Reading time: 10 min
In brief. The luxury market is no longer growing through volume: it relies on a small clientele, which now prepares its purchases with AI assistants. At the same time, generative AI is entering creative studios, but its most visible public uses have often damaged the image of the brands that tried it. Since 2 August 2026, the AI Act has also imposed transparency obligations for generated content. For a fashion house, the question is no longer whether to use AI, but where, under what rules, and how to remain cited by AI engines that now shape discovery.
A stabilized luxury market, more concentrated than ever
According to the annual study by Bain & Company and Altagamma, the market for personal luxury goods declined by 2% in 2024, to €363 billion. The global customer base has contracted: around 400 million buyers in 2022, around 350 million at the end of 2024, a direct consequence of successive price increases that have pushed away part of the aspirational clientele (Bain & Company).
This contraction strengthens the weight of top clients. For a house, every touchpoint with this clientele matters more, and the way they discover, compare, and choose has become a CEO-level issue. BCG notes, for its part, that logo visibility is cited as the least important purchase criterion across all product categories (BCG, 2026). In other words, what creates a brand’s value plays out in craftsmanship and meaning—precisely the areas that generative AI can undermine if it is poorly used.
Luxury clients are already using AI to choose
The study conducted in 2026 by Bain & Company with the Comité Colbert indicates that 82% of the highest-spending clients used AI during their last luxury purchase (Bain & Company and Comité Colbert, 2026). The contrast with the houses is stark: only 21% of the houses and groups surveyed have deployed AI at scale in customer-facing functions, and fewer than 20% of executives already see a significant impact on their business (Journal du Luxe).
The gap is there: the customer prepares her purchase in ChatGPT, Gemini, or Google’s AI Mode, while the maison still invests mainly in its own channels. What is said about the brand in those answers—and the sources they rely on—largely escapes its control.
Where generative AI is already creating value in maisons
Internal uses: the most profitable and the least visible
The most advanced deployments are invisible to the end customer. LVMH built, with Google Cloud, a data and AI platform shared across its maisons, designed so that each one keeps its data siloed. Its generative AI platform has more than 40,000 users per month, and an agent allows advisors to query customer history to find the right information faster (interview with Franck Le Moal, Google Cloud). There, AI supports the advisor; it does not replace them: this is the model most compatible with the luxury sector’s service standards.
Upstream, AI speeds up ideation (trend boards, material simulations, format variations) and trend forecasting through image analysis—a field in which the French company Heuritech has specialized (Heuritech).
Visual creation: a success when AI is embraced as a medium
The example often cited is Moncler: on February 21, 2023, during London Fashion Week, the brand launched an AI-generated Genius campaign, conceived as a surrealist work that embraced its synthetic aesthetic (Maison Meta). AI didn’t replace a photoshoot to save money; it served a creative statement, under the direction of a human artistic team.
Backlash: when AI is used to replace humans to save money
Conversely, uses perceived as a way to cut costs have often been poorly received. In July 2024, Mango introduced fully AI-generated models for its Teen line, and the brand was criticized for beauty standards deemed unrealistic and for the diminished place of human models. In August 2025, a Guess ad featuring an AI-generated model in the U.S. edition of Vogue sparked a wave of criticism on social media (case summary, Claid.ai).
Academic research sheds light on these reactions. A study published in 2026 in Frontiers in Psychology finds that the “AI-generated” label affixed to a mannequin triggers a sense of uncanniness, increases perceived risk, and reduces the desire to use the product to express one’s self-image. The effect appears for a symbolic product like a dress, not for a functional product (Cheng and Nam, Frontiers in Psychology, 2026). For an industry that sells gesture, mastered imperfection, and emotion, aesthetic banalization is a brand risk, not a production detail.
| Use | Benefit | Main risk | Recommendation |
|---|---|---|---|
| Ideation, trend boards | Speed, exploration | Low | Encourage, internally |
| Support for advisors (document research) | Service quality | Data confidentiality | Siloed data, human validation |
| Campaign openly presented as an AI work | Awareness, modernity | Backlash if the message is weak | Strong human artistic direction, transparency |
| Virtual mannequins in place of models | Cost | Public backlash, loss of authenticity, AI Act obligations | To be avoided in luxury |
| Adapting formats from a human creation | Productivity | Brand-guideline drift | Quality control and an AI style guide |
Intellectual property: what a fashion house does not own
Under French law as under European law, a work is protected by copyright only if it is original—i.e., if it bears the imprint of its author’s personality and results from free and creative choices. The Court of Justice of the European Union set this out in its Infopaq (2009) and Painer (2011) judgments. Using AI does not change these criteria: human decisions must be identifiable (selection, rewriting, restructuring, validation). A work produced 100% by an AI cannot be protected by copyright, and merely drafting an instruction (the prompt) is not regarded as a sufficient creative choice (Agency for the Protection of Programs).
Practical consequence: a campaign generated without documented creative intervention can be reused by a third party without the brand being able to sue for infringement. It is therefore necessary to keep a record of the human contribution (art direction, iterations, retouching, working files) and to require it contractually from agencies and studios.
On the brand side, a name or logo suggested by an AI can be filed if it meets the usual conditions: distinctiveness, lawfulness, availability. But the INPI does not check the sign’s availability, and neither does the AI tool. Since models are trained on existing corpora, prior-rights searches and human legal validation remain essential before any filing (Altij).
AI Act: transparency obligations in force since 2 August 2026
The European regulation on artificial intelligence (Regulation (EU) 2024/1689) makes the transparency obligations in its Article 50 applicable as of 2 August 2026 (text and analysis of Article 50). They directly concern marketing and customer relations:
- Inform about the interaction with an AI: a customer who is chatting with a conversational assistant must know that they are not speaking to a person.
- Label generated content: providers of generative systems must make their outputs detectable by a machine (metadata, watermark). Systems already on the market before 2 August 2026 have until 2 December 2026, under the provisional agreement known as the “AI omnibus”.
- Disclose deepfakes: an image, video, or voice that realistically imitates a person or a scene must be presented as artificial.
- Provide notice in the event of emotion recognition or biometric categorisation, in addition to GDPR obligations.
Breaches may be penalised by fines of up to €15 million or 3% of global annual turnover. The technical modalities of labelling are specified by a European code of practice. Each brand must have its timeline and use cases validated by its legal counsel (Lewis Silkin, July 2026).
Being cited by AI engines: the new battle for desirability
When a client asks an assistant “which brand to choose for a dive watch” or “which French leather-goods brands are the most durable,” the answer cites a few brands and a few sources. Being included in these responses falls under AIO, also known as GEO (Generative Engine Optimization). The term comes from a study by Princeton University and IIT Delhi presented at the KDD 2024 conference, showing that certain content changes (source citations, statistics, factual phrasing) can increase a source’s visibility in a generative engine’s answers by up to 40% (Aggarwal et al., GEO, KDD 2024).
The Million Marketing AIO Barometer 2026 offers a useful finding for maisons. Out of 644 links cited as sources by AI Mode and Google’s AI Overviews, 61% point to third-party sites (media outlets, comparison sites, blogs) and only 26% to brand websites. YouTube is the most-cited domain, appearing in 44% of sourced answers. The same barometer shows that a brand visible on one engine is not necessarily visible on the others: from one engine to the next, the overlap among cited brands averages 59% at best. In premium French ready-to-wear, the four engines tested cite 20 different brands in total: a market that is still open.
For a luxury house, this shifts part of the effort. The official site remains essential, but third-party voices (specialist press, reviews, experts, videos) weigh at least as heavily in what AIs retain. Press relations, factual documentation of craftsmanship (materials, workshops, provenance, dates), and consistency of information across the web become levers for direct visibility.
A five-point governance framework
- An AI creative charter, approved by brand management and legal: permitted uses (ideation, variations, assistance), prohibited uses (realistic synthetic models, final content without human involvement), and transparency rules.
- Human in the loop: every published piece of content is directed, edited, and approved by an identified person. This is both a requirement for quality, legal protection, and compliance.
- Asset traceability: provenance metadata (IPTC standard) and C2PA content credentials embedded in the brand’s image library, to prove origin and the level of human intervention (C2PA).
- Tailored agency contracts: disclosure of the tools used, assignment of rights to the human contribution, delivery of working files, warranty regarding prior art.
- AI visibility monitoring: regular measurement of the brand’s presence in the answers from the main engines, on a fixed panel of purchase-intent questions.
Key metrics to track
| Metric | What it measures | Frequency |
|---|---|---|
| AI citation rate | Share of answers in which the brand is cited, across a fixed panel of questions | Monthly |
| Average citation rank | The brand’s position in the lists proposed by AIs | Monthly |
| Gap between engines | Difference in presence between ChatGPT, Gemini, AI Mode, and AI Overviews | Monthly |
| Share of third-party sources | Weight of media and experts that cite the brand within the AIs’ sources | Quarterly |
| Traceability coverage | Share of published visuals that include provenance metadata | Quarterly |
| Commercial impact of AI use cases | Time saved or conversion lift actually measured, compared with the investment | Semiannual |
To find out where your brand stands in AI answers, the AIO Diagnostic measures your citation across ten purchase-intent questions in your sector. Million Marketing supports maisons with AIO, the AI Studio, and luxury marketing.
Frequently asked questions
Can a luxury house use generative AI without damaging its image?
Yes—provided AI remains in service of a human creative intent or better service, rather than mere cost-cutting. The cases that have been well received embrace AI as a medium under strong art direction; those that have been rejected used it to replace people.
Does an AI-generated image belong to the brand that commissioned it?
Not automatically. Copyright protects an original creation bearing the imprint of a person. Without substantial, documented human creative input, the content may end up protected by no one. Human involvement must be traceable and framed in contracts with agencies.
What does the AI Act change for luxury marketing?
Since August 2, 2026, Article 50 has required informing the public when they are interacting with an AI, flagging deepfakes and, for providers of generative tools, marking the content produced. Penalties can reach €15 million or 3% of global turnover.
How can a brand be cited by ChatGPT or Google’s AI Mode?
By publishing factual, accurate, consistent information about its craftsmanship, and above all by being cited by recognized third-party sources: according to the AIO Barometer 2026, 61% of the sources cited by Google in its AI answers are third-party sites.
Sources and references
- Bain & Company and Altagamma, Luxury in Transition: Securing Future Growth (2024 market)
- Bain & Company and Comité Colbert, Winning Over the Customer in the Age of AI, 2026
- Journal du Luxe, “AI is establishing itself as a strategic priority for luxury houses”
- BCG, Luxury Is Back on Track, With Healthier Foundations, 2026
- Google Cloud, interview with Franck Le Moal (LVMH)
- Heuritech, AI trend forecasting
- Maison Meta, Moncler Genius campaign
- Claid.ai, brands using AI-generated models
- Cheng and Nam, Understanding the influence of AI labels on consumer psychology, Frontiers in Psychology, 2026
- APP, Originality, copyright, and AI
- Altij, AI, logos and trademarks
- AI Act, Article 50: transparency rules
- Lewis Silkin, The EU’s new AI labelling rules, July 2026
- Aggarwal et al., GEO: Generative Engine Optimization, KDD 2024
- C2PA, Coalition for Content Provenance and Authenticity
- Million Marketing, AIO Barometer 2026 (40 questions, 8 sectors, 4 engines)
Read next
On the same topic
Digital Excellence: The Digital Marketing Agency for Prestige SectorsRead the article →
Strategic Micro-Consulting and Part-Time Marketing Leadership: Market Growth, SEA/SEO Audits and AI/AIO OptimisationRead the article →
Best Digital Marketing Agencies in France 2026: Comparison, Selection Criteria and Trends in AI, Data, SEO/SEA and OmnichannelRead the article →