I have run hotels for more than thirty years. I have stayed in 290 of them in Manhattan alone, never repeating one, and traveled to 93 countries doing the same thing everywhere else. I know what a good room feels like before I open the door. That instinct now makes me the wrong customer for the machine that is planning your next trip.
AI has fixed the boring part of travel. Feed it a city, a budget, a set of dates. In seconds, you get neighborhoods, flights, restaurants, a clean itinerary. The three hours you used to lose to browser tabs are gone. Good riddance.
Planning got faster. Discovery got smaller.
The machine only trusts what it can count
A recent study ran twelve AI models through more than sixty thousand hotel-selection tests. Researchers built five synthetic hotels and randomized the ratings, the prices, the review counts, the brand names, the list position.
The machine had opinions. A 4.7 rating over a 3.9 moved selection by 31.6 points. A $249 rate against $129 dropped it by 30. A hotel with 2,100 reviews beat one with 45 by 8.3 points. Simply appearing first on the list was worth about twelve dollars a night.
None of that is wrong. It is just thin. Stars, price, and review count describe a transaction. They say nothing about how a place makes you feel walking through the door at eleven at night after a delayed flight.
Competence is easy to score. Character is not.
What actually stays with you never fits in a filter
A lobby full of locals at six in the evening. A bartender who reads your mood before you order. A building that looks like it grew out of its own street instead of a catalog. A front desk agent who knows the difference between a guest who needs help and one who needs to be left alone. None of that shows up next to “free Wi-Fi” and “rain shower.” Hospitality is timing, restraint, and judgment. You cannot filter for judgment.
So hotels do the only thing they think the machine understands. They feed it more inputs. More optimized copy. More manufactured reviews. Longer amenity lists. Photographs of rooms that no human has ever slept in.
Search enough hotel websites and every one of them promises an “immersive experience” and “unparalleled service.” That language describes nothing. It turns a hundred different buildings into siblings wearing the same beige suit, and the algorithm cannot tell a good liar from a good hotel any better than a tired guest can.
Consensus is not the same as quality
Ask for the best hotel in Rome or Tokyo and the same six names surface. They have earned it, mostly. Thousands of reviews. Big marketing budgets. Years of digital residue built up like sediment. The problem is not that those hotels are undeserving. The problem is what gets buried under them.
Independent hotels do not have the review volume. New hotels have no history yet. Small hotels cannot buy the distribution. The best places are often the hardest to describe, because what makes them work is a feeling, not a feature. Recommendation engines run on the same physics everywhere. Popular gets more exposure. More exposure books more rooms. More rooms generate more reviews. More reviews feed the algorithm the popularity it already believed in. It is a closed loop, and it starves anything that has not already won.
Travel gets smoother. It also gets smaller. You fly six thousand miles to have an experience you could have found at home.
Stop asking the machine lazy questions
“Best hotel” tells the machine nothing about you. Best for a honeymoon, a family of five, or a man who wants a lobby with no conference badges in it.
Ask better, and it answers better:
- Find an independent hotel in Rome where locals actually drink at the bar.
- Show me a hotel whose architecture belongs to its neighborhood, not a brand deck.
- Skip the global chains and anything famous mainly from Instagram.
- Which reviews mention staff by name?
- Does the hotel restaurant pull people who are not sleeping upstairs?
Then make it show its work. Ask what evidence backs the answer. Ask if locals eat there. Ask if the hotel connects to anything real in the neighborhood, a chef, an artist, a shop down the block.
You are not asking AI to remove the risk from travel. You are asking it to widen what it is willing to show you.
Run the human test before you book
Ten minutes. Before you hand over your card. Look at the hotel’s social feed. Are there guests actually using the space, or an empty chair styled for a furniture catalog? Read the recent reviews. Do people name an employee? A named employee means someone made a human connection. “The room was clean” means housekeeping cleared the minimum bar of civilization.
Check who eats at the restaurant. Locals with reservations mean the hotel is part of the city. Only hotel guests at every table means it is an amenity wearing a menu.
Read the copy. Does it say something specific, or does it recite the standard luxury vocabulary that means nothing because it means everything?
Then ask one real question. Would you remember this place six months from now without opening your camera roll.
Use the machine for speed. Use it for range. Use it to kill the browser-tab purgatory that used to eat your Sunday afternoon.
Keep the last decision for yourself. That is the one part of travel no algorithm has earned the right to make.
A good hotel meets your preferences.
A great one changes them.
-longing for belonging™
