Why ChatGPT Shows a Lower Star Rating Than Google

When ChatGPT shows your business a lower star rating than Google does, the card is usually linked to Yelp. In the Cheers Market Baseline we matched 1,624 businesses that appeared on a Yelp-linked ChatGPT map card between September 18 and 22, 2026, and on Google AI Mode in the same city: the ChatGPT cards averaged 4.3 stars against 4.9 on Google AI Mode, and 24% showed under 4.0 stars against 0.25%. ChatGPT's other cards sat within 0.1 star of Google AI Mode's rating 98% of the time.

For a brand with dozens of branches, that means a customer who asks ChatGPT for a plumber can see a branch at 3.9 stars on the map while Google shows it at 4.8, and the gap comes from the branch's Yelp page. The fix is to manage those pages inside Yelp's rules, which rule out the obvious move of asking customers for Yelp reviews. You can see which card a location gets by asking ChatGPT your top question for that city, or by running the AI Visibility Grader for one location.

Last verified: September 23, 2026.

A third of ChatGPT's business cards link to Yelp

When ChatGPT answers a local question it often adds a map with business cards, each carrying a name, a star rating, a review count and a link. Between September 18 and 22, 2026, 34% of the rated cards on those maps linked to a Yelp business page, and 72% of ChatGPT answers with rated cards showed at least one Yelp card. The share was much higher in our 60 US metros than in our 40 Canadian ones: 46.4% of cards against 12.8%.

We only captured ChatGPT's cards reliably from September 18, so every ChatGPT card figure here comes from those five days, which covered about 23 of our 32 weekly questions. The citation figures further down use the full three weeks.

Same business, two ratings

To compare like with like, we matched businesses by name within the same metro. Of the businesses on Yelp-linked ChatGPT cards, 1,624 of 2,305 also appeared on Google AI Mode in the same metro. Their ChatGPT cards averaged 4.3 stars, and Google AI Mode showed the same businesses at 4.9. On ChatGPT, 24% were under 4.0 stars and 43% were at least half a star lower than on Google AI Mode; on Google AI Mode, 0.25% were under 4.0. The review counts sat even further apart, with a median of 31 on the ChatGPT card and 469 on Google AI Mode.

The cards that didn't link to Yelp behaved differently. Across 2,250 matched businesses they sat within 0.1 star of Google AI Mode 98% of the time, and 0.4% were under 4.0 against 0.44% on Google AI Mode. Their review counts matched too: a median of 152 on the ChatGPT card and 157 on Google AI Mode. Our data doesn't label where those cards get their figures; it only shows that they match Google's AI.

When we narrowed the comparison to the same question, city and moment, asked of both engines, the gap was wider: 4.2 stars against 4.9 across 1,083 businesses, with 28.5% of the ChatGPT cards under 4.0.

Why the Yelp number runs lower

The plainest difference is volume. The median Yelp-linked card showed 32 reviews, against 168 for ChatGPT's other cards, and 41% of Yelp-linked cards had fewer than 20 reviews; 28% had fewer than 10. In Canada the median Yelp-linked card showed 3 reviews, and 39.4% of Canadian Yelp-linked cards were under 4.0 stars, against 18.2% in the US.

Small samples swing hard. A business with eight reviews, seven at five stars and one at one star, averages 4.5, and one more one-star review takes it to 4.1. The same extra review on a profile of 469 reviews averaging 4.9 leaves it at 4.9.

Yelp also filters its own reviews. Its recommendation software sets some reviews aside as less reliable, and Yelp says those "not recommended" reviews "do not factor into a business's Yelp star rating." Customers also pick different platforms, so a branch's Yelp reviewers aren't necessarily the same people as its Google reviewers. Our data shows the gap. It can't tell you which of these reasons explains it for a given branch, so open each branch's Yelp page before you assume anything.

The influence is on the card, not in the citations

This is why the problem is easy to miss. Of the ChatGPT answers that showed any business under 4.0 stars on the map, 98% (617 of 630) involved a Yelp-linked card. Yet across 9,434 ChatGPT answers from August 31 to September 20, ChatGPT cited Yelp as a source in 0.6%. It cited a directory, review platform or "best of" list in 84%, led by Expertise.com at 32% and Angi or HomeAdvisor at 29%.

A team that audits ChatGPT by reading its citations will conclude that Yelp doesn't matter. A team that opens the map will find Yelp behind almost every low-rated card. The citation side of that audit, engine by engine, is in Do Google Reviews Affect AI Recommendations?

What OpenAI and Yelp say

OpenAI's help page on searching the web with ChatGPT, which I read on September 23, 2026, says ChatGPT may share a user's location "with trusted third-party providers that help provide more local information, such as nearby business listings." It points to Microsoft's and Shopify's privacy policies for its search providers, says results on iOS and Android may include a map, and mentions Yelp only for restaurant reservations. It doesn't say which provider supplies the ratings on business cards, and I couldn't find an OpenAI page that does.

Yelp is more direct. Announcing its ChatGPT reservations integration on August 10, 2026, Yelp said its existing content integration surfaces "Yelp's trusted reviews, ratings, photos, and business details directly within the chat experience," and that platforms including ChatGPT deliver "Yelp-powered recommendations" and help people request quotes from service providers.

Put that next to our numbers and the practical reading is that a Yelp-linked ChatGPT card shows a branch roughly the way Yelp shows it, and Yelp's picture of a branch is often thinner than Google's.

What to fix inside Yelp's rules

Yelp's rule is blunt: "Don't ask anyone to review your business." It names customers, mailing lists, friends and family, and it adds that its software tries to identify requested reviews and not recommend them. So the work on a Yelp-linked ChatGPT card is page management, not review generation:

• Claim every location's Yelp page. Claiming is free, and it's what lets a branch reply to reviews and update its address, phone number, hours and website.
• Match the page to the branch. The name, category, address or service area, phone and hours should agree with the Google Business Profile and the location page, so the card can't send a caller to the wrong market.
• Reply to recent reviews in public. A calm, specific reply helps the next reader far more than a template does.
• Report reviews that break Yelp's content guidelines, such as fake reviews, with the Report Review option in the business account. Yelp generally leaves honest criticism up.
• Open the ChatGPT card for each market. If it goes to a duplicate page, an old address or a closed location, fix that page first, because that is the rating customers see.

If a card shows the wrong branch or the wrong facts rather than a low rating, the correction workflow is in How Do I Fix Wrong Business Information in ChatGPT? Keep the Google side growing the way Google allows, too, which means asking every customer with a link or QR code, as Google's review guidance describes; how many Google reviews you need for AI recommendations has the benchmark by trade.

What to do this week

Take your ten highest-revenue branches. In the ChatGPT mobile app, ask the question a customer would ask for each branch's trade and city, open the map, and note whether the branch appears, whether its card links to Yelp, and the rating it shows. Put that next to the branch's Google rating. For every branch where the ChatGPT card sits half a star or more below Google, open the Yelp page the card links to and work through the list above. Then leave the reviews alone and run the same check again in a month.

Methodology

Every week, Cheers asks ChatGPT, Gemini, Perplexity and Google AI Mode the same 32 home-services questions in 100 metro areas across the United States and Canada, and records every business each engine names, with its star rating, review count and map position where the engine shows one, and every source it cites. This analysis covers the three weeks from August 31 to September 20, 2026: 34,742 usable answers (9,434 from ChatGPT, 9,435 from Google AI Mode, 9,364 from Gemini and 6,509 from Perplexity) across 60 US and 40 Canadian metros. We excluded empty or cut-off answers, most of them from a Perplexity collection outage. Each question was asked once per city per week, so week-to-week changes include normal answer-to-answer variation. Business names were extracted automatically; a hand check of 40 answers found the extraction caught about three in four of the businesses a reader would count, so counts are lower bounds. Star ratings and review counts are the figures each engine displayed; for ChatGPT and Gemini we captured them from September 18 and September 14 onward. Findings describe associations in this sample, not how any engine ranks businesses.

ChatGPT card figures come from September 18 to 22, 2026. Before September 18 our collector captured ChatGPT's map cards on under 5% of answers, and from then on it captured them on 87% to 99%, so we don't use the earlier card data. A card counts as Yelp-linked when its provider link is a yelp.com or yelp.ca business page, or when it is tagged as coming from Yelp's feed. Matched comparisons pair a ChatGPT card with Google AI Mode cards for the same normalized business name in the same metro, averaging Google AI Mode's rating across its observations from August 31 to September 22, 2026; the same-moment comparison uses both engines' answers to the identical question, city and run. We also checked a random sample of matched Yelp-linked cards by hand against the full answers. Citation shares use every usable ChatGPT answer from August 31 to September 20, 2026, with URLs de-duplicated within each answer.

Sources

• Cheers Market Baseline, August 31 to September 20, 2026, with ChatGPT map cards from September 18 to 22, 2026. Cheers Research methodology. Source of every card, rating and citation figure above.
• Searching the web with ChatGPT. OpenAI Help Center, shown as updated last month when read on September 23, 2026. Source of the location-sharing sentence on business listings, the map on iOS and Android, and Yelp's restaurant-reservation role.
• Yelp Brings Reservations and Waitlist to ChatGPT. Yelp Official Blog, August 10, 2026. Yelp's description of its content integration with ChatGPT.
• Don't Ask for Reviews. Yelp Support Center, undated, read September 23, 2026. Source of the rule against asking for reviews and the filtering of requested reviews.
• Don't Ask For Reviews: Why Yelp Does Not Recommend Solicited Reviews. Yelp Official Blog, January 18, 2017. States that not recommended reviews do not count toward the Yelp star rating.
• What is a claimed business?. Yelp Support Center, read September 23, 2026. Claiming is free and lets the owner reply to reviews and edit the address, phone, hours and website.
• How do I report a review?. Yelp Support Center, read September 23, 2026. The Report Review path for business owners.
• Tips to get more reviews. Google Business Profile Help, undated, read September 23, 2026. Google's guidance that a business can ask customers to leave a review with a Google link or QR code, and its ban on incentives.

Dylan Allen-Arnegård is the CEO & Co-Founder of Cheers, the local search platform for multi-location service businesses.