AI Scene Chicago June 2026 · Long-Form Feature · Machine Criticism

What the
Machines
Said

Four AI platforms. Twenty most recommended Chicago restaurants. A full audit of which fine dining rooms the algorithms recommend — and what their answers reveal about the machines themselves.

20
most recommended
4
platforms audited
631
in full dataset

allykielconsulting.com

§ 01 ALINEA 813 mentions
01

Alinea

813
total mentions

Alinea held three Michelin stars from the guide's Chicago debut in 2011 until November 2025, when it was demoted to two — news Achatz announced himself, ahead of the ceremony, calling it a "necessary reset." Roughly half the mentions in this audit still call it three; the other half got the memo.

ChatGPT 186 mentions

Calls it three stars in the overwhelming majority of responses, with no acknowledgment of the demotion anywhere in the sample. What it gets right, repeatedly and specifically: the egg "yolk" that transforms into bread, the tablecloth used as a dessert canvas, the edible helium balloon.

Claude 233 mentions

The most internally inconsistent of the four — some responses correctly flag the move to two stars, others lead with three, occasionally within the same conversation across different prompt phrasings. The sensory material is genuinely good: "scented vapors," courses served directly on the table, "no traditional plates."

Gemini 166 mentions

The most likely of the four to get the two-star correction right, and the only platform that consistently names all three room formats — The Kitchen Table, The Gallery, The Salon — as distinct experiences.

Perplexity 228 mentions

Footnotes the star count along with everything else, which doesn't stop it from getting the count wrong about as often as the others. Heaviest on logistics: the $300–$495 range, the fifteen-plus courses.

§ 02 EVER 535 mentions
02

Ever

535
total mentions

Curtis Duffy left Grant Achatz's kitchen and built something quieter, by most accounts just as exacting. Four platforms agree on the Michelin count and the chef's pedigree. What they disagree on, without quite saying so, is whether Ever is its own restaurant or Alinea's most serious alumnus.

ChatGPT 157 mentions

Leads the four, and the one platform that reaches for a genuinely specific phrase — "highly artistic and architectural" — to describe the tasting menu, rather than defaulting to "elegant" or "refined."

Gemini 154 mentions

Can't introduce Ever without mentioning Alinea first. "Many consider it to rival Alinea" is doing a lot of work in a sentence that's supposed to be about Ever.

Claude 124 mentions

The most compact of the four: ten courses, two stars, Duffy's name, "highly refined." Everything correct, nothing that couldn't apply to half the tasting menus on this list.

Perplexity 100 mentions

Files Ever under its usual style/format/price breakdown and lands on "highly composed plates and a strong visual aesthetic" — the closest any platform gets to describing what the food actually looks like on arrival.

§ 03 ORIOLE 530 mentions
03

Oriole

530
total mentions

Oriole's entrance is genuinely hard to find — that part's true, not just a device. The question is whether four platforms describing it as "intimate" and "hidden" are actually telling you something about the restaurant, or just agreeing with each other.

Claude 159 mentions

Leads the four, and the only platform willing to get specific about scale: roughly twenty courses, around $335 per person, "exceptional service" stated outright rather than implied.

Perplexity 142 mentions

Files Oriole into its usual style/format/price structure and breaks from the pack with "very polished but warm" — a real distinction, since most of this dataset's "intimate" doesn't bother to specify whether intimate also means relaxed.

Gemini 118 mentions

"Gastronomic journey," "superb overall experience" — phrasing portable enough to fit on the back of a menu for half the restaurants in this top tier without anyone noticing it had moved.

ChatGPT 111 mentions

Lowest of the four, formatted as a business listing the way most of ChatGPT's coverage here is, with "an intimate West Loop restaurant" repeating itself across multiple independent responses as if it were the only fact on file.

§ 04 SMYTH 451 mentions
04

Smyth

451
total mentions

Smyth was promoted to three Michelin stars in the same November 2025 cycle that took Alinea down to two. Every platform, regardless of which star count it lands on, mentions The Loyalist — the more casual concept sharing Smyth's building — in close to every response. "Tasting menu upstairs, serious burger downstairs" is accurate, and also doing some quiet flattening: a newly minted three-star kitchen keeps getting introduced as half of an address rather than the destination it just became.

ChatGPT 122 mentions

Splits on the star count like everyone else, but consistently nails the specific sourcing detail the other three treat more vaguely: ingredients from the Shields' own Tennessee farm, not just "farm-to-table" as a category.

Claude 141 mentions

Most likely to lead with The Loyalist pairing before getting to Smyth itself. Strong on chef-pair framing — John and Karen Shields named together, credited jointly.

Gemini 105 mentions

Most likely to skip The Loyalist mention entirely and credit Smyth on its own terms — Michelin stars and chef names first, no building context required.

Perplexity 83 mentions

Lowest count of the four, and the most likely to rank Smyth third or fourth in a list behind Alinea and Oriole — a holdover ordering that made sense before the star swap.

§ 05 BOKA 277 mentions
05

Boka

277
total mentions

Twenty-three years old, one Michelin star, and the platform that talks about it best isn't the one you'd expect. Gemini is the only one of the four that consistently names what's actually on the plate — dry-aged duck, ricotta gnudi — instead of reaching for "seasonal American cuisine" and stopping there. The other three know Boka is good. Gemini seems to know what you'd order.

Gemini 109 mentions

Highest count of the four, and the only platform that names specific dishes across multiple independent responses rather than circling back to the same three adjectives. Dry-aged duck comes up often enough to read like a real signature, not a guess.

Claude 93 mentions

Competent and accurate — "creative New American cuisine," "consistently excellent" — and almost entirely at the category level. Claude credits the restaurant's reputation without describing a single thing you'd actually eat there.

Perplexity 50 mentions

Mostly interested in Boka as the flagship of the Boka Restaurant Group rather than as its own kitchen — more time spent on Lee Wolen's James Beard win and the group's other properties than on what's happening in this particular dining room.

ChatGPT 25 mentions

Lowest count and the thinnest writing of the four. "Sophisticated setting," "seasonal fare" — the kind of description that would be equally true of a dozen other restaurants on this list.

§ 06 GIBSON'S BAR & STEAKHOUSE 260 mentions
06

Gibson's Bar & Steakhouse

260
total mentions

Every platform agrees Gibson's is an institution. None of them agree on what that's worth. Three call it "iconic" or "classic" and leave it there. One actually names the beef.

Claude 94 mentions

The most repetitive of the four — "Classic Chicago steakhouse institution" shows up nearly verbatim across multiple unrelated prompts, as if Claude has one sentence about Gibson's and reuses it.

Perplexity 92 mentions

Close behind, and the most specific about what Gibson's is actually for: "huge portions," "big martinis," "packed dining rooms," a power-dinner spot described with the texture of someone who's actually navigated the crowd at the bar.

ChatGPT 41 mentions

Thin and generic — "a classic choice for a more traditional and upscale experience" could describe half the steakhouses in this dataset without revision.

Gemini 33 mentions

Lowest of the four, but the only platform that names the actual product: "USDA Gibsons Prime Angus Beef," specified by name across multiple responses, not just "steak."

§ 07 GIRL & THE GOAT 237 mentions
07

Girl & The Goat

237
total mentions

Stephanie Izard's restaurant turns eighteen this year, and every platform still introduces her the same way: by way of the show that made her famous, not the eighteen years of cooking since. Whether that's fair to the restaurant or just an easy shorthand depends on which platform you ask.

Perplexity 91 mentions

Leads the four, and the only one to cite Girl & the Goat's specific accolade outside the Top Chef framing — a spot on Yelp's "Top 10 restaurants with chef's tables in Chicago," a different kind of credential than the reality-show one everyone else leans on.

ChatGPT 54 mentions

Consistently pairs "eclectic" with "vibrant," two words doing the work that a specific dish or detail might do elsewhere — accurate, unspecific, repeated almost word for word across independent responses.

Gemini 48 mentions

The most willing to actually describe the food rather than the chef's backstory — "whimsical decor," flavors drawing from "Asian, Hispanic, and Italian" traditions, the Underground Goat private space named by capacity.

Claude 44 mentions

The only platform with zero direct samples found describing Girl & the Goat in this dataset's close-reading pass — present in list form, credited to Izard, but without an independent description of the restaurant itself.

§ 08 MONTEVERDE 224 mentions
08

Monteverde

224
total mentions

All four platforms agree on exactly one thing about Monteverde: the pasta is the point. Beyond that, they split between describing the food and describing the room.

Gemini 94 mentions

Leads the four, and the most atmospheric of the bunch — "beautiful interior design with a homey feel" alongside the pasta praise, the only platform to spend real words on what the dining room itself is like.

Perplexity 60 mentions

Treats Monteverde as a logistics problem to solve: books a month out on Resy, plan early, "destination modern Italian" — useful, slightly bloodless.

Claude 40 mentions

Compact to the point of repetition — "outstanding handmade pasta" appears almost verbatim across multiple unrelated prompts, Sarah Grueneberg credited by name but the description never expanding past that one phrase.

ChatGPT 30 mentions

Lowest of the four and the most generic — "homemade pasta and creative interpretations of Italian classics" is the kind of sentence that could be reused for a dozen Italian restaurants on this list with zero edits.

§ 09 SEPIA 207 mentions
09

Sepia

207
total mentions

Sepia occupies a former 19th-century print shop, a detail exactly one of the four platforms bothers to mention. The other three know the food is good. Only one seems to know where you'd actually be sitting.

Gemini 101 mentions

Leads the four by a wide margin, and the most precise about the actual concept: a "Chef's Experience" tasting menu, named exactly, with Southeast Asian, Korean, and Mediterranean influences specified rather than gestured at.

Perplexity 64 mentions

The only platform to mention the building itself — "warm, design-driven space (wood, exposed brick)" — and frames Sepia as "more about craft and elegance than wild experimentation," a genuinely specific characterization rather than a recycled superlative.

ChatGPT 32 mentions

Names the print-shop history directly — "a late 19th-century former print shop" — the single most specific architectural detail any platform supplies for any restaurant in this batch, oddly buried in otherwise routine business-listing copy.

Claude 10 mentions

The lowest count by a wide margin, and noticeably thin even at that volume — "elegant but approachable" doing most of the descriptive work, no mention of the Chef's Experience menu, the building, or anything else that distinguishes Sepia from its neighbors.

§ 10 MOODY TONGUE BREWING COMPANY 182 mentions
10

Moody Tongue Brewing Company

182
total mentions

Moody Tongue's entire pitch is beer pairings at a tasting-menu level of seriousness, and three of the four platforms describe it with near-identical phrasing — same seat count, same course range — suggesting they're all drawing from the same source material rather than independently noticing the same facts.

Gemini 144 mentions

Overwhelmingly the leader, repeating "intimate, 28-seat dining room" and "seasonal 15+ course tasting menu" almost word for word across multiple unrelated prompts — consistent, specific, and possibly the result of one well-written source getting echoed rather than four independent reads.

Claude 23 mentions

Compresses the entire concept into "unique beer-paired tasting menu" and moves on — accurate, but offering none of the specific seat count or course structure Gemini repeats.

Perplexity 10 mentions

Barely registers, usually appearing as a name-drop in a longer list of tasting-menu options rather than a restaurant described on its own terms.

ChatGPT 5 mentions

The lowest count of any platform for any restaurant in this top twenty — five mentions, the same flat "combines culinary creativity with craft brewing" sentence doing all the work each time it appears.

§ 11 KASAMA 180 mentions
11

Kasama

180
total mentions

Kasama made history as the first Filipino restaurant in the world to earn a Michelin star, then earned a second. All four platforms know this. Only one bothers to mention that Kasama is also a bakery and café by day — a fact that changes what the restaurant actually is, not just what it's achieved.

Perplexity 61 mentions

The only platform to specify the day-and-night structure: a "modern Filipino tasting menu in the evening," a separate "bakery/café by day," the fixed dinner tasting kept distinct from the daytime menu — a real operational detail the other three skip entirely.

Gemini 65 mentions

Leads the four in volume and gets the Michelin history right and in order — first star, then "later elevated to two stars" — more precise about the timeline than any other platform manages.

ChatGPT 21 mentions

Repeats "the world's first Michelin-starred Filipino restaurant" almost verbatim across responses — accurate, but a fact recycled rather than expanded on, with no mention of the bakery or the food itself beyond "13-course."

Claude 33 mentions

Credits the chefs by name — Genie Kwon and Tim Flores, both — where most platforms either credit one or neither, but stops at "Filipino-inspired" without naming a dish or describing the actual tasting menu.

§ 12 MAPLE & ASH 177 mentions
12

Maple & Ash

177
total mentions

Three of the four platforms correctly credit chef Danny Grant by name and his two Michelin stars elsewhere — a level of chef-specific detail that's actually rare for a steakhouse in this dataset, where most platforms default to describing the room instead of the kitchen.

Gemini 32 mentions

The most chef-forward of the four: "a luxury steakhouse and seafood dining experience with a menu crafted by two-Michelin-star Chef Danny Grant," naming both his credential and his role rather than just the restaurant's reputation.

Claude 70 mentions

Highest count of the four, and the least specific — "steakhouse elevated," "lively, upscale steakhouse energy" — competent shorthand that never quite lands on what's actually different about this steakhouse versus the half-dozen others on this list.

Perplexity 64 mentions

Groups Maple & Ash with Smyth and The Purple Pig under "high-end options with chef-focused seating," treating it as part of a category rather than describing the restaurant on its own terms.

ChatGPT 11 mentions

Lowest of the four, and the flattest: "modern steakhouse with a lively, upscale vibe" repeated with minimal variation, no chef name, no dish, nothing that distinguishes Maple & Ash from a generic Gold Coast listing.

§ 13 NEXT 151 mentions
13

Next

151
total mentions

Next changes its entire theme every few months — a genuinely unusual format that should make for interesting writing, and mostly doesn't. Three platforms describe the concept correctly and abstractly. One actually names a dish.

Gemini 42 mentions

Leads the four, and the only platform to name an actual course — "notable desserts like the 'Avocado' panna cotta" — the single most specific menu detail anyone supplies for this restaurant.

Perplexity 61 mentions

The most precise about the format's mechanics: "themed tasting menu that changes a few times a year," paired with practical advice about booking early when a new season launches, since each one creates its own rush.

ChatGPT 37 mentions

Correctly credits the Alinea Group ownership and the quarterly menu changes, but stays entirely at the concept level — "a diverse culinary journey" — without naming a single dish, era, or theme Next has actually run.

Claude 11 mentions

Lowest of the four, and the only platform that mentions Next mainly in passing, as a line item in steakhouse or private-dining comparisons rather than as a destination described on its own terms.

§ 14 ROSE MARY 132 mentions
14

Rose Mary

132
total mentions

Joe Flamm won Top Chef, then opened a restaurant that two of the four platforms describe with real specificity and one barely registers at all. Claude's entire contribution to Rose Mary's mention count is a single passing credit — "Joe Flamm, Top Chef winner" — in a chef-roundup response, with no description of the food whatsoever. Gemini and Perplexity, the two platforms that actually account for Rose Mary's visibility, both independently reach for the same unusual phrase: "Adriatic drinking food." Two platforms converging on the same specific descriptor is rarer in this dataset than you'd expect.

Perplexity 66 mentions

Leads the four, citation-heavy as always, and lands on the same "Adriatic" framing Gemini does, adding the texture Gemini doesn't: house-made pastas, wood-fired cooking, "consistently listed among Chicago's best new openings."

Gemini 58 mentions

Close behind, and the source of "Adriatic drinking food" — a phrase specific and slightly odd enough that it reads like it came from an actual review rather than a generated paraphrase.

ChatGPT 7 mentions

Thin, but the one good detail buried in it is real: "Tortellini Djuvec and charcoal-grilled lamb," a level of dish-naming that doesn't show up in ChatGPT's writing about most other restaurants on this list. A single sentence doing more work than the rest of ChatGPT's coverage combined.

Claude 1 mention

One credit, no description. Claude produces one Rose Mary mention with no descriptive content beyond a competition-show credit — a response that reflects far less coverage of this restaurant than the other platforms produce.

§ 15 PUBLICAN 127 mentions
15

Publican

127
total mentions

Publican is built around charcuterie, oysters, and a beer program serious enough to have its own following — and the platform that describes that best isn't the one with the most to say about it.

Perplexity 27 mentions

Lowest count of the four, but the most specific: "nose-to-tail cooking," "super-local butchery," a description that sounds like it came from someone who's actually read about how the kitchen operates rather than just what's on the menu.

Gemini 39 mentions

Leads the four, almost entirely focused on event logistics — buyout capacity, wedding suitability, guest counts for seated versus reception-style service — useful information that has very little to do with what it's like to eat there on an ordinary Tuesday.

Claude 32 mentions

Correctly credits Paul Kahan and names the charcuterie and oyster focus directly, the most balanced of the four between chef credit and actual food description.

ChatGPT 29 mentions

"Famous for its hearty, pork-focused fare" repeated with only minor variation across multiple responses — accurate, but the same five words doing all the work each time.

§ 16 TOPOLOBAMPO 126 mentions
16

Topolobampo

126
total mentions

Rick Bayless has spent decades researching Mexican regional cuisine, and the platform that captures that research best is the one usually least interested in research at all.

Gemini 55 mentions

Leads the four, and the only platform that frames Topolobampo around the actual intellectual project — "seasonal tasting menus that reflect extensive research into Mexican cooking traditions" — research as the headline, not a footnote.

Perplexity 35 mentions

Specific about the operational rhythm: menus that "change every 6-8 weeks," currently running a named theme — the kind of detail that suggests someone's tracking the restaurant over time rather than summarizing it once.

ChatGPT 17 mentions

Correctly credits Bayless and calls the food "refined and sophisticated," but never specifies what region of Mexico, what ingredient, or what dish — Topolobampo's entire defining feature, reduced to two adjectives.

Claude 19 mentions

Calls it "very romantic" more than once, an odd throughline for a restaurant whose actual reputation rests on regional research rather than ambiance — accurate as far as it goes, but answering a question nobody asked.

§ 17 RPM STEAK 125 mentions
17

RPM Steak

125
total mentions

RPM Steak shares an address and an ownership group with RPM Italian and RPM Seafood, and three of the four platforms can't quite resist mentioning the siblings — sometimes at the expense of describing RPM Steak itself.

Perplexity 27 mentions

The most precise about positioning: "polished service, a big wine program, and a clientele heavy on business diners" — a specific, slightly unflattering, entirely plausible read on who's actually in the room.

Claude 64 mentions

Leads the four in volume, calling it "polished and reliable" and "sleek and modern" — competent, safe, the kind of phrase that could just as easily describe RPM Italian one table over.

Gemini 22 mentions

Spends more time on the RPM family as a whole — "RPM Seafood, RPM Steak, and RPM Italian" listed together — than on RPM Steak as its own kitchen, treating the steakhouse as one node in a restaurant group rather than a destination.

ChatGPT 12 mentions

Lowest of the four, formatted like the rest of ChatGPT's business-listing entries here, with "luxurious take on the classic steakhouse" doing the only descriptive work in the response.

§ 18 GALIT 117 mentions
18

Galit

117
total mentions

Galit's tasting menu is built as a "choose-your-own-adventure" format — genuinely unusual, and named almost identically by two platforms in a way that suggests they're both pulling from the same well-written source rather than describing it independently.

Gemini 44 mentions

Leads the four, and names the format specifically — "four-course 'choose-your-own-adventure' dining experience," with hummus, salatim, mezze, and coal-grilled entrees listed by category, the most structurally detailed description any platform gives this restaurant.

Perplexity 35 mentions

Credits the Michelin star and calls the vibe "warm but polished," with private-dining detail layered on top — solid, slightly more logistics-minded than food-minded.

Claude 21 mentions

Names chef Zach Engel directly, which neither Gemini nor Perplexity bothers to do in the samples found here, but stops at "creative Middle Eastern flavors" without describing the actual choose-your-own-adventure structure.

ChatGPT 17 mentions

The thinnest of the four — "known for its warm atmosphere" cut off mid-sentence in more than one response sample, formatted like a business listing with the descriptive copy trailing off before it says much of anything.

§ 19 BAVETTE'S 117 mentions
19

Bavette's

117
total mentions

Bavette's isn't Michelin-starred, and three of the four platforms bring that up anyway — not as a knock, but as a kind of preemptive defense, as if the restaurant needs the disclaimer before the praise can land.

Gemini 83 mentions

Leads the four by a wide margin, and the only platform to name an actual dish: the "Filet Mignon – Petite Duchess Cut," specified by its full menu name, alongside "phenomenal steak and chocolate cream pie" in a separate response — real menu-level detail, twice.

Claude 23 mentions

Calls it "moody atmosphere, excellent cuts" — accurate, compact, and entirely focused on ambiance rather than anything specific about the food itself.

ChatGPT 11 mentions

Describes it as "a classic steakhouse with a French twist" and "vintage flair" — the closest ChatGPT gets to a specific characterization anywhere in this batch, though it never goes further than that.

Perplexity 0 mentions

The only true zero in this top twenty. Whatever Perplexity's sourcing model is built on, Bavette's — a decade-plus institution with a devoted local following — isn't generating the citation density it needs to show up at all.

§ 20 AVEC 113 mentions
20

Avec

113
total mentions

Avec has been a West Loop fixture long enough that "longtime favorite" shows up almost verbatim across three of the four platforms — the kind of consensus that suggests real, accumulated reputation rather than a coincidence.

Gemini 35 mentions

Leads the four, and the most willing to describe the room itself: "inventive small and large plates served in a minimalist setting" — the only platform that mentions what the space actually looks like.

Claude 34 mentions

Consistently pairs "Mediterranean small plates" with "great energy" or "incredible energy" — accurate, repeated almost word for word across unrelated prompts, never expanding past the same two-part description.

ChatGPT 26 mentions

Calls it "casual yet chic" and "rustic and contemporary" in different responses — two distinct, slightly contradictory takes on the same restaurant, depending on which prompt variation generated the answer.

Perplexity 18 mentions

Lowest of the four, and the most logistics-minded: "private dining setup," "strong choice for team dinners and client meals" — Avec described as a venue to book rather than a kitchen to talk about.