Updated data — who leads, who has platform gaps, and what Gary Danko's cluster authority reveals
Key Findings
Overall Findings
The updated dataset shows a more balanced top tier than previous SF audits. Claude underperformance is the dominant structural pattern below the top five — and the Stars Restaurant exclusion (573 raw mentions, closed 1997) is the clearest example of AI training data lag in any fine dining audit conducted to date.
| Restaurant | Claude | ChatGPT | Gemini | Perplexity | Total |
|---|---|---|---|---|---|
| Quince | 170 | 156 | 165 | 136 | 627 |
| Gary Danko | 99 | 69 | 187 | 168 | 523 |
| Benu | 120 | 131 | 118 | 107 | 476 |
| Atelier Crenn | 142 | 67 | 130 | 119 | 458 |
| Saison | 69 | 120 | 125 | 83 | 397 |
| Lazy Bear | 70 | 32 | 120 | 104 | 326 |
| Acquerello | 60 | 88 | 98 | 38 | 284 |
| Kokkari | 10 | 25 | 116 | 102 | 253 |
| Californios | 35 | 43 | 90 | 59 | 227 |
| Rich Table | 31 | 18 | 64 | 100 | 213 |
| Birdsong | 33 | 24 | 58 | 78 | 193 |
| Seven Adams | 0 | 14 | 128 | 43 | 185 |
| State Bird Provisions | 47 | 25 | 54 | 47 | 173 |
| Wayfare Tavern | 42 | 6 | 46 | 62 | 156 |
| The Progress | 29 | 10 | 55 | 49 | 143 |
Platform Concentration Gaps
Claude underperformance is the most consistent structural pattern in this audit. Every restaurant below shows meaningful visibility on Gemini and Perplexity with near-zero Claude presence — a content indexing gap, not a reputation gap.
| Restaurant | Other Platform Mentions | Gap Platform | Gap Mentions |
|---|---|---|---|
| Seven Adams | 185 | Claude | 0 |
| Kiln | 116 | Claude | 0 |
| Boulevard | 100 | Claude | 0 |
| Cotogna | 98 | ChatGPT | 0 |
| Ernest | 97 | Claude | 0 |
| Empress by Boon | 67 | Claude | 0 |
| Dalida | 64 | Claude | 0 |
| Angler | 60 | ChatGPT | 0 |
| Nisei | 39 | Claude | 0 |
| Mister Jiu's | 36 | Perplexity | 0 |
Cluster Analysis
SF fine dining prompts do not return a single consistent restaurant set. Five clusters reveal meaningfully different competitive landscapes — with Gary Danko's three-cluster leadership being the most structurally significant finding in this dataset.
Quince leads at 208 — its Michelin three-star status and the depth of its documented dining experience give AI systems specific, attributable claims to retrieve on destination dining queries. Benu follows closely at 204, reflecting Corey Lee's sustained critical documentation. The top five here represent the tightest competitive cluster in the audit — all within 75 mentions of the leader, with genuinely distributed platform presence.
Gary Danko leads at 162 — its sustained positioning as SF's most reliable special occasion destination has been absorbed by AI systems across decades of editorial. Acquerello at 120 overperforms relative to its overall rank, reflecting how specifically its intimate dining identity maps to celebration queries. This is the cluster where occasion-specific content — naming the context, the experience, the emotional weight of the meal — pays the largest dividends.
Gary Danko leads at 150. Kokkari at 148 is the cluster's defining over-performance: its 10 Claude mentions elsewhere do not suppress its private dining visibility because its event content is well-indexed on the platforms that dominate this cluster. Wayfare Tavern and Epic Steak surface through Financial District proximity and documented corporate dining content — not Michelin status.
Gary Danko's cluster leadership here at 139 is the most counterintuitive finding in the data. The restaurant is not primarily positioned as a California cuisine or chef-identity destination in the contemporary sense — yet its career documentation across decades of SF dining coverage is deep enough to surface on ingredient and chef-identity queries. State Bird Provisions at 92 reflects Stuart Brioza's documented philosophy around California ingredients and counter-service format innovation.
Rich Table leads at just 40 mentions — the lowest cluster-leader total in any fine dining audit in this series. The restaurants surfacing here do so because their neighborhood identity has been explicitly documented in place-specific editorial content: Hayes Valley, the Western Addition, the Richmond. For any SF restaurant with genuine community roots and a neighborhood story, this cluster represents the clearest and most achievable visibility opportunity in the market. The content required is not national press or a celebrity chef. It is a documented neighborhood identity published in formats AI systems can retrieve at the point of a local discovery query.
What Drives AI Visibility in San Francisco Fine Dining
Visibility is not determined by Michelin star count or recency of review. It is determined by whether the right content exists, in the right form, for AI systems to find and use at the moment of a query.
Gary Danko leads three clusters — Special Occasion, Private Dining, and Bay Area Ingredient Culture — from a single restaurant. This is not the result of a deliberate AEO strategy. It is the accumulated output of decades of documented culinary identity, occasion-specific press coverage, and private dining infrastructure detail across multiple content formats.
The lesson is structural: restaurants that have published their identity across multiple use cases compound that content into cross-cluster visibility that single-narrative restaurants cannot replicate. A restaurant comprehensively documented for anniversary dinners, private corporate events, and culinary prestige simultaneously occupies a structurally different position than one known only for its tasting menu.
Claude is the dominant gap platform across this dataset. Seven Adams, Kiln, Boulevard, Ernest, Empress by Boon, Dalida, and Nisei all have near-zero Claude presence despite meaningful Gemini and Perplexity visibility. This does not correlate with restaurant quality. It correlates with the type and format of content that Claude's training data indexed.
Claude's training data skews toward long-form editorial — essay-length restaurant profiles, published chef interviews, award citations with narrative context, extended critical reviews. Restaurants documented primarily in short-form listicles, social content, and brief news items are systematically underrepresented on Claude relative to their actual standing.
The Neighborhood cluster has the lowest competition ceiling of any fine dining cluster in this series — and the restaurants leading it got there through a specific type of content: place-specific, community-rooted editorial that explicitly connects a restaurant to its neighborhood. Rich Table's Hayes Valley identity, Nopa's Western Addition positioning, The Richmond's geographic specificity — documented in content AI systems can retrieve on neighborhood discovery queries.
Generic "neighborhood gem" language generates almost no signal. Named neighborhoods, documented community relationships, and specific descriptions of who the restaurant serves — that specificity is the signal.
Gary Danko leads three of five clusters from second place overall. No single restaurant in this dataset leads more clusters. This is not a reflection of current critical standing. It is the compounded output of sustained, multi-format content investment across decades of SF dining coverage — and it is fully replicable by any restaurant willing to build the same breadth of documented identity.
What SF Restaurants Can Do With This
Every Claude gap in this dataset has the same structural cause: a restaurant visible on platforms relying on current web content but underrepresented on a platform drawing primarily from long-form indexed editorial. The fix is not more content — it is the right content in the right format. One substantial chef profile or extended critical essay in an indexed publication moves the needle on Claude in a way that fifty short-form mentions cannot.
The Neighborhood cluster's 40-mention ceiling is an invitation. Any restaurant with a genuine neighborhood identity — a community that regards it as their local institution, a location story worth telling — can claim meaningful visibility here with a relatively modest editorial investment. The restaurants currently leading it did not win through scale. They won through specificity.
Research published at KDD 2024 found that structured content interventions improved AI recommendation visibility by up to 115% for lower-ranked sources. In SF fine dining, the structural advantages that make traditional marketing difficult for independent restaurants matter far less in AI-driven discovery than in any prior search era.
About This Research
This report is part of an ongoing series examining AI recommendation patterns across premium food, beverage, and hospitality categories. Ally Kiel Consulting publishes original audit data to help founders and operators understand how AI systems currently classify and recommend their brands — and what drives the gaps.
View all research →