AI Travel Recommendations: Will Your Best Trip Make the List?

AI assistants are becoming a filter between your trips and your travellers. In a new Skift Research brief, ChatGPT, Gemini and Claude were tested across six traveller personas and 21 destinations, and 29% of pairings that were a strong fit were never recommended. The best-known destinations were the most exposed. The sample was small, so read the numbers as direction rather than law. The practical answer for an agency is still clear: describe your signature trips specifically and consistently, on your own website and on every listing, so an AI assistant can find them and repeat them accurately.

What did Skift Research actually find?

The brief is The Disappearing Destination: AI as a New Filter in Discovery, published by Skift Research in October 2026. The researchers posed as six different traveller personas and asked ChatGPT, Gemini and Claude for destination recommendations across 21 destinations, repeating the prompts to check consistency. Three findings matter for an agency:

• 126 persona-destination pairings were a strong fit on paper. 29% of them were never recommended by any model.
• The gap was widest for the famous names: marquee destinations went unrecommended 60% of the time, against 42% for mid-tier and 38% for emerging ones.
• The models appeared to draw heavily on OTA and tour-operator listings when deciding what to say about a place.

Two related findings from the same brief. AI already accounts for 21% of the touchpoints where travellers get inspiration, even though only 6% call it their main discovery platform. And in data cited in the brief, vacation rental sites with 40 or more in-depth local guides were cited by AI 60% of the time, against 32% for sites with almost none.

One caveat, stated plainly: the persona test was a small sample. Treat it as an early signal, not a settled measurement.

Why would an AI assistant skip a destination that fits?

A search engine gives a traveller hundreds of links. An AI assistant gives a shortlist with reasons attached. That changes what it means to be visible. There is less shelf space, and the assistant can only recommend what it can describe.

If the public web says nothing specific about a trip, the assistant has nothing to work with. "Best of Kerala, 5 nights" looks like ten thousand other listings. A described route, a named season, a clear sense of who the trip suits: that is material a model can pick up and repeat. The destinations and operators that vanish are not always the weak ones. They are the ones the models cannot say anything definite about.

What does this mean for an Indian travel agency or DMC?

Your edge has always been specific product knowledge. The exact ghat where the evening aarti is least crowded. The safari gate that pairs with the right lodge. The houseboat route that avoids the convoy. That knowledge wins the booking today. Increasingly it also decides whether you get found at all.

The agencies most at risk are the ones whose best trips live in a PDF, a WhatsApp catalogue or a senior consultant's head, while the website carries three generic packages. If you sell outbound, the same logic applies to how you describe the destinations you send people to. An assistant that cannot read anything specific about your Manali or your Vietnam will recommend someone else's.

How do you describe a signature trip so AI can find it?

Pick your five best-selling trips and write each one up properly:

• Name the route and the places, day by day.
• Say when to go, and why that season.
• Say who the trip suits: a family with young kids, honeymooners, a group of friends, a senior couple.
• Say what is included, and what is not.
• Say what makes your version different from the standard circuit.

Then use the same words everywhere: your website, your Google Business Profile, your OTA and marketplace listings, your brochures. Consistency matters because the models assemble a picture from many sources. Five descriptions that agree with each other are stronger than five different ones.

Where does NAMA fit in all this?

NAMA is software that runs a travel company end to end, from enquiry to invoice. The discovery work in this article is yours to do on your own channels, and no tool can do it for you. Where we come in is the moment after: when the enquiry arrives, however the traveller found you, NAMA's stated aim is a first quote built from sourced rates in 15 minutes, so the interest you worked to earn does not go cold while someone re-keys hotel rates.