How to use ChatGPT to answer hotel guest e-mails

I run hotels and I write software, so I have tested this myself: a general AI chatbot can write a friendly, well-structured reply to a guest in a few seconds. The catch is that it only knows what you tell it. This guide shows how to get good drafts from ChatGPT (the same approach works in Claude and other chatbots), how to check them, and where the copy-paste approach stops being worth it.

What you need before you start

The single biggest improvement you can make is to stop asking the chatbot to guess. Collect the facts your guests ask about in one plain document — a hotel fact sheet:

• Address, how to get there, parking and its price
• Check-in and check-out times, early check-in and late checkout rules
• Room types, what is included, breakfast times and price
• Cancellation and payment policy, deposits, children and extra beds, pets
• Who handles groups, events and complaints — and how to reach them
• Your sign-off: the name and signature you use

Keep it short and factual. If it is not in the fact sheet, the chatbot should not answer it.

The prompt template

Paste this at the start of a new chat, fill in the brackets, and keep the chat for the day so you only add the guest e-mails. In ChatGPT you can also save the first part as custom instructions or in a project so you do not paste it every time.

The [CHECK: …] rule matters most. It turns "I don't know" into a visible gap instead of an invented answer, and you will spot it before you press send.

Example: a draft based on the prompt

An illustrative inquiry and the kind of draft you get when the fact sheet and availability are pasted in. The reply depends entirely on those facts — without them, the chatbot cannot know that two rooms are free.

How to check the draft (60-second checklist)

1. Dates and weekdays — are they the guest's dates, and do the weekdays match?
2. Prices — do they match your booking system for those dates and that room?
3. Availability — is the room still free right now?
4. Policies — does the reply promise anything your policy does not allow (free cancellation, early check-in, a discount)?
5. All questions answered — guests often ask two or three things in one e-mail.
6. No [CHECK] left and no placeholder text.
7. Tone and length — would you sign it yourself?

The honest limits of copy-paste AI

ChatGPT is a fine start, especially for a small hotel with a few e-mails a day. But it is worth knowing where it falls short before you build your routine around it:

• No live availability or prices. The chatbot is not connected to your booking engine, PMS or channel manager. Every number has to be pasted in and checked.
• Copy-paste for every e-mail. Open the e-mail, copy, switch window, paste, wait, copy back, check, send. For a few messages that is fine; for dozens it becomes a job in itself.
• Confident wrong answers. Language models can state a policy, a time or a price that sounds right but is not yours — and the text reads just as convincingly as a correct answer. The [CHECK] rule reduces this but does not remove it.
• Inconsistency between staff. If each person uses their own prompt and their own chat, guests get different answers to the same question.
• Privacy. Guest e-mails contain personal data. Paste only what you need to write the reply, and check your account's data settings and terms (see below).
• No 24/7 coverage. The chatbot only works when someone is there to use it. Inquiries that arrive at night or during a busy check-in still wait — and a waiting guest often books elsewhere. See how fast hotels should reply.

Privacy: what to paste and what to leave out

You are responsible for the guest data you share with any outside service. Practical rules that cost nothing:

• Remove phone numbers, addresses, card details, passport details and booking references before pasting. The chatbot rarely needs them to write the reply.
• Use the first name only, or "the guest".
• Check your plan's data settings. Both OpenAI and Anthropic let users control whether their chats are used to improve the models, and business plans come with different terms from personal accounts.
• If staff use the chatbot for hotel e-mails, decide which account and settings they should use — not everyone's private account.

Sources: OpenAI Help Center, Data Controls FAQ, retrieved 2026-10-07; Anthropic Privacy Center, "Is my data used for model training?", retrieved 2026-10-07

Manual vs ChatGPT vs a dedicated AI assistant

When a dedicated tool makes sense

A dedicated AI e-mail assistant does the same writing work but removes the steps around it. BookingWatch, which we built for our own hotels, connects to the reservation inbox, uses the hotel's own facts and rules, answers in seconds day and night, and hands a case over to staff when it needs a human decision. If you want to see what that would be worth for you, estimate it with the lost revenue calculator; prices are on the pricing page. For what to look for in any such tool, read AI e-mail assistant for hotels.

Signs it is time to switch:

• Copy-pasting into the chatbot has become a daily routine of more than a few e-mails.
• Inquiries arrive in the evening and wait until the morning.
• Guests get different answers depending on who is on shift.
• You have caught the chatbot inventing a policy more than once.

Common mistakes

• Asking "write a reply to this guest" with no hotel facts — and getting a generic or invented answer.
• Sending the draft without checking dates, prices and availability.
• Letting the chatbot confirm bookings, discounts or exceptions you have not approved.
• Pasting the whole e-mail thread, including phone numbers, addresses and booking references.
• Using a private account with default settings for business e-mails.
• Long, flowery replies. Guests want a clear answer and a next step.