I counted a year of searches on my flight search demo, and only 36% of them had fixed dates.
I build FlightPowers, a flight and hotel price API. The demo at demo.flightpowers.com lets you write something like "cheap flight to Europe sometime in October, 3-4 nights" and get live fares back. Between April 2025 and March 2026 the engine ran 27,320 searches there. About 4,900 signed-in users ran 21,822 of them, and visitors who weren't signed in ran the rest (most searches started in Tel Aviv).
How I counted
The database keeps the structured search the model turned each prompt into (origins, destinations, the date window, nights, stops, sort), never the prompt text. I grouped everything in SQL and never exported a row. Any group with fewer than 20 people stays out of this post.
One prompt can fan out into a couple of searches, about 1.5 per ask when you count one ask per signed-in user per minute, so "searches" below means searches the engine ran.
Right now the data covers one site, and over half of it comes from a busy August and September 2025.
1. Most searches didn't pin the dates
Only 36% of searches had fixed dates. The other 64% left room in the departure date or the length of stay, and nearly a third of all searches left more than two weeks.
2. Nearly one in ten searches listed 10+ airports
71% had a single destination airport. But 9.5% listed 10 or more, and since the prompt isn't stored, my guess is a lot of those were "anywhere cheap" asks.
3. Thursday
When the dates were fixed, 32% of searches asked to leave on a Thursday, 1.8 times Friday's share. Most searches on the site start in Tel Aviv, where the weekend is Friday and Saturday, so my guess is Thursday after work.
4. Short trips, about 40 days ahead
38% of round-trip searches set their shortest stay at 3-4 nights, and the median search looked about 40 days ahead.
5. Bangkok and New York
In searches from Tel Aviv that named a single city, Bangkok drew the most people (about 290), then New York (about 260), and both beat every European city. European cities still fill the top of the overall count, mostly through searches that listed several cities.
What it means if you build flight search
When most searches carry a range, one search is really many API requests. A 31-day month with 3 cities and 3 stay lengths is already 279 requests (31 × 3 × 3).
That's what I built FlightPowers around. A round trip is one request, with both legs priced together and one total. Over REST you fan the dates out yourself, and the 10$ PRO plan allows 150 requests a minute, so those 279 requests take about two minutes. With a RapidAPI key, the MCP server (Claude, ChatGPT or Cursor) takes a date range and a list of destinations in one call, up to 300 billed combinations.
You can try it free from Claude with ads (free-trial.flightpowers.com/mcp, Google sign-in, 50 searches a day), and PRO is 10$ for 2,500 requests a month with a RapidAPI key: https://rapidapi.com/mtnrabi/api/google-flights-live-api
Which do you search with more often, fixed dates or a range? :)
Top comments (3)
Great insight! The 36% fixed-date stat is eye-opening. I noticed that for destination-focused trips (like China rail travel), travelers often search by route/destination first rather than dates. Would be interesting to see if that pattern holds — maybe date flexibility correlates more with business travel than leisure. Thanks for sharing the data!
Thanks Jun! I can't really test business vs leisure, the database keeps the structured search but never the prompt, so I don't know why anyone travels. My guess leans leisure though: 38% of round trips had 3-4 nights as the shortest stay, and fixed-date searches piled up on Thursday, which in Tel Aviv (weekend is Fri-Sat) reads like leaving after work. Your destination-first idea kind of shows up too, 71% named a single destination airport and 64% left room in the dates, though those are two separate counts.
Matan, this is such a sharp read — thanks for writing it up. The Thursday pile-up as 'leaving after work' is a clever catch, and it mirrors what I see in China rail travel: folks fix the destination first, then hunt for the cheapest flexible-date ticket inside a tight window. Your 71%/64% split really backs up that 'destination-first, dates-flexible' instinct. One idea on your blind spot: since the DB keeps structured search but not intent, an optional 'trip purpose' tag at search time could separate business vs leisure with little friction. Either way, great analysis!