Whenever a fashion brand announces a new AI tool — a virtual try-on, a personalised styling assistant, a demand-forecasting engine — the press release follows a familiar shape: efficiency gains, sustainability promises, a better experience for you. What it rarely contains is the full picture. Below are six things that tend to get left out, and why they matter to you as a consumer.
Key takeaways
- Fashion brands seldom disclose where the data that trains their AI tools actually comes from.
- AI sizing and fit tools can perform unevenly across body types, yet brands rarely publish accuracy breakdowns.
- Regulatory pressure on AI transparency is growing, but consumer-facing disclosure remains thin.
- Job displacement in fashion production is a documented concern that AI announcements almost never address.
- The gap between an AI-generated image and a garment you can actually wear is wider than marketing suggests.
1. Where does the training data come from?
Every AI tool learns from data. A sizing recommendation engine learns from body measurements and past purchases; a trend-forecasting model learns from images scraped across the internet. Brands almost never tell you whose data was used, whether it was collected with consent, or whether the people whose images or measurements trained the model were compensated in any way.
This matters for two reasons. First, data sourced without clear consent raises ethical questions that sit squarely with the brand, not the technology vendor. Second, if the training set skews toward a particular demographic — say, measurements collected primarily from one region or one body type — the tool will perform less well for everyone outside that group.
When you see an AI feature on a brand's website, it is reasonable to ask: whose information made this possible?
2. Fit tools are not equally accurate for all bodies
Virtual try-on and size-recommendation tools are among the most consumer-facing AI applications in fashion. They promise to reduce returns and help you find the right fit without visiting a store. In practice, their accuracy varies considerably depending on body shape, height, and proportions that fall outside the statistical centre of the training data.
Brands rarely publish accuracy figures broken down by body type. A tool that is right for a particular size range most of the time may perform much less reliably for plus sizes, petite frames, or bodies with proportions that differ from the modal customer in the dataset. The result is that the people who have historically had the hardest time finding clothes that fit are often the ones for whom AI fit tools work least well — a pattern that tends not to feature in launch announcements.
3. AI-generated images are not the same as clothes you can buy
Generative AI has made it straightforward to produce images of garments that look finished, wearable, and beautiful. Some brands use these images in early-stage marketing or social content before a physical sample exists — sometimes before it is confirmed that the garment can be made at all.
The gap between a rendered image and a production-ready pattern is significant. Fabric behaves according to physics that image models do not simulate accurately. A collar that looks elegant in a generated image may be structurally impossible, or may require expensive hand-finishing to achieve. When you see an AI-assisted campaign image, there is no reliable way to know whether it represents something you will actually be able to purchase, or something that existed only as pixels.
This is not a hypothetical concern. Trade journalism covering the industry — including Just Style, now part of GlobalData — has noted that trust and readiness are struggling to keep pace with fashion's AI ambitions.
4. The regulatory picture is changing, but disclosure is still thin
In Europe, the EU AI Act introduces transparency obligations for AI systems that interact with people directly. Under Article 50, providers of AI systems must design them so that individuals are informed they are interacting with an AI, unless this is obvious. Deployers must also inform people when emotion recognition or biometric categorisation systems are in use.
In practice, most fashion brands that deploy AI styling assistants or virtual try-on tools do not make these disclosures prominently. A chatbot that recommends outfits may technically comply with the letter of the law while doing very little to help you understand what it is, what data it uses, or how its suggestions are generated.
Regulators are paying closer attention. The US Federal Trade Commission announced a crackdown on deceptive AI claims in September 2024, targeting firms that overstated what their AI tools could do. Fashion is not immune to that scrutiny.
5. Demand forecasting AI can entrench overproduction
One of the most common AI applications in fashion is demand forecasting: using historical sales data and trend signals to predict what to produce and in what quantities. Brands position this as a sustainability win — make less, waste less. The reality is more complicated.
Forecasting models trained on historical data tend to recommend producing more of what sold well before. In a market where consumer preferences shift quickly, that can mean large quantities of items that miss the moment, ending up in landfill or incinerated. The McKinsey State of Fashion report, produced annually in partnership with the Business of Fashion, tracks how brands are navigating AI adoption alongside persistent overproduction challenges — and the picture is not straightforwardly optimistic.
AI forecasting can reduce waste when it is well-designed and well-governed. But a tool is only as good as its objective. If a model is optimised for revenue rather than sustainability, it will recommend producing more, not less.
6. Job displacement is a real and documented concern
Fashion brands that announce AI tools almost never address what those tools mean for the people whose work they automate. Pattern-making assistance, image retouching, customer-service chatbots, and trend analysis all touch roles that real people currently hold — often in parts of the world where alternative employment is scarce.
An International Labour Organization working paper on generative AI and jobs found that women are disproportionately represented in the occupations with the highest exposure to generative AI, with the gap widening in higher-income countries. Fashion's workforce — which skews female, particularly in production and administrative roles — sits squarely in that exposure zone.
This does not mean AI tools should not exist. It means the conversation about them should include the people whose livelihoods they affect, and that brands have a responsibility to be honest about that dimension rather than presenting AI adoption as an uncomplicated good.
What you can do with this information
None of this means you should distrust every AI feature a brand deploys. Some tools genuinely reduce waste, improve fit, or make the shopping experience more useful. The point is that you deserve a more complete account than a press release provides.
When a brand announces an AI tool, a few questions are worth asking: What data trained it, and was that data collected ethically? How does it perform across different body types? Does it replace human roles, and if so, what is the brand doing about that? Is the image I am looking at something I can actually buy?
Brands that can answer those questions clearly are the ones worth trusting.
FAQ
What does the EU AI Act require fashion brands to disclose about their AI tools? Under Article 50, brands must inform users when they are interacting with an AI system and when biometric categorisation or emotion recognition is in use. The rules apply to systems deployed in Europe, but consumer-facing disclosure remains inconsistent in practice.
Why do AI fit tools sometimes work less well for certain body types? Fit tools learn from historical data. If that data over-represents a narrow range of body shapes, the tool will be less accurate for bodies outside that range. Brands rarely publish accuracy breakdowns by demographic.
Can AI-generated fashion images show clothes that do not actually exist? Yes. Generative AI can produce images of garments that have not been sampled, patterned, or confirmed as manufacturable. There is no standard disclosure requirement that distinguishes AI-rendered images from photographs of real samples.
Is AI demand forecasting actually more sustainable? It can be, but only if the model is designed with sustainability as an objective. A forecasting tool optimised for revenue may recommend higher production volumes, not lower ones.
Are fashion workers at risk from AI automation? Research from the International Labour Organization indicates that women — who make up a large share of fashion's workforce — are disproportionately represented in roles with high exposure to generative AI. The industry has been slow to address this publicly.
Further reading
- Transparency obligations under Article 50 of the AI Act
- FTC Announces Crackdown on Deceptive AI Claims and Schemes
- ILO Working Paper: Generative AI and Jobs
- Week in review: Fashion's AI ambitions meet a reality check