AI Recommendation Patterns:

Chicago Fine Dining

Ally Kiel · AI Discoverability Strategy for Hospitality + F&B Brands · April 2026

This is original audit data collected in April 2026 by Ally Kiel, AI Discoverability Strategist, across Claude, ChatGPT, Gemini, and Perplexity using 50 prompts across five intent clusters, mapping not just which restaurants appear, but which intent categories they own and where platform concentration creates vulnerability.

Research from KDD 2024 (Princeton and IIT Delhi) found that structured content interventions improved AI recommendation visibility by up to 115% for lower-ranked results, meaning the gaps identified here are not fixed. Results reflect which restaurants are being recommended — and which are invisible — when diners use AI to plan a meal, celebrate a milestone, or entertain a client.

Methodology: Queries were run via API across Claude, ChatGPT, Gemini, and Perplexity — not consumer web interfaces. API responses reflect static training data; consumer-facing products may return different results due to live web access. Each prompt was run twice and results averaged to reduce single-run variance. Semantic query variations were tested alongside original prompts. Brand mentions were extracted using named entity recognition. Results represent baseline AI training data visibility — the floor, not the ceiling.

Note on legacy concepts: AI training data reflects the web as it existed at a prior point in time. Closed or rebranded restaurants with strong historical web presence — press coverage, reviews, award citations — may continue to surface in results. This audit identified three such instances: Acadia (closed 2020), Sixteen (closed 2018), and Fat Rice (closed 2021, now NoodleBird). These are flagged where they appear.

Platform Divergence — Top 15 Restaurants

The table below shows mention counts per platform for the fifteen most visible Chicago restaurants. The patterns reveal which properties have built durable cross-platform signals versus those whose visibility is concentrated on one or two platforms — a meaningful distinction for any restaurant relying on AI-driven discovery.

Results by Cluster

What Drives AI Visibility

The restaurants dominating these results share specific content signal patterns. Visibility is not determined by food quality, critical reputation, or review volume — it is determined by the depth, specificity, and accessibility of structured content that AI systems can find and use. Three signal types account for the majority of high-visibility patterns in this audit.

Signal 1 — Named Chef with Editorial Presence

Every restaurant in the top five of the Chef & Culinary Identity cluster has a chef whose name, credentials, culinary philosophy, and awards are documented in publicly accessible, AI-legible formats — interviews, press profiles, award citations, and first-person content. This is the single strongest predictor of cross-cluster visibility. A restaurant with an exceptional chef whose story is not published in structured, findable content is invisible in chef-identity queries regardless of actual talent.

Alinea, Ever, Oriole, Smyth, Kasama, and Girl & the Goat all demonstrate this pattern. Next and The Publican demonstrate the inverse — strong overall visibility that drops sharply in chef-identity queries.

Signal 2 — Specific Private Dining Infrastructure Content

Gibsons and the RPM Restaurant Group dominate the Private Dining cluster not because they are the only restaurants with private rooms — but because their private dining content is specific, structured, and findable. Room names, capacity figures, event menus, booking processes, and dedicated landing pages give AI systems the structured data needed to surface them confidently for corporate and event queries. Generic descriptions of "available for private events" are invisible.

Gibsons (193 private dining mentions) versus The Signature Room (100 overall mentions but absent from the private dining cluster top five) illustrates the gap between specific infrastructure content and general availability language.

Signal 3 — Occasion and Neighborhood-Specific Editorial Content

Restaurants that surface in occasion and neighborhood clusters have content that explicitly addresses the consumer's decision context — not just what they serve but when, where, and for whom. Wood appears in neighborhood queries because its content signals a specific location and community identity. Bavette's appears in anniversary and romantic occasion queries because its atmosphere and intimate dining framing maps directly to the intent behind those queries.

Neighborhood-specific queries outside West Loop and River North return thin results because almost no Chicago restaurants have published hyperlocal, neighborhood-identity content that AI systems can use to surface them confidently for those queries.

Key Findings

Platform concentration is a hidden vulnerability. Several top-15 restaurants have significant platform concentration — Kasama generates 105 of its 169 total mentions from Gemini alone; Next generates 85 of its 180 from ChatGPT alone. A restaurant whose visibility depends on a single platform is exposed to meaningful recommendation loss if that platform's training data or behavior changes. Cross-platform visibility requires deliberate signal building across all four ecosystems — not just a strong presence on one.

Private dining visibility is determined by content specificity, not product quality. Gibsons and the RPM Restaurant Group dominate the private dining cluster not because they have the best private rooms in Chicago — but because their private dining content is specific, structured, and findable. Room names, capacity figures, event menus, and dedicated booking pages give AI systems the data needed to surface them confidently. Restaurants with equivalent infrastructure but generic "available for private events" language are invisible in this cluster entirely.

Neighborhood queries outside West Loop and River North are effectively unclaimed. AI platforms return thin and inconsistent results for neighborhood-specific dining queries outside the two dominant areas. This is the most accessible visibility gap in the entire audit — the competition for recommendation space is demonstrably lower and the content signals required are achievable without national press coverage or a celebrity chef. For any restaurant in an emerging or under-indexed neighborhood, this cluster represents the clearest near-term opportunity.

AI visibility can be engineered and Sepia is the clearest proof in this data. Sepia ranks 9th overall with 237 total mentions, but outperforms restaurants ranked 3rd and 4th in the PrivateDining and Special Occasion clusters — the two highest-value commercial contexts in the audit. What makes this result instructive is not just that it happened, but why. Sepia operates a fully developed private dining product: a named venue (Private Dining by Sepia), 1,100 square feet, capacity for 60 seated or 100 standing, three menu tiers with per-person pricing, named chefs, wine pairings, A/V specs, and a detailed FAQ covering deposits, vendors, and custom desserts. Every element AI systems need to surface a venue confidently for private dining queries already exists. It is documented in a private dining menu PDF, which AI crawlers cannot read. The cluster performance comes not from the restaurant's own site, but from off-site platforms where the offering is named, specific, and consistently described. The finding is not that Sepia has strong content. It is that Sepia's content is in the wrong format and the wrong place and it is still outperforming tier-one properties. Publishing what already exists as crawlable HTML would be among the highest-leverage content moves available to any Chicago restaurant in this audit.

Platform Concentration Gaps — Gemini Visibility

The following restaurants have meaningful AI visibility across ChatGPT, Perplexity, and Claude, but zero mentions on Gemini. For any active property in this list, Gemini represents an unaddressed platform gap. Note that three entries — Acadia, Fat Rice, and Sixteen — are closed or rebranded concepts whose historical web presence continues to generate AI mentions. This is itself a finding: AI systems are recommending restaurants that no longer exist, reflecting the lag between real-world closures and the decay of training data signals.

What Restaurants Can Do With This

The gaps identified in this audit are not fixed. AI visibility is not a function of how long a restaurant has been open, how many reviews it has, or how well-known it is to local diners. It is a function of whether the right content exists, in the right form, in the right places for AI systems to find and use it.

Restaurants that close these gaps typically do so through two types of interventions: structured content development that gives AI systems specific, named, verifiable signals to work with; and site architecture changes that surface existing content in the formats and at the depth levels AI crawlers prioritize. Schema markup may support entity recognition over time as part of broader technical infrastructure, but recent evidence suggests it does not directly drive AI citation outcomes; the content itself is the primary lever.

The restaurants that move from invisible to recommended are not always the ones that spend the most or have the highest profiles. They are the ones that understand what AI systems are looking for and build deliberately toward it.