On 17 January 1985, American Airlines started selling cheap tickets called Ultimate Super Saver fares. It did this to compete with People Express, a low-cost airline. The fares were cheap enough that they could bring in extra travellers.
But there was a hard problem. How many seats on each flight should be sold at each price? And how should that change as people booked? American called this yield management. Its goal was "selling the right seat to the right customer at the right time." In 1988, American built DINAMO, a computer system that did yield management automatically. A 1992 paper estimated that it added $1.4 billion in revenue over three years, according to a case study of the system.
Old ticket prices came in steps, like a ladder
Airlines still use something called fare buckets. A fare bucket is a group of seats that all sell for one price. Each bucket usually has a letter from A to Z. So two people in the same kind of economy seat can pay different prices.
When the cheapest bucket sells out, it closes and the next bucket, which costs more, opens. If not many people are booking, the airline can open a cheap bucket again. If lots of people want the flight, it can close a bucket early. For years, analysts wrote pricing rules, like raising prices once a quarter of the seats are booked.
So prices move in steps, like a ladder. The classic system has 26 price points, one for each letter, but many airlines add two-character codes to get more. A ticket might jump from $150 straight to $210, with no price in between.
The drawing below shows that ladder next to the newer way, continuous pricing.
AI can pick a price in between the steps
Continuous pricing gets rid of the steps. The airline sets the price at the moment you search, and it can pick any number in a set range, like $193 or $204. This mostly works on the airline's own website and app, or on travel sites that get offers straight from the airline.
One company doing this is Fetcherr, an Israeli company that sells AI pricing to airlines. Its computer models learn from data about prices, seats and routes, plus live information about other airlines' fares, demand and seasons. Fetcherr says this data is aggregated and anonymised, so it does not show who anyone is.
Fetcherr uses reinforcement learning, where a computer gets better by trying things and learning from rewards and penalties. It tests pricing ideas in simulations to find the ones that make the most money.
The AI also has to follow rules. In Fetcherr's examples, prices have to stay between set limits, they can't jump too much from one change to the next, and an economy seat can never cost more than business class. Fetcherr also says its "Glass Box" approach gives clear explanations for every AI decision.
Four airlines testing, choosing or using AI pricing
Virgin Atlantic signed up with Fetcherr in 2023 to update its fares often and in real time. Delta said in a 2025 letter to US senators that it was testing Fetcherr's price suggestions in test markets. Delta also said its analysts "oversee and fine-tune the recommendations".
In Brazil, the airline Azul added the system one market at a time. André Américo, Azul's director of planning and revenue, said revenue went up by 3 to 5% from the moment the airline started using it. Américo said this in a sponsored 2025 interview, with Fetcherr's chief AI officer taking part too.
In February 2026, Thai Airways chose a tool called Air Dynamic Pricing from Amadeus, a travel technology company. Amadeus's announcement says the tool changes fares in real time and lets the airline get an average revenue increase of more than 3%.
Pricing the flight, not the passenger
Pricing a flight means looking at how busy it is. Pricing a person means using their data to guess what they would pay. The second one is called surveillance pricing. US Congressman Frank Pallone describes it as using "a consumer's online data—location, demographics, browsing history, shopping habits, or device type—to set individualized prices".
Airlines say they don't use personal data to set individual prices. Delta's letter says it has never used, tested or planned a fare that targets customers with individual prices based on personal data. JetBlue told CBS News in April 2026 that it "does not use personal information or web browsing history to set individual pricing." JetBlue said this in answer to a lawsuit that claims it does.
Pallone's August 2026 announcement says letters went to eight US airlines, including American, Delta, United and JetBlue. They are part of an inquiry into surveillance pricing, and the airlines were asked to answer by 25 August.
Fewer cheap seats on busy routes, maybe more on quiet ones
Bloomberg reported in July 2026 that AI lets airlines change seat prices faster and close the price gaps that used to help travellers find cheaper flights. So finding a lucky cheap seat on a popular flight is getting harder.
Bryan Terry of Alton Aviation Consultancy told Bloomberg that airlines will get smarter about pricing, raising fares where possible and cutting prices where they have room to attract more buyers. Terry expects cheaper fares on off-peak and less popular routes, because airlines will use AI to fill empty seats.
There is good news if you plan ahead. Guy Leitch, another aviation analyst in the report, said "by booking smart or early, passengers can get below-average fares."