A travel agency answers the same questions on every package. Is the flight included? Which hotel on day three? What do I pack for Ladakh in May? Which documents do you need from me? What if I cancel a month before? Nearly all of it is on the itinerary page or in the PDF you send after an enquiry, and most of these questions arrive in the evening, when the office is closed.
An AI chatbot that reads your own pages can answer them. Travel also has three questions a web page can never answer safely: today's price, whether a seat is free, and whether someone will get a visa. This post covers both sides. The TailGPT steps and plan figures are as of 4 October 2026.
Which traveller questions can a travel chatbot answer?
It can answer any question whose answer is written on your package, itinerary and policy pages, or in an itinerary PDF you upload. TailGPT answers only from the pages you tick and the documents you add, with no general travel knowledge, so the quality of each answer is the quality of the page behind it.
| What travellers ask | Where it usually lives | How to write it |
|---|---|---|
| What is included and what is not | Package page | Two plain lists, "Included" and "Not included", item by item |
| The day-by-day plan | Itinerary page or PDF | Each day: places, travel time, activities, where you stay that night |
| Hotels and meals | Itinerary | Hotel name or category, "or similar" if that is the deal, which meals on which days |
| What to pack | Packing list page or PDF | By season and by trip, especially for treks and cold places |
| Documents you need from the traveller | Documents page | The list as you publish it, and the official source for visa rules, by name |
| Payment and cancellation | Terms page | Deposit, when the balance is due, cancellation charges by date, in words |
| Group size and departures | Package page | Minimum and maximum group size, fixed departure dates as published |
| Age and fitness | Package page | Limits for treks and adventure activities, as you set them |
The pattern holds for very different businesses. A houseboat-and-hills operator in Kochi gets asked whether the houseboat has air conditioning at night. A trekking company in Manali gets asked about altitude, porters and what boots to bring. A desert safari operator in Dubai gets asked about pick-up points and dress. A walking-tour company in Edinburgh gets asked whether tours run in the rain. Each answer is one clear sentence on the right page.
Follow-ups work too. A traveller who asks about the eight-day Rajasthan tour and then types "and is breakfast included on day four?" gets an answer about that tour, because the chatbot is given the last three exchanges of the conversation.
What must a travel chatbot never answer?
It must never quote a live price, say whether seats or rooms are free, confirm a booking, or advise anyone on whether they will get a visa. All four depend on facts that change daily or on a decision someone else makes, and none of them can be read off a web page.
- Live prices. Fares and hotel rates move every day. If your package page shows a "starting from" price, the chatbot will quote it as written, with whatever conditions you put beside it. So put them there: per person, on twin sharing, for which departures, and that the final price is confirmed at booking. Asked "how much for four of us on 12 December?", it will say your pages do not cover that rather than invent a figure.
- Seats and rooms. The chatbot cannot see your booking system. TailGPT does no booking or availability lookups. Avoid writing "only 2 seats left" on a page: once read, it becomes a stale fact the chatbot repeats.
- Bookings and payments. TailGPT cannot take an order, collect details or pass the visitor to a person. It can explain how booking works and name the page or phone number to use, if that is written on a page it reads. Answers are plain text with no clickable links, so it names pages in words, such as "the Enquire page".
- Visa advice. Visa rules are set by governments, change without notice and depend on the traveller's passport and history. Publish the documents you ask clients for, exactly as you publish them today, and on the same page name the official source travellers should check, such as the destination country's official visa website or embassy. The chatbot will repeat that list and name that source. Questions such as "will my visa be approved after a past refusal?" are not on your pages, and the chatbot will say so.
Two TailGPT behaviours back this up. If an answer cites something that was not among the pages it read, the answer is thrown away. And a phone number or email address that appears on none of your pages is removed, while your own published contact details pass. The role you choose for the chatbot changes its manner, never what it is allowed to say.
Before it goes live, use the Try demo button to ask five questions: a price for a specific date, "are there seats on the 14 November departure?", "please confirm my booking", "will I get a Schengen visa?" and "what do I get back if I cancel 20 days before?". None of the first four should produce a price, a seat, a confirmation or a verdict. The fifth should quote your cancellation terms exactly. If any answer does otherwise, fix the page it came from.
Watch older blog posts as well. A 2022 article saying a country offered visa on arrival is a page like any other, and a chatbot that reads it will quote it. Untick destination articles that carry rules which may have changed.
How do you use itinerary PDFs?
Upload them as documents. TailGPT takes PDF, Word (.docx) and text files, with no limit on how many, and documents do not count against your page allowance. That suits travel well: many agencies keep the detailed day plan, the packing list and the terms in PDFs they email after an enquiry.
- Text, not pictures. A brochure designed and exported as a PDF usually has real text; try selecting a sentence. A brochure made of images has none, and TailGPT marks it failed with a clear reason rather than reading nothing.
- One package per file, named in its first line. "Rajasthan Heritage, 8 days, October to March departures" at the top helps the chatbot keep the eight-day and six-day versions apart.
- Keep the file name stable. TailGPT treats the name as the document's identity. Upload a corrected
rajasthan-heritage-8-days.pdfand it replaces the old one within seconds. A different name is a second document, so remove the old one. - Remove what has expired. A withdrawn early-bird offer in an old PDF answers as confidently as a current one.
Preparing documents is covered in more depth in how to train a chatbot on your PDFs and documents. Uploading needs a plan: the free trial reads up to 500 pages of one website and gives 20 test questions in the dashboard, without documents.
How do you keep seasonal packages current?
Change the page or the PDF first, then have TailGPT read it, because the chatbot keeps quoting the old version until you do. There is no automatic refresh.
- Before each season, publish the new packages and dates, untick last season's pages (that takes effect at the next read), then press "Read the N selected pages" once on the Websites page.
- Batch your re-reads. As of 4 October 2026, Starter includes 2 re-reads a month and Growth 8, shared across all your websites. Collect a round of edits, then read once.
- Use documents for what changes often. A departures list or a terms PDF replaced under the same name uses no re-read at all.
- Check new pages are listed. A package page published after you added the website may not appear in the page picker. If it is missing, upload its itinerary as a PDF instead.
The full routine is in how to keep your chatbot's answers up to date.
How should a travel agency set up the chatbot?
Add your website, tick the package, itinerary, terms, documents and contact pages, upload the itinerary PDFs, choose a role, test, then paste one line of code before </body>. The whole path is on how it works. A few choices are specific to travel:
- Role. Sales suits package pages: it leads with what is included and points to the next step your site already has. Support suits travellers who have booked and want the packing list or the payment schedule. Two chatbots can share one website's content, each with its own role and look; running separate sales and support chatbots shows how. As of 4 October 2026, Starter includes one chatbot and Growth three.
- Greeting. Say what it can help with: "Ask about any of our packages: what is included, the day plan, what to pack." Visitors then ask the questions it can answer.
- Hindi. In beta, a traveller can ask in Hindi and get an answer from your English pages. Test a few with Try demo first.
How many questions will a travel season need?
Plan for your busiest month: summer holidays, the Diwali break, December, or the weeks before a popular fixed departure. As of 4 October 2026, Starter is ₹4,000 a month with 4,000 questions and Growth is ₹8,000 a month with 10,000, shared by every chatbot on the account. TailGPT emails you at 80% and at 100%, and at the limit the bubble disappears until the month resets, you upgrade, or you buy a top-up of questions. Upgrading takes effect at once, and top-up questions never expire.
A question your pages do not cover uses one of the month's questions; only refused ones are free: attempts to give the chatbot orders, requests to write code, and too many questions too fast. So the "how much for these dates" questions count, which is one more reason to write your price conditions on the page. See the pricing page for every plan.
How do you see what travellers are asking?
Open Show chats on the chatbot. It lists every conversation as chat bubbles, which you can search for words like "visa", "refund" or a destination, and filter to the last 7, 30 or 90 days. The Home dashboard also lists "Questions your chatbot could not answer", with how many times each was asked. The top of that list after a new season usually names a missing fact, such as whether the Kashmir package includes a shikara ride. Add it to the package page, read the site again, and the next traveller gets the answer.