Gemini mentioned Moody Tongue Brewing Company 144 times. ChatGPT mentioned it 5 times. That difference is not a glitch. It is a window into how AI platforms build fundamentally different versions of the same city.
Out of 182 total Moody Tongue mentions across all platforms
Nearly 30× more than ChatGPT — same restaurant, same city
Bar width = share of platform maximum. Total mentions: 182.
Gemini mentioned Moody Tongue Brewing Company 144 times. ChatGPT mentioned it 5 times.
Let that land. This is the same restaurant. The same menu, the same James Beard-recognized brewing program, the same Michelin star. Same city. Same category of questions — Chicago fine dining, Chicago tasting menus, where to go in Chicago for a special occasion.
Within a structured audit, a 29× gap is hard to dismiss as simple fluctuation. It is a signal that something structural is different about how these platforms understand Moody Tongue — and by extension, Chicago.
Claude mentioned Moody Tongue 23 times. Perplexity mentioned it 10. But Gemini at 144 is the outlier, and ChatGPT at 5 is the absence. The gap between those two figures — 139 mentions — is the story.
The instinct when you see a gap this large is to explain it as noise — a quirk of query phrasing, a statistical fluctuation, something that would smooth out with more data. It is worth resisting that instinct.
We ran structured, repeated queries across all four platforms. The methodology was consistent. The gap is real and persistent within this audit. What this data reflects is not error — it is behavior. Two major AI systems. Radically different answers.
That matters because millions of diners now consult AI assistants to decide where to eat. They are not choosing from an objective list. They are choosing from a platform-filtered version of reality. And if those versions diverge by a factor of 29, the diner who happens to use ChatGPT and the diner who happens to use Gemini are navigating different cities.
The platforms do not produce consistent recommendations about a city — each returns its own version of the dining scene, and sometimes those versions barely resemble each other.
Each AI assistant appears to draw on a different mix of signals — the outputs differ in ways that suggest different sources or different weighting, though we do not have direct access to those inputs. What sources each platform was trained on, which review ecosystems it draws from, what associations formed during training: these are plausible hypotheses, but they are hypotheses. What we can observe is the output.
The output suggests Gemini is far more likely to classify Moody Tongue as a Chicago fine dining destination — its Michelin recognition, its brewery-meets-tasting-menu format, its destination dining credentials — while ChatGPT rarely does so. Gemini consistently includes Moody Tongue. ChatGPT rarely does.
This is not unique to Moody Tongue. Across the Chicago data, we see similar patterns: Bavette's appearing 83 times in Gemini and 0 times in Perplexity; Cariño appearing 74 times in Gemini and not once in Claude; Rose Mary appearing 66 times in Perplexity and 1 time in Claude. Each gap tells the same underlying story. The platforms diverge — and they diverge consistently, not randomly.
Moody Tongue is the most extreme example in this edition's data. But it is not an exception. It is the sharpest version of a pattern that runs through the entire dataset.
Most Michelin-recognized Chicago restaurants are legible to AI systems. Alinea is clearly a fine dining restaurant. Ever is clearly a fine dining restaurant. The category is unambiguous, and AI has absorbed abundant coverage of both.
Moody Tongue is something different. It is a brewery that decided to serve a tasting menu. It is Michelin-recognized at the highest level — a distinction it has held — while still being identifiable as a brewing company. The name contains the word "tongue," which is neither obviously fine dining nor obviously a gastropub. The concept is layered: culinary ambition, beverage program, destination dining, and brewer identity, all at once.
That kind of layered identity can create highly variable behavior across AI systems. One platform may have trained on sources that contextualize Moody Tongue primarily as a fine dining destination — press coverage of its tasting menu, Michelin entries, food media. Another may have trained on sources that contextualize it primarily as a brewery — brewing industry coverage, craft beer media, beverage journalism. Those are not wrong framings. They are different framings, and they produce different recommendation patterns.
A restaurant that sits cleanly inside a recognizable category is easy to recommend consistently. A restaurant that lives at the edge of multiple categories is where platform divergence becomes most visible — each platform produces a classification result, and those results differ significantly for restaurants with layered identities.
The Moody Tongue gap has different implications depending on who you are.
Depending on which AI assistant you open, Moody Tongue may look like one of the defining dining experiences in Chicago — or it may barely appear at all. The platform you use is not a neutral search engine. It is a curated, idiosyncratic version of the city. The gap between those curation choices is 139 mentions wide.
Visibility in AI recommendations is not simply a function of reputation. It depends on whether AI systems understand which category you belong in — and whether they classify you consistently. A restaurant with a hybrid identity may find its AI footprint wildly uneven across platforms, with no clear way to know which version of itself each platform is holding.
The question is not whether Moody Tongue belongs in the conversation about Chicago fine dining. It clearly does. The question is what produced a 29× platform gap for the same restaurant in the same city — and what that pattern says about how platform-specific AI recommendations may be affecting which restaurants diners encounter.