Generative AI in Hospitality: 7 Mistakes to Avoid (and How to Fix Them)

The most serious mistake with generative AI in hospitality has nothing to do with technology: it's pasting guest data, names, room numbers, card details, into a public tool, then publishing what it produces without review. Rule number one is simple: no identifiable personal data should ever go into a consumer AI. Here are the 7 mistakes I see most often in the field, with a concrete fix for each.

My name is Tiffany Weltman and I train hotel and restaurant teams to use generative AI. Accor, Paris Society, Airelles and the Plaza Athénée are among my clients. My training organization is Qualiopi-certified, so the courses can be funded through your OPCO, and I have already trained more than 2,500 hospitality professionals. These seven mistakes come up in almost every property, and the good news is that they can all be avoided with a few simple habits.

The mistakes that expose your data (the most serious)

Mistake #1: pasting guest data into a public AI

This is the most dangerous habit, and yet the most common. To draft a review reply, an apology or a confirmation, people copy-paste the guest's message as-is, with their name, email, sometimes their booking or card number. In a consumer AI, that information can be stored and reused to train the model. You lose control over data you are legally required to protect.

The fix: always anonymize before you write your prompt. Replace the name with "the guest", the email with "[email]", the room number with "[room]". The AI doesn't need the real identity to produce quality text.

You are a receptionist at a 5-star hotel. Write a warm, professional reply to this guest review. The guest is called "[GUEST]"; never reuse any real personal data. Here is the review: […]

Mistake #2: ignoring GDPR and consent

Many teams think "AI is just another tool". But the moment a process involves personal data, GDPR applies. Using AI to analyze guest profiles, segment a database or generate personalized messages without a legal basis or informing the individuals exposes you to real legal and reputational risk.

The fix: favor the professional versions of these tools, which offer contractual guarantees (no training on your data). Keep a simple log of use cases, never enter identifiable data, and train your teams to tell the difference between personal and anonymized data. When in doubt, your DPO or compliance lead decides.

The mistakes that make your content ring false

Mistake #3: publishing without human review

AI writes fast and well… on the surface. It can also invent a rate, an opening time, a service that doesn't exist, or slip in a tone-deaf phrasing. Publishing a post, sending a guest email or printing a menu without proofreading means risking spreading an error at scale.

The fix: always keep a human in the loop. AI proposes, a professional validates. Set the rule "no AI output published without review", especially for anything touching guests or numbers.

Mistake #4: settling for generic, soulless text

"Our establishment offers you a unique experience combining comfort and elegance…": you spot that kind of sentence instantly. An AI left without guidance produces smooth, interchangeable content that could describe any hotel in the world. It's the opposite of what sells hospitality: the concrete and the emotional.

The fix: feed the AI real details, your chef's name, the view from the terrace, the building's history, an anecdote. The more specific and embodied your brief, the more distinctive the result.

Mistake #5: not giving the AI your brand voice

A palace doesn't speak like a seaside boutique hotel. Yet without a tone instruction, AI always writes the same way. The result: content that betrays your positioning and dilutes your identity.

The fix: create a "brand voice sheet" once and for all, tone, vocabulary, what you do and never say, and paste it at the start of every prompt.

Here is our brand voice: warm yet refined tone, short sentences, never hollow superlatives, formal address, a touch of subtle humour. From now on, write all my texts in this style. Confirm you've understood before you start.

The mistake about what AI really knows

Mistake #6: assuming the AI knows your prices, menu and hours

This is a frequent misunderstanding: the AI knows neither your evening rate, nor your daily menu, nor your spa hours. When it doesn't know, it invents a plausible answer, this is called a "hallucination". A guest told a wrong price or service is a guaranteed disappointment and sometimes a dispute.

The fix: never ask for factual information "from memory". Provide your real data in the prompt (rates, menu, hours) and ask the AI to stick to it strictly.

Here is our up-to-date menu and rates: […]. Answer the guest's question based ONLY on this information. If the answer isn't there, say so instead of inventing.

The process mistake that cancels out all the others

Mistake #7: one-off training with no follow-up

You run a two-hour session, everyone is enthusiastic… then, for lack of practice and support, the habits fade within a few weeks. AI evolves fast, teams turn over, and the tool falls back into a drawer. It's the mistake that ruins the return on investment of every other good practice.

The fix: think of upskilling over time: an initial training, then regular check-ins, a shared prompt library and an AI champion per department. I detail this approach in my guide to training your teams on generative AI.

Where to start: your anti-mistake checklist

Before every use of AI, keep these habits in mind:

• I anonymize any guest data before sending it to the tool.
• I proofread systematically before publishing or sending.
• I give my brand voice and concrete details in every prompt.
• I provide my real information (prices, menu, hours) rather than trusting the AI's memory.
• I check GDPR compliance and use professional versions whenever possible.
• I build training over time, with follow-up and champions.

These seven mistakes have one thing in common: they don't come from the tool, but from how it's used. That's exactly what I pass on to your teams, with concrete exercises drawn from your own property. Discover my AI training programs for hotels and restaurants, fundable through your OPCO.