Artificial intelligence is becoming part of a decisive moment in travel: the stage when nothing has yet been booked. A destination is being considered, several hotels are being compared, a budget is taking shape and the practical constraints of the stay are beginning to emerge. At this stage, a conversation with an assistant can help travellers organise their ideas, explore possibilities and prepare their questions.
For tourism professionals, this change deserves more than a choice between technological enthusiasm and anxiety. It invites us to examine what makes an offer understandable, credible and easy to book. A property can have comfortable rooms, a capable team and an appealing location, yet describe its services too vaguely to answer the specific expectations of future guests.
In autumn 2026, studies and platform announcements provide concrete indications of this development. They do not justify claiming that assistants will replace every intermediary. They do, however, show why hotels, destinations and service providers should start working on the quality of their information now.
Planning a trip with AI does not mean entrusting it with the booking
In a study published by lastminute.com on 24 September 2026, 45% of the holidaymakers surveyed said they had used artificial intelligence to plan their holidays during the previous twelve months. Only 12% said they were ready to let it make the booking. This gap highlights an important distinction: requesting a suggested trip and delegating a decision involving one's money require two different levels of trust. [1]
These figures must be presented within their scope. The Censuswide survey covered 5,001 respondents questioned in August 2026 across five markets: France, Germany, Italy, Spain and the United Kingdom. It measured reported behaviour and willingness, rather than the proportion of worldwide bookings actually made by autonomous agents. It was also commissioned by a commercial travel business. [1]
Our interpretation for professionals is therefore measured: AI can contribute to the discovery of an offer without automatically becoming the transaction channel. A customer might ask an assistant for an itinerary, then visit the hotel's website, compare conditions and telephone before booking. This illustrative journey reminds us to support several stages rather than rely on a single interface.
Search is becoming more conversational
On 28 October 2026, Expedia Group announced more than thirty new or improved features across Expedia, Hotels.com and Vrbo. The developments presented included natural-language search, conversational activity planning and AI-assisted accommodation comparisons. These are product announcements, and availability must be checked for each service and market. [2]
The principle is easy to understand through a fictional request: “I am looking for a hotel for two nights, near a railway station, with arrival after eleven at night, an early breakfast and a room suitable for two adults travelling with a child.”
This wording brings together needs that a traveller previously had to check across several pages. It also shows why a generic description soon reaches its limits. “Ideal location”, “warm welcome” and “quality services” do not answer questions about opening hours, room layouts or arrival conditions.
Conversation therefore does not remove the need to produce good information. On the contrary, it makes gaps more visible.
For hotels, start with questions asked at reception
The first resource to draw on is often within the property itself: questions received by telephone, email and at reception. They reveal the uncertainties customers genuinely want to resolve.
Is the car park inside the hotel or in a neighbouring building? Does a space need to be reserved? Is there a height restriction? Is breakfast included in the rate being viewed? Does a late-night arrival require a code, a call or the use of a particular entrance?
Our recommendation is to turn these questions into precise answers, then check that they are consistent across the channels the business uses. Information may be correct on the official website but outdated on another listing. A commercial package may also have changed without its translation being updated.
The aim is not to accumulate pages. It is to build a body of information clear enough to serve customers, staff and distribution partners. This work remains useful even if no assistant recommends the property.
Describe services without turning a possibility into a guarantee
Problems arise when communication oversimplifies a service. “Parking available” can be understood as a guaranteed space. “Family rooms” specifies neither capacity, sleeping arrangements nor possible restrictions. “Accessible” can conceal very different situations depending on the facilities and spaces concerned.
In an AI-assisted journey, these ambiguities can be repeated in a summarised answer. To reduce them, professionals should distinguish what is always included, what depends on the room category booked and what remains subject to confirmation.
Consider a fictional example: instead of simply advertising “late arrival possible”, a hotel could explain the applicable procedure, the information guests must provide and the limits of the service. The precise wording must, of course, reflect how the hotel actually operates.
Precision does not detract from appeal. It helps travellers choose an offer that suits them and saves staff from having to correct, on arrival, an expectation created before booking.
Destinations must provide identifiable local knowledge
For a tourism office or specialist publication, the opportunity is not simply to produce more guides. A succession of interchangeable descriptions offers little to visitors trying to organise their stay in practical terms.
A useful page explains why to choose a particular visit, how to combine it with other activities and which information needs confirmation. It distinguishes documented facts from itinerary suggestions and editorial judgements.
A one-day itinerary should not imply that every monument is open at the same time or that a route is accessible to everyone. Mobility constraints, closures and travelling times need to be examined at the appropriate level of detail.
Local knowledge can also come from professionals and residents, provided their contributions are identified and factual information is checked. AI can help organise these elements; it cannot replace a field visit or confirmation from the person responsible for a venue.
Visibility is about more than writing for assistants
It may be tempting to look for a formula that guarantees a place in AI answers. Such a promise would be excessive. Systems, their sources and selection criteria vary, and the presence of content does not guarantee that it will be used.
Instead, we suggest working on observable fundamentals: clear presentation, accessible pages, consistent information, relevant links and explicit identification of the business. Important questions should have understandable answers that do not depend solely on an image or an advertising slogan.
The commercial logic of the website must also be preserved. Once visitors understand the offer, they should be able to check conditions, contact the property or access the booking journey. Extensive content that facilitates none of these steps can create work without improving the experience.
The challenge is to make the offer easier to evaluate while measuring what actually happens.
Automate preparation and organise verification
Artificial intelligence can help prepare a summary, suggest an article structure, adapt content for different audiences or produce an initial translation. These uses must remain linked to checks proportionate to what will be published.
A neighbourhood description does not involve the same stakes as information about cancellation conditions, specific accessibility arrangements or entry formalities. The more an error could influence an important decision, the more explicit the source and validation should be.
For a small team, a simple arrangement may be enough: one person prepares the material, another checks sensitive points, and publication is then recorded with its date and sources. If resources do not allow this separation, a review conducted separately from drafting is still preferable to immediate distribution.
Tools should reduce repetitive work without making editorial responsibility invisible. Professionals retain control over the promises made in their name.
Translations must preserve the conditions of the offer
Multilingual content is a promising area, but a fluent translation is not necessarily a faithful one. A supplement can disappear, a service available on request can become implied, or an opening time can lose its precision.
Before extending a page into several languages, it is better to stabilise the reference content. Service names, capacities, conditions and exceptions should then be compared across versions.
Imagine an offer that includes parking subject to availability. If this qualification is missing in one language, two customers may understand different commitments for the same service. The problem is no longer merely stylistic: it concerns commercial consistency.
Our advice is to begin with the languages that genuinely matter to the clientele and the pages that influence decisions. Translating the entire archive can wait if the essential information about the stay is not yet reliable.
Human relationships remain necessary when situations become complicated
A stay does not always go according to plan. Transport arrives late, a room fails to match an expectation or a special request requires judgement. At such moments, customers primarily want a solution and someone who takes responsibility.
An automated system should therefore explain how to reach a person, which requests it can handle and when it hands over. Adding a contact button is not enough if nobody is organised to take over the conversation.
For teams, the potential benefit lies in better-prepared exchanges: fewer repetitive questions when answers are available, and more time to understand individual situations. This benefit needs to be verified within the property, rather than presented as an automatic consequence of installing a tool.
Hospitality thus retains a dimension that the speed of an answer alone cannot measure: the ability to understand and act.
A realistic first project for tourism professionals
To get started, we recommend a limited scope. Selecting five important pages, collecting frequently asked questions and correcting contradictory information provides a more concrete starting point than a broad transformation programme.
The second step is to test a specific use: preparing replies, improving a destination page or translating a services page. Every result should be open to review and comparison with its previous version.
Finally, define a few useful indicators. How many answers needed correcting? Are repetitive enquiries decreasing? Can customers find the conditions they need more easily? Can contacts or bookings be attributed to the improved journey? When such attribution is impossible, it is better to acknowledge this than produce a misleading figure.
Success is not measured by the number of paragraphs generated, but by the usefulness of the work completed.
An opportunity to make the offer clearer
AI is changing some ways of planning travel without making tourism's fundamental requirements obsolete: understanding the customer, presenting an honest offer and honouring the commitments made.
Professionals have a concrete foundation to build on. They know their places, their services and their visitors' questions. By organising this knowledge, they can better inform travellers and better equip their teams.
The priority is therefore not to promise guaranteed visibility in every assistant. It is to build reliable, maintained information connected to a real experience. Technology can contribute to this work; trust grows when the stay delivers what was promised.
Sources
[1] lastminute.com, Travel Horizons, 24 September 2026. Censuswide survey of 5,001 respondents in five European markets, August 2026.
[2] Expedia Group, product announcement, 28 October 2026.
Sources consulted on 16 November 2026. Fictional examples and recommendations are editorial analysis, not measured findings from these studies.
There is no comments yet.
You must login Login Sign up