AI-generated data: evidence, or fluff at scale?

It was about time! The EU’s tighter rules on environmental claims give hospitality a reason to examine what sits behind the word “green”.

For me, the real test is whether a hotel makes life better for the people and place around it. Lower energy use matters. So do fair work, local spending and the health of nearby ecosystems. Calling an activity regenerative, however, does not make its benefits real.

AI sharpens this challenge. It can connect information from across a hotel and reveal where action is needed. It can also turn an untested estimate into a remarkably confident sentence. A figure with two decimal places may look precise while telling us very little. However thereis also a dark side to AI (see my longer article).

I would start with a simple discipline: for every public claim, keep the original records, state the starting point and explain what changed. Show estimates as estimates. Have significant claims checked independently. Then make the results understandable enough for guests and communities to question them.

That is how we move from attractive promises to something more useful: evidence that hospitality is doing more good.

With EmpCo but without Green Claims Directive our industry finds itself in an uncomfortable position: the obligation is here, the instruction manual is not.

For most hotels, the data behind disclosures would not yet survive serious scrutiny — not because of bad intent, but because we built sustainability reporting as a communications exercise before we built it as a data exercise. Technology can help, but it does not replace real hotel input. If the underlying data is thin, AI just produces better-formatted uncertainty. The goal should be to help hotels improve performance, not to help corporate teams produce reports faster.

We also need more honesty about what counts as sustainability. Bike rentals, EV charging stations, refillable bottles — these are guest service attributes, not sustainability achievements. Labeling them as such inflates narratives and dilutes credibility.

Real impact disclosure means reporting what changed, by how much, verified by whom, and whether progress is on track against a credible target. Data should serve performance improvement and guest confidence, not just compliance. The industry needs less narrative and more verified evidence.

Credible sustainability reporting starts with integrity: clear definitions, traceable sources and transparent methods. AI can help organise and analyse information, but measured data, estimates and projections must remain clearly distinguished. Claims and badges need evidence, with geographical and operational scope clearly explained. Social outcomes deserve the same scrutiny as environmental performance. The goal is to demonstrate real improvement for people and places.

In my recent contribution to Hospitality Net, I argued that sustainability should move beyond reporting and become a leadership capability. The debate about AI reinforces that conviction.

AI is not the problem. Confusing generated knowledge with verified evidence is.

Hospitality already had a sustainability data problem before generative AI arrived: fragmented information, different methodologies and, sometimes, limited verification. AI did not create these weaknesses, but it can amplify them—and make weak evidence look remarkably convincing.

We therefore need to distinguish between data, evidence and narrative. AI can analyse data, identify gaps, estimate missing information and communicate results. But an estimate must remain an estimate. It should never become “evidence” simply because AI presents it convincingly.

The question is not whether AI can produce a sustainability report. It clearly can. The question is whether the organisation can demonstrate what sits behind every material claim in that report.

Technology can generate the narrative. Responsibility for the evidence remains with us.

Perhaps this is the opportunity to change the sequence in sustainability reporting:

Evidence first. Communication second.

Because if sustainability is ultimately about deciding impact, leadership must also decide what evidence is strong enough to stand behind it.

Personally, I am pleased to see the direction of EmpCo. It encourages a shift away from vague promises and towards transparent, verifiable performance. In the long run, that is better for consumers, better for businesses, and better for credibility. The era of sustainability-by-marketing is ending. The era of sustainability-by-evidence is beginning.

Less storytelling. More evidence. Less greenwashing. More accountability. That feels like progress.

Do we need a Green Claims Directive to understand if the claims are verifiable or not? Probably not. But would it help? Certainly.

What we really need instead? An integrated strategy for data collection, and AI that is only allowed to work within the designated network to help with data interpretation and best practices, but not with data creation.

Would the data survive scrutiny? Honestly, if it lives in spreadsheets with no audit trail, fuzzy boundaries and no link back to documented sources, probably not. The real test is whether a third party can follow a claim back to the evidence behind it. That comes down to traceability, a clear methodology, defined boundaries and being upfront about uncertainty.

Evidence or faster paperwork? For me it depends on where AI sits in the process. If it helps clean, reconcile, cross-check and flag inconsistencies in measured data, it can strengthen the evidence. If it drafts narratives or fills gaps with estimates nobody labels, we are just producing paperwork faster. What decides it is whether the underlying data is sound and whether we are open about AI's role.

Badges or real impact disclosure? Badges probably are not going away, but they should rest on clear criteria, documented evidence and a defined scope. But how can a guest tell a rigorous scheme from a weak one? To me, a badge is only as good as the evidence behind it.

In the end, it comes down to one question: would the evidence behind a claim hold up if someone checked?

Using AI tools  - powered by renewable energy net-zero data centers – to accurately disclose a hotel’s progress towards science-based climate related targets seems like a worthy goal, and use of AI technology.

Efforts are underway to agree on the metrics and methodologies which should be adopted by the industry - like USALI was for financial disclosures. These will be imperfect and require regular revisions but should move us towards our collective sustainability goals. AI systems can help companies to collate data, spot anomalies, and produce reports faster with that data. The reliability of the data will depend on the quality of the measurement and reporting systems across the supply chain which need to be carefully monitored.

Hotel owners should measure, disclose, and act on their sustainability metrics, climate and biodiversity endangering or mitigating activities, and understand their exposure to climate risks, it is good business.  Policy makers must accept the metrics, accept the targets and penalties for non-compliance must be proportionate. EMPCO adopted on 27th September is a step in the right direction to help protect European consumers and investors from spurious sustainability claims, which must now be backed up with certified data, but more is needed.

As I argued in the recent Lemongrass Travel Trend Report 2027, drawing from a long bike ride from Italy to Nordkapp this summer: the algorithm has the road, but people have the conditions.

And sustainability reporting has the same split (template versus evidence).

AI has made sustainability text abundant and cheap, but to be clear, under EmpCo, I doubt that anybody will ask how well-written a claim is...they will only ask what sits behind it.

My take:

Would the data survive scrutiny? Energy and water would, because they come from meters and invoices. I would argue that Scope 3 and anything "estimated" rarely would.

Evidence or paperwork? Probably faster paperwork because AI would pull meter data, reconcile invoices and map data to framework. It is what has already happened in the hotel. If AI starts inventing and modelling data nobody can trace, that is problematic and honestly, that should not go public. I would argue, similar to what I say to students: Simply state where and how AI touched the final report.

Badges or impact? High time to move to impact disclosure, showing what is changing where and using which baseline.

AI can sharpen evidence. It shouldn't create it.

Our industry is scrambling for sustainability badges. The new EU rules have made a logo feel like a get-out-of-jail card, and AI offers the same shortcut: polished prose that reads like proof.

For The Long Run, it is not whether a disclosure reads well, but whether it traces back to something real. Feed AI thin data and you get confident paperwork, only faster. Our own review of members' self-reported figures found typing errors and double-counting. This was data collected through a rigorous, albeit imperfect process. AI flagged the red flags; we ratified the numbers with team members who know their properties intimately.

AI earns its place analysing information already held, spotting gaps and reusing what members have published. It should never estimate what was never measured.

After 16 years supporting impact-led tourism, we know policy and implementation are two different things. Every Long Run property receives an on-site visit, certified or not, and we are bringing that transparency forward to support our claims.

In an AI-led world, anyone can produce a convincing disclosure. What sets one apart is transparency: where the data came from, who has seen it, and where AI helped.

I have been exploring the topic of “AI and Nature/Biodiversity” for several years.

In this context, it is safe to say that there are AI applications that help monitor wildlife populations, resolve or at least mitigate human-wildlife conflicts, and make data collected by tourists available for scientific analysis.

In this way, AI can not only contribute to nature conservation but also help protect a key asset of the hospitality sector. And this opens up new opportunities for tourists, who see that guests’ experiences of nature help preserve exactly what may have been the reason for their trip in the first place.

This triad—operator, guest, and science—may hold a treasure that has yet to be uncovered.

Of course, amid all this, we must not forget the significant negative impacts that the use of AI has on nature. Its immense consumption of water, energy, and resources harms nature. However, this does not mean that we cannot and should not use AI wisely; rather, it means that we must first learn how to use it properly.

Evidence has friction.

Would the data survive scrutiny? Often not, and dishonesty is rarely why. Hotels collect sustainability data to fill reports, not to run buildings. When Cornell added a waste index to its 2025 benchmark, only about 4,500 of 9,800 submissions passed validation. My test: trace any figure, within an hour, to a meter, an invoice or a weighbridge ticket. If the trail ends in a spreadsheet cell, it is an opinion with decimals.

Evidence or faster paperwork? It depends on distance from the meter. Technology close to the physical event creates evidence; technology close to the report creates paperwork. In a University of Auckland experiment, informed postgraduate readers rated AI-drafted sustainability statements more credible than real ones, although the AI versions scored higher for greenwashing. Real reports carried real problems. A disclosure without a bad month should worry an owner.

Badges or impact? When Booking.com replaced its own badge, shown on over 500,000 properties, with third-party certification, just over 16,500 qualified. EmpCo now pushes the EU market the same way. The next question is what a label certifies: a management process, or verified outcomes in the place. Ultra-luxury, which sells trust, will be judged on the second.

AI-generated data: evidence, or fluff at scale?

As EmpCo begins to apply, hospitality faces a simple but important question: how do we turn sustainability ambition into credible, verifiable disclosure? The Green Claims Directive may be paused, but consumer expectations and regulatory direction are not. This is an opportunity for our sector to strengthen trust.

AI can help us analyse, summarise and communicate faster. But it cannot replace measurement, assurance or accountability. Its value depends on the quality of the data and systems beneath it. That is why the Alliance is working with members to build harmonised metrics, practical tools and assurance-ready approaches that make credible disclosure easier.

Our members are already demonstrating leadership, and we want to support every stage of that journey. No one is expected to have perfect data overnight. What matters is steady progress, transparency and a shared commitment to real impact.

Let’s use AI to accelerate understanding, not to manufacture claims. Together, we can ensure hospitality’s sustainability story is grounded in evidence, not fluff. That is the standard we are building together, and it is within reach.

EmpCo targets general sustainability claims without evidence, so stop making such claims, be specific and back it up with evidence.

Working with hundreds of hotels and sifting through the piles of evidence they submit, it is inspiring to see the positive impacts that many hotels are having on their community and environment. Yet many hotels and travel partners are getting anxiety around EmpCo. Teams need clarity on what counts as evidence. My advice is to stop claiming a level of sustainability and simply describe what you're actually doing. And only talk about what you can backup with evidence.

Supporting local communities? Tell us how. Document your contributions and show what changed before claiming an impact.

Sourcing local materials? Tell us which materials, where they come from and how you define “local,” backed by purchasing and supplier records.