The Essential View: You don’t need to become a supply-chain investigator before buying a luxury product. But if you’re about to spend thousands of dollars or euros, five minutes of basic due diligence is reasonable. Generative AI makes that easier: give it the right prompt and it can quickly report on a maison’s supplier disclosures, audits, human-rights record, sustainability claims, regulatory history, and independent reporting. It won’t prove that a company is clean, but it can help you judge whether its story deserves greater confidence—or greater scrutiny.
Image: PeopleImages / Shutterstock
In this week’s feature article, I wrote about the recent scandals in Italian luxury goods production and what they mean for us as consumers. I wrote:
Does that mean that the next time we visit our favorite luxury boutique, we should take our consumer muckraker checklist with us and start peppering the staff with questions? No. The average salesperson wouldn’t be able to answer our questions, and I suspect they’re not trained to do so, at least not in a meaningful fashion.
It does mean, though, that we can learn to distinguish transparency from reassurance. Before spending thousands of dollars or euros, spend a few minutes searching the brand’s name and its public record in this domain.
I had a simple web search in mind, but then I realized this was an obvious use for generative AI.
Using generative AI for quick maison research
As a test, I gave the following prompt to ChatGPT, using Hermès as the subject:
Provide me with a 200-word report on Hermès and their human rights and sustainability practices. How confident should I feel that the firm isn't engaged in practices similar to those that firms such as Dior, Loro Piana, and Tod's have been accused of in Italy?
I chose 200 words—roughly a minute of reading. And this is what came back:
Hermès has a stronger risk profile than many luxury peers, but not a clean bill of health. Its structural advantage is control: 55% of products are made in exclusive workshops, 75% in France, and suppliers are predominantly European. Its supplier code requires compliance with ILO standards and disclosure and monitoring of subcontractors. Hermès reports independent audits; its 2025 statement records more than 700 audits, including over 100 each at tiers one and two.
Environmentally, Hermès has SBTi-validated 2030 targets and reports reductions since 2018 of 68.9% in absolute Scope 1–2 emissions and 59.2% in Scope 3 emissions intensity; purchased electricity is renewable. Ninety percent of tier-one leather suppliers hold Leather Working Group certification. Durability and repair are genuine strengths, although extensive leather and exotic-skin use creates biodiversity, animal-welfare, and traceability risks.
My judgment: moderate-to-high confidence—perhaps 70–80%—that Hermès is not presently replicating the systematic Italian abuses alleged at Dior, Loro Piana, and Tod’s. I found no comparable public allegation involving Hermès. Nevertheless, confidence should stop short of “high”: audits can miss concealed subcontracting, Hermès does not publish a complete factory-level supplier list, and most evidence comes from Hermès itself. It appears lower-risk, not independently proven abuse-free.
Counting starting up ChatGPT, writing the prompt, letting it do its research, and reading the result, that was less than five minutes.
At first glance, the result seemed useful. On closer inspection, however, it offered only the reassuring half of some of the evidence. Hermès did report more than 700 supplier audits in 2025. But the same report says those audits found working-hours violations, inadequate overtime compensation, and undeclared subcontracting. The answer omitted those findings, supplied no links, and converted an incomplete public record into a suspiciously exact 70–80% confidence estimate.
Back to the drawing board.
The copy-and-paste prompt
I worked through several more rounds with ChatGPT, refining the prompt to produce answers that would be specific, grounded in evidence, and no more definitive than warranted by the evidence. Here’s the final version. Copy and paste it into a web-connected generative AI tool with search enabled, then fill in the luxury house and desired reading time:
Using live web search—not model memory—prepare a current, evidence-based pre-purchase ethics and sustainability check on [luxury house]. Write approximately [n] minute(s)’ reading at 200 words per minute aimed at an ordinary luxury shopper. Use plain English, short paragraphs, and only numbers that materially affect the purchasing decision. Explain unavoidable technical terms. Investigate the house, its parent company, and relevant subsidiaries. Search relevant local-language sources where practical. Do not infer ethical conduct from European production, proximity to headquarters, or the absence of public allegations. Determine:
• Who makes its products, whether it identifies suppliers and factories, and how it controls unauthorized subcontracting.
• Whether it audits suppliers, whether independent auditors are involved, what problems the audits found, and what was corrected. Treat audits and certifications as safeguards, not proof.
• What protections exist against forced labor, child labor, unsafe conditions, and other human-rights abuses; whether workers can report problems; and whether outcomes or remedies are disclosed.
• Whether credible investigations, court cases, regulatory actions, or allegations involve the company or its suppliers.
• Whether its environmental impact is genuinely declining overall—not merely per product or per dollar of revenue. Consider materials traceability, water and chemicals, biodiversity, animal welfare, renewable energy, repair, reuse, and circularity.
Structure the answer as:
Bottom line
What reassures
What gives pause
What remains unknown
Strength of public evidence: very weak, weak, mixed, moderately strong, or strong.
Clearly label company claims and independent findings. If information is not disclosed, say so. Distinguish “I found no allegation” from evidence that abuse is absent. Give a direct working link and publication date for every important factual claim. Do not invent, reconstruct, or repair citations. If live browsing or source verification is unavailable, do not complete the assessment.
The bottom line
To be clear, a report produced by the copy-and-paste prompt above isn’t a certificate of good conduct. Generative AI can miss relevant data, misinterpret evidence, or even invent facts, events, or citations. That’s why the prompt asks for direct links: if something would materially affect your decision, open the source and check it yourself.
Five minutes won’t tell you everything about the people and supply chains behind a luxury product. But five minutes can tell you whether there are reasons to look more closely—and whether the story the maison tells about itself is reasonably consistent with the available evidence. In other words:
If luxury asks us to pay a premium for provenance, spending five minutes checking that provenance doesn’t seem like too much to ask.
We do not accept consideration of any kind in exchange for product or service reviews.
