Key takeaways
- Yelp now licenses its 330 million reviews, ratings, and photos directly to OpenAI, but that partnership is brand new (disclosed February 2026, reported in depth by Axios in July 2026), and ChatGPT's actual citation behavior hasn't caught up to the hype yet.
- One large study (Cheers.tech, 9,434 ChatGPT home-services answers across US/Canada metros) found ChatGPT cited Yelp in just 0.6% of answers, while Expertise.com (32.1%) and Angi (29.4%) led by a wide margin.
- A separate four-engine study (Foundation Inc./AirOps, Q4 2025) found Yelp dominating local citations overall, but almost all of that dominance came from Google AI Mode (72.5% share) and Perplexity (62.1% share), not ChatGPT.
- Promptwatch's own data shows reviews make up only 1.7% of all ChatGPT citations generally, but jump to 5.6% when someone names a specific business, meaning reviews matter most for reputation checks, not discovery.
- Google reviews still carry the most weight for the local pack and Maps, and recency now matters more than raw volume: 74% of consumers only care about reviews from the last 90 days.
So does Yelp actually move the needle in ChatGPT?
I'll be honest, this question doesn't have a clean answer right now, and anyone who tells you it does is probably cherry-picking one study. The honest answer is: it depends which AI engine you're asking, which industry you're in, and whether you're measuring citation share across four platforms or isolating ChatGPT specifically.
Here's what actually happened. Yelp disclosed in its February 2026 earnings release that it was licensing its review data to OpenAI. Axios broke the fuller story in July 2026: Yelp CEO Jeremy Stoppelman told the outlet, "If you want to answer local queries, you really need Yelp," and the deal gives ChatGPT access to around 330 million reviews and roughly 8 million business listings, plus ratings, photos, and business attributes. A "Request a Quote" feature is reportedly coming to ChatGPT too, letting users contact local service providers without leaving the chat.
That's a real partnership with real scale. What it is not, yet, is proof that Yelp reviews are driving ChatGPT's local recommendations today.
The data contradicts itself, and that's the real story
Two credible studies looked at this question around the same time and landed in very different places. Worth sitting with both.
Cheers.tech ran 32 home-services questions per week across 100 US and Canadian metros between August 31 and September 20, 2026, generating 9,434 usable ChatGPT answers. Yelp showed up in just 0.6% of them, 52 answers total. Expertise.com led at 32.1%, Angi/HomeAdvisor followed at 29.4%, with HomeStars (Canada only) at 16% and BBB at 14.5%. Of the ChatGPT answers that cited anything, 84% cited at least one directory or review platform, and 44.2% cited the contractor's own website.
Foundation Inc. and AirOps ran a much bigger study, 28 million local-business queries spanning ChatGPT, Gemini, Perplexity, and Google AI Mode in Q4 2025. In that dataset, Yelp received 512,680 citations, more than the next five competitors (BBB, Angi, Thumbtack, HomeAdvisor, Nextdoor) combined. But when you break that down by platform, Yelp held 72.5% of citations within Google AI Mode and 62.1% within Perplexity. The study doesn't isolate what share ChatGPT alone gave Yelp, and given how overwhelming the Google AI Mode and Perplexity numbers are, it's reasonable to assume ChatGPT's piece of that pie was much smaller.
Put those two studies side by side and a pattern emerges. Yelp has been a heavyweight source for Perplexity and Google AI Mode for a while now because both of those engines crawl and surface Yelp's category pages organically. ChatGPT, historically, leaned on directories like Expertise.com, Angi, and BBB instead, likely because of how it sources through Bing's index rather than crawling Yelp directly. The new Yelp-OpenAI data partnership could change that going forward, since it gives ChatGPT a structured data feed instead of relying on whatever Bing happened to index. But the Cheers.tech data was collected after that partnership was already public, and Yelp still barely showed up. Either the data feed hadn't ramped yet, or OpenAI hasn't prioritized it in citations the way the headlines suggest. Treat any confident claim about "Yelp now dominates ChatGPT" with some suspicion until a newer study measures post-rollout behavior specifically.
What Promptwatch's citation data adds to this picture
Promptwatch's citation-type tracking for the week of September 28 to October 4, 2026 gives a useful sanity check: reviews made up just 1.7% of all ChatGPT citations across all query types, trailing well behind product pages (22.4%), landing pages (16.8%), and listicles (13.6%). That's a small slice.
But the number that matters more is what happens on branded prompts, questions that name a specific business. There, review citations jump to 5.6% of citations, more than triple the baseline, and social post citations roughly quadruple too. In other words: ChatGPT doesn't reach for reviews much when someone is discovering businesses in a category, but it reaches for them a lot more when someone is checking out a specific company by name. That lines up with the LinkedIn post referenced in the research above, where ChatGPT reportedly excluded businesses from "best of" recommendations because of negative or insufficient reviews, even when their SEO was strong.
Promptwatch's data also shows ChatGPT has almost stopped citing social platforms generally, just 0.46% of citations across Reddit, YouTube, LinkedIn, Facebook, Instagram, TikTok and X combined, a massive drop from an earlier 5.8% share after Reddit's citation share collapsed from over 5% to near zero in August 2026. Google AI Overviews, by contrast, pulls 12.2% of its citations from social platforms, about 26 times ChatGPT's rate, with YouTube leading. This matters for local businesses because Yelp, TripAdvisor, and similar review sites aren't technically "social media," but they sit in a similar bucket of user-generated content, and the platforms clearly treat that bucket very differently depending on which AI engine you're optimizing for.
Where Google reviews still win
While ChatGPT's reliance on any single review platform is murky, Google reviews haven't lost their grip on the channels that still drive the most local traffic. Whitespark's 2026 Local Search Ranking Factors survey, which polled 47 local SEO practitioners, found that review signals have risen in importance year over year while link signals have fallen, a genuine inversion of how local ranking used to work. Review recency ranked as the 11th most influential factor for local pack and Maps visibility, and for the first time the survey asked experts to weight factors for AI search specifically, where reviews also scored as influential.
Google's own AI layer is expanding fast, too. A Whitespark case study found AI Overviews now appear in 68% of local-business queries versus 39% for the traditional local pack on the same queries, and Google's "Ask Maps" feature, a Gemini-powered conversational layer inside Google Maps, ranks off the Places database and review signals directly.
Here's the part that should change how you think about reviews: trust in reviews as a signal "as much as a personal recommendation" has been dropping for years, from 79% in 2020 to 42% in 2025, ticking back up slightly to 49% in 2026, per BrightLocal's Local Consumer Review Survey. Meanwhile AI as a trust channel is climbing fast, 40% of consumers now trust AI platforms to provide business recommendations, and 42% trust AI as much as reviews. That's a real shift in where the "last mile" of trust-building is happening, even if the underlying review content is still what feeds the AI's answer.
Comparing what each AI engine actually sources from
| AI engine | Primary local data source | Role of reviews | Notes |
|---|---|---|---|
| ChatGPT | Bing index, plus new Yelp partner data feed | Low overall (1.7%), higher on branded queries (5.6%) | Historically cited Expertise.com, Angi, BBB more than Yelp in home-services queries |
| Google AI Overviews / AI Mode | Google's own index and Places database | High, especially via Yelp organically (72.5% share in AI Mode per Foundation Inc.) | Appears in 68% of local-business queries |
| Perplexity | Live web search, review-platform category pages | Very high, review platforms are the most-cited source type for local prompts | Has a formal TripAdvisor partnership |
| Gemini / Ask Maps | Google Places, GBP completeness, reviews | High | Conversational layer inside Google Maps |
What local businesses should actually do
Skip the panic about chasing a specific ChatGPT ranking formula, because there isn't a published one. OpenAI's only public statement on the matter is that it "ranks search results using multiple factors intended to help users find relevant, reliable information," and placement isn't guaranteed. That's it. No weighting breakdown, no minimum star rating, no magic review count.
What you can control:
- Fix your Yelp listing first, even if you've ignored it for years. Business name, phone, hours, categories, and photos all need to be accurate, because that's the exact data OpenAI is now licensing wholesale.
- Don't neglect Bing Places. It relaunched in October 2025 and is reportedly the primary data source both Microsoft Copilot and ChatGPT lean on for local answers through Bing's index. Wrong data there can mean wrong answers in ChatGPT, independent of what Yelp says.
- Keep NAP (name, address, phone) consistent across Google Business Profile, Yelp, Bing Places, and any industry-specific directories. Inconsistent data makes AI systems treat the same business as multiple unreliable entities, which tanks citation odds.
- Respond to negative reviews everywhere, not just on Google. A complaint left unanswered on TripAdvisor or Yelp still gets read by the models behind ChatGPT, AI Overviews, and Perplexity when they're deciding who to recommend.
- Prioritize recency over raw volume. 74% of consumers only weight reviews from the last 90 days, and that recency bias shows up in AI search behavior too.
- Focus on the handful of high-intent, category-specific prompts people actually ask in your market, rather than trying to cover every conceivable query. Foundation Inc.'s research suggests 10-75 recurring queries cover most of the real search volume in a given category.
For businesses that want to see what's actually happening rather than guessing from industry studies, tools built specifically for AI visibility tracking can show you which platforms are citing your business, how often, and from which sources. Promptwatch tracks citations across ChatGPT, Gemini, Perplexity, Claude, Grok, and Google AI Overviews and AI Mode, including Reddit and YouTube citation patterns that most trackers ignore entirely, which matters given how differently each engine weights review platforms.
For local-specific listing management, tools like Birdeye, Synup, or BrightLocal focus on keeping your Yelp, Google Business Profile, and directory data consistent across the board, which is the unglamorous but foundational work underneath all of this.
The honest bottom line
Google reviews still run the show for the channels that drive the most volume today, the local pack, Maps, and increasingly AI Overviews and Ask Maps. Yelp just bought itself a seat at ChatGPT's table, but the data so far suggests that seat isn't doing much yet, at least for ChatGPT specifically, compared to how dominant Yelp already is inside Perplexity and Google AI Mode. If you run a local business, the practical move isn't to pick a side. It's to make sure the information sitting on Yelp, Google Business Profile, Bing Places, and the directories relevant to your industry all say the same accurate thing, because none of these AI systems are publishing a ranking formula you can game. They're just reading whatever's out there and trying to pick the business that looks most trustworthy in the moment someone asks. If you want to dig deeper into tools built for this exact problem, the GEO software directory at bestgeosoftware.com is a good place to compare platforms side by side.