Proof, Not Prose: Sustainability Disclosure and the Case for Verified Luxury

I have asked that question in hotels on four continents, first as an operator and now as an assessor. Sometimes there is a meter, a reading log and a clean line to the number. More often, the figure comes from a monthly utility invoice that bundles the laundry, the staff accommodation and the spa; from a waste contractor who weighs on their own scale; or, in island resorts, from fuel deliveries used as a proxy for what the generators actually burned. The number then travels through two or three spreadsheets before it reaches corporate reporting, where it acquires decimals and an authority it never earned.

The framing of this panel is right: a generative layer on top of thin data does not produce evidence. I would go one step further. It produces plausibility, which is more dangerous, because plausibility is exactly what a reader cannot tell apart from proof.

The previous panel asked what is material. Materiality tells you what to measure; provenance tells you whether you actually did. And when this panel examined hotel sustainability audits in 2024, Willy Legrand set aside the reliability of the underlying data as "a topic for another day." EmpCo has now set the date.

Would the data survive scrutiny?

In most hotels, not yet, and dishonesty is rarely the reason. The problem is structural. Sustainability data is usually collected to fill a report, not to run the building, and when data has no operational use, nobody notices when it is wrong.

The industry's own benchmark shows the scale of the issue. When the Cornell Hotel Sustainability Benchmarking Index introduced a waste index in its 2025 edition, only about 4,500 of roughly 9,800 waste submissions passed validation. And these came from companies committed enough to join what Cornell calls the sector's most robust benchmark.

The test I apply is simple, and any owner can use it tomorrow. Take the five figures your hotel quotes most often: energy per occupied room, water per guest night, waste diverted, local procurement, local employment. For each one, ask the team to walk you back to the physical source within an hour. A meter reading, an invoice, a weighbridge ticket, a payroll record, a purchase ledger. If the trail ends in a spreadsheet cell typed by someone who left the property two years ago, the figure is not evidence. It is an opinion with decimals.

I call this the Traceability Test. It requires no new software, only curiosity at the level of the owner and the general manager.

Verification is only half of what this panel describes. The other half is context. The same water figure per guest night means one thing for a city hotel on a municipal network and another for a resort drawing on a stressed aquifer. Data never read against its place cannot tell an owner whether the hotel is restoring or depleting what it depends on.

Is technology generating evidence, or paperwork faster?

PwC's Global Sustainability Reporting Survey 2025, covering 496 companies across sectors, found that AI use in sustainability reporting nearly tripled in a year among companies that had already reported, from 11% to 28%. For hotels specifically, we do not know, because nobody is required to say that a narrative was drafted, or a gap filled, by a model. That absence of data is not merely a research problem. It is a governance problem.

The most useful evidence on the consequences comes from the University of Auckland. Ruth Dimes and Charl de Villiers asked postgraduate students who had studied sustainability reporting and greenwashing to blind-rate AI-drafted sustainability statements against real corporate ones. The students found the AI versions more credible, even though those versions scored higher for greenwashing, and student edits did not improve them. It was a classroom experiment, not a market study. But de Villiers' explanation deserves the attention of every hotel board: the real reports contained real problems they could not smooth over, and the generated ones did not.

That is the point. Evidence has friction. A real hotel year includes a bad July, an irrigation leak, a laundry contract that changed scope halfway through, a desalination plant that underperformed. A disclosure without friction, where every indicator improves smoothly and every gap is neatly closed, should worry an owner rather than reassure one.

None of this argues against AI. Estimation is not the problem; unlabelled estimation is. Nor are all estimates equal. A forecast of next month's energy use is valuable because next month's meter will confirm or correct it. A figure that fills a gap nobody will ever measure can never be contradicted, which is exactly why it should never pass as fact.

A sound discipline is a short provenance note at the front of every disclosure: what share of the figures is measured, calculated or estimated, the method behind each estimate, and where AI was used for drafting or gap-filling. Nobody requires it yet, which is precisely why it would be believed.

The useful question about any technology is its distance from the meter. Tools that sit close to the physical event — such as sub-metering, automated reads, invoice capture linked to source documents and audit trails — make data harder to fake. Tools that sit close to the report mostly make paperwork faster.

The second test is distance from the ledger. Impact data kept in a separate sustainability file stays a reporting exercise. Placed in the same system that runs finance and operations, it is checked every day by people who need it to be right, from the chief engineer to the financial controller. Data becomes reliable when it becomes useful.

Are we still collecting badges?

Hospitality has already seen what happens when the bar moves. In 2024, following action by the Dutch consumer authority ACM, Booking.com withdrew its Travel Sustainable badge, which by 2023 appeared on more than 500,000 properties, and replaced it with a label for third-party certification. At the time, just over 16,500 properties displayed one.

Since 27 September, the EU's EmpCo Directive applies similar logic to environmental claims made to consumers across the EU. Generic claims without recognised excellent environmental performance, sustainability labels not based on a certification scheme or public authority, and neutrality claims based on offsetting are now prohibited in all circumstances. Claims about future performance, such as "net zero by 2035," are misleading unless backed by a detailed implementation plan with measurable, time-bound targets, regularly verified by an independent expert whose findings are public.

EmpCo regulates what a hotel says to guests, not what it files. But a guest-facing claim is only as strong as the data in the file behind it.

The directive will also push the market from self-declared badges towards certification, and that is progress. The useful question about any label, including the ones my own organisation issues, is what exactly it certifies: that a management system exists and has been checked, or that outcomes in the place have been measured, independently audited and can be checked by anyone. Impact disclosure answers what a badge cannot: what changed, by how much, against which baseline, who verified it, and what got worse.

The same is true of awards. Recognition that rewards the best-written submission rather than the best-evidenced performance teaches an industry to write better, not to perform better. Generative AI makes that weakness acute, because writing well has just become almost free.

This matters most in luxury. A guest paying several thousand euros a night is buying, among other things, the assurance that what they are told is true: about the provenance of the fish, the vintage of the wine and, increasingly, the reef in front of the villa and the village behind it. In luxury, credibility is part of the product.

That conviction shaped the definition of regenerative luxury I proposed in a working paper this July: an operational practice carried out "under conditions of independent audit, continuous measurement and intergenerational accountability." Those last words carry the argument. Regeneration that cannot be measured remains an intention; luxury that cannot be verified remains a story. I expect the next definition of ultra-luxury to include a word the industry rarely applies to itself: verified.

There is also an uncomfortable asymmetry. The outcomes that matter most in regenerative terms — the health of a watershed, the livelihoods of a community, the recovery of a habitat — are the hardest to measure and the easiest to narrate. That is precisely where generative prose will arrive first, and where verification matters most.

What should change

Before buying another reporting platform, run the Traceability Test on your most-quoted numbers. Label every estimate. Disclose AI use. Ask what your certificate actually certifies. Publish the bad months.

What a hotel leaves behind cannot be drafted. It has to be measured.