How AI models measure trust across fifteen luxury fashion brands

We asked four AI models to research fifteen luxury fashion brands. The median trust score is 68, and the whole category sits within 24 points, from Hermès at 80 to Tom Ford at 56. No house reaches the top trust tier.

The headline pattern is easy to state. The models describe almost every house accurately, and all but Balenciaga score 15 or more out of 20 on Clarity. They struggle to tell the houses apart, and no house scores above 13 on Identity Fidelity.

Each attribute score is built from five sub-scores, and each sub-score comes with the scorer's notes on what the models said. Luxury's hard-to-measure qualities show up in that layer. It shows whether the models can name a house's voice and whether they know who designs its collections now. It also shows whether what they repeat about a house builds trust or wears it down.

Hermès, Louis Vuitton and Chanel score the maximum 20 on AI Recommendation. Ralph Lauren scores 2 with some of the strongest fundamentals in the category. Prada had the best quarter, and Balenciaga the worst.

ABOUT THIS READING

We score every house on five attributes, each out of 20, for a total out of 100: Findability, Clarity, Sentiment, Identity Fidelity and AI Recommendation. Five tiers sort the total: AWESOME 83–100 / STRONG 70–82 / AVERAGE 56–69 / WEAK 40–55 / INVISIBLE 0–39. Six houses are STRONG and nine are AVERAGE. With a median of 68, the category sits two points under the STRONG line.

Two counts stay separate. Named is how often a house appears in its 60 moments. First is how often it is the first answer. A maximum score on AI Recommendation means a house appears often and high, and it does not mean the house leads. Where we claim primacy, we give the first-place count. Thirteen of this quarter's fifteen questions carry over word for word from Q2, and two are new. Movement is measured on those thirteen.

Winners

Twelve of the fourteen brands with a Q2 reading lost points this quarter, and the median change was -5. The four houses at the ends of the board show why scores moved.

Hermès leads at 80, down 6 and close to that median. All four models open with Hermès on value retention, craft tradition and handbags, and it comes first in 15 of its 60 moments. The drop came from Identity Fidelity, down 5, and Sentiment, down 3, where the models cite specific criticism of its quota system.

Prada rose 5 points to 75, the biggest gain in the category. The whole gain came from AI Recommendation, 6 to 16. It is the conversion our Q2 recommendation bet predicted, and we grade that bet delivered. Three models now name Prada on what is shaping fashion, where none did in Q2, and Gemini calls it “the intellectual's fashion house”.

The other two Prada bets did not land. No positioning content shipped, and Identity Fidelity fell 3 points. Recommendation rose without the handbag bet. The gain traces to fragrance, eyewear and a stronger commercial story, and nothing Prada published names the Galleria, Re-Edition 2000 or Cleo. Prada still comes first in only one answer, on distinctive design point of view.

Losers

Tom Ford closes the board at 56, down 14. The models name it on one question, menswear heritage, and never first. They miss Haider Ackermann's recent collections, and the Tom Ford website returns an error to AI agents. Identity Fidelity fell 6 points, Sentiment 3 and Findability 3.

Balenciaga fell furthest, 17 points to 61. Fifteen of those points came from the three areas our Q2 bets targeted. The biggest loss was the risk we flagged first: the models never picked up the Piccioli era. One model still names Demna as creative director, and none of the 60 answers names Piccioli. Perplexity invents fashion-tech features the house does not offer. Clarity fell from 18 to 12 and Identity Fidelity from 16 to 10.

The other two bets explain the rest. The craft bet widened the gap. The models still leave Balenciaga out of craft, investment and handbag questions, and name it almost only on streetwear. The reputation bet half landed, and all four models now tell the recovery story. Sentiment still fell 3 points, held down by a 1.6 out of 5 consumer rating and resale below retail. About 12 of the 17 points are Balenciaga's own, since the category fell by a median of 5.

The tie has a limit. The work happened, from a couture debut to new collaborations, but it reached press and customers, not text a model can find.

The Leaderboard.

A third of the luxury shelf goes to houses outside the fifteen we audit

Across all the answers, the models made 318 brand choices across 72 houses. The fifteen audited houses hold 66.0% of them, and the rest go to houses outside the cohort.

Those outside houses fall into clear groups. Brunello Cucinelli (6 choices), Loro Piana (5) and Brioni (4) form a quiet-luxury and tailoring group. Comme des Garçons (4), Maison Margiela (4) and Issey Miyake (3) form an avant-garde group, and Versace (6) and Elie Saab (3) hold eveningwear. No house we audit leads any of these lanes.

The cohort-only view makes luxury look like a settled order with Hermès on top. The full shelf shows a third of the choices going to houses these fifteen compete with every season. The full shelf is also what a buyer sees.

15 buyer questions x 4 models / audited brands in colour, every other brand in a stepped grey. Toggle between Cohort Only and Full Shelf.

KEY FINDING 01

The models can hear every house's voice and still can't say who the house is

All fifteen houses score 3 out of 4 on machine legibility. The models pick up a register for each house, and some reproduce it closely. Claude writes Chanel's aphoristic, restrained voice with precision, and it writes The Row in lines such as “The lining is silk. It matters.”

Naming the house is where it breaks down. No house gets all four models to agree on a single archetype. Chanel splits two against two, Ruler against Lover. Gucci, after three creative directors in two years, splits across three different archetypes. When two of the models write in Louis Vuitton's voice, they fall back on “crafted with precision” and “a study in timeless elegance”. A Hermès or Chanel answer could use the same lines.

Understood vs Distinct.

KEY FINDING 02

Taste moves faster than the models, and four houses keep AI agents out

Luxury changes creative directors often, and the models trail behind. In seven houses, at least one model presents an earlier designer or an outgoing aesthetic as current. Two models still name Chiuri and Jones at Dior. Claude names Sabato De Sarno at Gucci in one answer and Matthieu Blazy at Bottega Veneta. Loewe's design language is still framed through Jonathan Anderson after his move to Dior. None of the four models names Haider Ackermann at Tom Ford.

The gaps go past names. All four models still present the Margaux as The Row's flagship bag after the house discontinued it. For Balenciaga, none of the four describes what the house stands for in the Piccioli era. Perplexity also invents fashion-tech features the house does not offer, such as 3D weaving and bioengineered silk.

Some houses also make the models' job harder. The websites of Hermès, Chanel, Dior and Tom Ford return an error to AI agent traffic. An agent that goes to the source for current information is turned away. No house in the category scores above 2 out of 4 on agent readiness.

KEY FINDING 03

The models remember the biggest houses in detail and report mixed views of them

Cultural memory is close to full. Fourteen of the fifteen houses score 4 out of 4, and the models tell the stories in detail. They know Grace Kelly and the Kelly bag, Dior's New Look, Tom Ford's revival of Gucci and Burberry's First World War trench coat.

What the models report people saying is less settled. Chanel, Louis Vuitton, Dior, Balenciaga, Gucci and Burberry all score 2 out of 4 on sentiment polarity. The rubric defines that as roughly balanced praise and criticism. The criticism is specific, from Chanel's price rises to Dior's labour case in Italy. Almost every Trustpilot score the models quote sits under two stars, and Loewe's sits at 4.6 to 4.7.

Institutional trust splits the category a different way. Miu Miu and The Row are the only houses at 4, backed by named critics, CFDA awards and fashion rankings. Chanel, Dior, Gucci and Balenciaga sit at 2, where regulators and courts weigh against the positive coverage. The notes cite the EU's €119M antitrust fine for Gucci and competition authorities for Chanel. They also cite court action in Italy for Dior and the revoked BoF award for Balenciaga.

The models also disagree about direction. For Louis Vuitton, one model describes a peak followed by a correction and another describes a stable rise. For Balenciaga, GPT-4o calls the revival a success while the other three call it incomplete. A buyer who asks two different models can hear two different stories about the same house.

KEY FINDING 04

Chanel is named most often, and Hermès is named first most often

Chanel appears in 36 of its 60 moments, eight more than Hermès and Louis Vuitton at 28 each. Chanel comes first in 11. Hermès comes first in 15, and Louis Vuitton, with the same reach as Hermès, comes first in 5. All three score the maximum 20 on AI Recommendation, so the leaderboard shows one score for three different results.

The Clarity notes add a detail. Three of the four models explain Hermès' difference through how the house works. They name one artisan per bag, saddle stitching, its own tanneries and the purchase history needed to buy certain bags. For eight houses, Perplexity explains the difference in business terms instead, such as pricing, scarcity, distribution or the parent group.

Below the top three, first places almost stop. Every other house comes first in two moments or fewer, and Saint Laurent, Ralph Lauren, Burberry and Tom Ford come first in none. Five of the top seven houses (Hermès, Louis Vuitton, Chanel, Dior and Gucci) are recommended above what their fundamentals support.

KEY FINDING 05

Four houses have strong fundamentals and little recommendation

Saint Laurent, Ralph Lauren, Burberry and Tom Ford share one position. Saint Laurent scores 14 on Findability, 17 on Clarity, 17 on Sentiment, 12 on Identity Fidelity and 7 on AI Recommendation. Ralph Lauren scores 15 / 17 / 17 / 12 and 2 on recommendation. Burberry scores 14 / 17 / 15 / 10 and Tom Ford 11 / 16 / 16 / 10, and both score 3.

The sub-scores show the models have plenty of material. Saint Laurent's notes list Le Smoking, the Mondrian dress and a 1,100-page Styleforum thread. Ralph Lauren's list Annie Hall, Gatsby and the Polo Bear. For both houses three of the four models agree on an archetype, which only Louis Vuitton, Balenciaga and Loewe also reach. The models still name Ralph Lauren in 2 of 60 moments, Tom Ford in 2, Burberry in 3 and Saint Laurent in 8.

This looks like a conversion problem more than a comprehension problem. The models know these houses well. The link between that knowledge and the specific occasions buyers ask about is what is missing.

Brand Fundamentals vs Recommendation.

WHAT THIS MEANS

The top-line scores say luxury fashion is well described and hard to tell apart. The layer underneath says why, and it points at things a house can act on. A house can write its voice down so the models have something to repeat. It can put current creative direction and product where agents can read them. The criticism the models carry about price and quality is specific enough to track.

None of this holds still. Creative directors change, collections turn over and sentiment moves from one season to the next, and the models trail each change. That is why we read the category every quarter. A single score tells a house where it stands. The quarterly reading shows whether the models are catching up with the house or drifting away from it.