Skip to content
PenzaıHUMAN TECH

The GuideItaly

Answering Reviews with AI: Hotels and Restaurants After the AGCM Guidelines (and Sigeric's Diary)

How to use AI to answer reviews for hotels, B&Bs and restaurants after the AGCM guidelines: method, mistakes to avoid and rules to follow.

The Professor at PenzaiReviewed by Agostino Cacciola4 min read

Titolare di un'attività ospitale di spalle, con il portatile aperto sul bancone, che legge le recensioni dei clienti

A negative review lands at eleven at night, after service, when you are tired and you are right. The temptation is double: answer at once, in anger, or paste it into an AI and publish whatever comes out. Both are mistakes. Here is how to do it well, now that the Italian competition authority has had its say.

An archbishop who kept reviews

In 990 Sigeric, newly named Archbishop of Canterbury, went to Rome to receive the pallium from Pope John XV. On the way back he noted the stages of his journey: a diary recording about eighty stops, from the walls of Rome to the Channel (sources disagree on the exact number, between 79 and 80). According to the Piedmont Region and the Council of Europe, it is the document behind the fame of the Via Francigena.

And you will say: what does that have to do with me? Plenty, because that diary did what a review does today: it told those who came after where to stop, and where it was worth doing so. Whoever ran a hospice on the road lived on word of mouth. The difference: back then word of mouth moved slowly; today it travels at the speed of a phone. Well, your inn today is on that same road. Only now the diary is written by a hundred people at once.

What the AGCM said

On October 19, 2026 the Italian Competition and Market Authority (AGCM) published its Guidelines on online reviews, provided for by Law No. 34 of March 11, 2026 (the annual SME law). They cover restaurants and tourist accommodation in Italy. In short:

  • reviews must be reliable, meaning written by people who actually used or bought the service;
  • whoever publishes them must explain how they are handled, ranked and moderated, and what makes a review "verified";
  • practices that manipulate reviews are to be avoided. Trade press lists, among them, rewards or discounts in exchange for positive ratings, pre-written texts and hidden negative reviews;
  • the guidelines will be reviewed within a year.

This concerns you even if you do not run a platform: it is an invitation to look anew at how you collect and treat reviews. The text is not a manual on AI-written replies: what follows is the Professor's method, not an instruction from the Authority.

Where AI helps (and where it does not)

AI is excellent at drafting and poor at being the host.

TaskWith AIStays with you
Reading and sorting 50 reviewsRecurring themes: breakfast, noise, waitingDeciding what to change
Draft reply to praiseFast, needs a touch-upThe name of the waiter to thank
Draft reply to criticismStructure: thanks, apology, facts, invitationChecking facts and tone
Foreign-language reviewsTranslation and draftFinal read-through
Serious cases (health, accusations)NothingYou, and a lawyer if needed

The method in five steps

  1. Write down your voice. One paragraph: how you speak to guests, what you would never say. You will paste it into every request.
  2. Give the AI only the review text. No guest name, room, stay dates or booking details: they are personal data, and answering in public with them is a mistake. If you use external tools, reread our guide to GDPR and artificial intelligence.
  3. Ask for a draft, not the answer. Then check the facts: if the AI writes "we have already fixed the problem" and it is not true, the lie is yours.
  4. Add a detail only you know. It is what separates a reply from a form letter.
  5. Reply to all of them, the good ones too. A reply to a complaint is read, more than by its author, by the person deciding where to sleep.

Mistakes to avoid

  • Answering everyone the same: guests notice copy-paste.
  • Contradicting the customer's version in public with personal details: keep a sober tone and move the detail to a private channel.
  • Hiding or discouraging unfavorable reviews: the guidelines cite exactly these practices among those to avoid.
  • Offering discounts for positive reviews or for withdrawing negative ones: same thing.
  • Leaving the reply to the AI with nobody rereading it.

On internal rules and tools, what we wrote in ChatGPT at work: rules for employees applies: whoever replies must know what can be pasted and what cannot.

What to do on Monday morning

  • Write your business's voice in ten lines and keep it in a file, ready.
  • Read the last twenty reviews and mark the three themes that recur: they are your real work plan.
  • Decide who replies, at what hours, and who rereads drafts before publishing.

Sigeric did not manage to change the roads: he noted them down, and those who came after benefited. Reviews work the same way. If you want to bring order to your flow of reviews and replies, write to Penzai for a free introductory call: we will talk it through calmly, with no commitment. You can also see our work with businesses on the page for entrepreneurs and SMEs.

Was this useful?
Buy us a coffee

Penzai Magazine

The day's stories, every evening in your inbox

Briefings, guides and analyses from the Penzai Professor: a preview and link for each article, one email a day. Free gift: the AI checklist for SMEs. Unsubscribe anytime.

Prefer Telegram? Get every article as soon as it is out Follow on Telegram

Comments

No comments yet. Be the first.

All articles
Portrait of the smiling Penzai mascot

Let's talk about your next project.

A focused consultation to understand how technology can really help you. First we understand your work, then we find the right technology.