, and it is the place the conversion actually happens.
FAQ
What did Yelp and OpenAI agree in July 2026?
Yelp licensed its reviews and business data to OpenAI. The practical implication for businesses is that Yelp profile content may now feed into ChatGPT's answers about local businesses.
Does this mean ChatGPT recommends businesses based on Yelp?
It means Yelp data is a licensed input. How heavily it weighs against other sources is not publicly documented, so treat any confident claim about the ranking effect with scepticism.
Should I prioritise Yelp over Google Business Profile?
No. Google Business Profile remains the most important local surface. Yelp has moved up considerably in priority because of the licensing deal, but it is an addition, not a replacement.
Do review counts still matter?
Probably, but differently. Under human browsing, volume is a trust badge. Under model summarisation, evidence density and specificity may matter more than raw count. This is a reasoned expectation, not documented behaviour.
Does review recency matter more now?
Plausibly yes, because a model answering a question about your current service can weight recent reviews and detect that older ones describe a different business. Maintaining a steady flow is a low-risk hedge either way.
What kind of reviews are most useful for local AEO?
Specific ones. A review naming the service, the outcome and the context gives a model concrete facts to repeat. Generic praise gives it nothing usable.
How do I check what AI says about my business?
Ask assistants the questions your customers actually ask, "best X near Y", "who does Z in [area]", and record the answers monthly. Track whether you appear, how you are described, and who is named alongside you.
Does NAP consistency still matter?
More than before. A model reconciling contradictory hours, addresses or service lists may express less confidence about you or default to a competitor whose data is coherent.
What is local AEO?
Answer Engine Optimisation applied to local intent: making sure the sources an AI assistant retrieves from describe your business accurately, completely and in language the model can reuse when answering questions about your category and area.
Will other review platforms sign similar deals?
The pattern of licensing deals across the industry suggests more will. Watch for announcements affecting the platforms that dominate your specific vertical, since those are the ones most likely to be authoritative for your category.
Work With Me
If you run a local or service business and want a clear-eyed read on where your visibility actually comes from, have a look at my work at younusfardeen.com and drop me a message through the contact form. I have 4+ years of marketing experience helping brands grow organically, most visibly at Masai School, and I would rather help you fix the two things that matter than sell you a twelve-platform programme you will not maintain.