Most hotel descriptions are written for humans who are already considering the hotel and need to be convinced. They use adjectives: “luxurious”, “stunning”, “world-class”, “unforgettable”. They describe feelings and aspirations. They are deliberately vague about price and specific about nothing.

AI systems do not use adjectives to make recommendations. When generating an answer to “family-friendly hotel near the Natural History Museum in London under £200 a night”, the AI is matching specific attributes against specific query terms. “World-class experience” does not match. “Family rooms with bunk beds, 7-minute walk from the Natural History Museum, rates from £150” does.

The four-fact opening

Every hotel description should open with four facts in the first two sentences:

  1. What type of property it is (hotel, boutique hotel, B&B, guesthouse, resort, apartment)
  2. Where it is specifically (neighbourhood and city, or district and landmark proximity)
  3. How many rooms it has (scale signal for AI)
  4. The one thing that makes it different from every other property in the same category and neighbourhood

Example of a good opening: “The Meadowbank House is a 19-room boutique hotel in Edinburgh’s New Town, two minutes’ walk from Princes Street Gardens. It occupies a converted Georgian townhouse and has the only rooftop terrace with unobstructed Castle views in this part of the city.”

Example of a bad opening: “Welcome to Meadowbank House, where luxury meets heritage in the heart of Edinburgh. Discover an unforgettable stay in our beautifully appointed rooms.”

The first example gives AI systems four matching attributes in two sentences. The second gives them nothing matchable at all.

The proximity section

After the four-fact opening, add explicit proximity data. This powers location-based queries, which are among the most common travel AI searches.

List: distances to relevant landmarks in minutes on foot, the nearest public transport link and its distance, the nearest airport and approximate travel time, and any relevant location advantages (quiet street vs. central location, views, noise level).

Write it in prose, not a bullet list. Bullet lists in website copy are sometimes stripped or deprioritised by AI crawlers. Prose reads more reliably.

Example: “The hotel is a 2-minute walk from Princes Street, 8 minutes from Waverley Station and 12 minutes from Edinburgh Castle. The Royal Mile is a 10-minute walk downhill. Edinburgh Airport is approximately 30 minutes by tram from the Princes Street tram stop, 400 metres from the hotel.”

Guest type matching

AI systems receive queries with guest type signals: “family with teenagers”, “solo female traveller”, “business trip”, “romantic anniversary”, “group of friends”. Your description should explicitly name the guest types you are best suited for and why.

“The hotel is particularly well-suited to couples visiting Edinburgh for the first time, solo travellers who prefer small, quiet properties, and literary tourists: the hotel is in the same street where Robert Louis Stevenson grew up.”

Attribute statements, not adjectives

Close with three to five attribute statements: specific, factual, matchable.

“The hotel restaurant serves Scottish breakfast until 10:30am. All rooms have blackout curtains and no-street-facing noise. The hotel has no lift and four rooms involve stairs. Parking is available at a public car park one block away at £18 per day.”

Note that negatives are useful too. An AI that knows your hotel has no pool and no lift can correctly exclude it from queries looking for those features, and correctly include it in queries from guests who don’t care about those features.

See also: What is llms.txt for hotels | What schema markup does a hotel need