Airline technology providers routinely advertise revenue gains of several percentage points from AI-powered pricing, and some airlines report positive results. The public evidence behind those figures varies widely, from vendor averages and projected gains to one airline’s account of a split-route test. TDN reviewed the published claims and found that very few give the baseline, the measurement period, the markets or the revenue measure an outside reader would need to reproduce the result.
That does not make the figures wrong. It means the public record cannot settle the question, and the figures should not be read as measurements of the same thing.
Most of the figures are averages or projections
Vendor averages come first. PROS lists an average revenue uplift of 2 to 3 percent for its airline revenue management product on Microsoft’s AppSource marketplace. Amadeus says airlines running its Air Pricing Optimization in production report an average of 3 to 5 percent, varying by airline, and separately cites more than 3 percent on airfares and 7 percent on ancillaries for its dynamic pricing products. Both are averages across unnamed airlines.
Next are advertised potentials. Sabre says its Continuous Revenue Optimizer, built with Riyadh Air, can unlock up to 3.5 percent more revenue without adding capacity. Fetcherr told Aviation Week in January 2025 that its system could lift revenue by 6 to 9 percent once implemented. Both describe what a customer may achieve.
Then come the airline-reported results. Azul’s director of planning and revenue described a 3 to 5 percent uplift and about 10 percent less manual work at World Aviation Festival 2025, in a session shared with Fetcherr. SAS said Amadeus’s tool produced more than 3 percent in its initial markets, on its website and NDC channels. Delta’s president at the time, Glen Hauenstein, said in November 2024 that early results were “amazingly favorable” against the beta, without giving a number.
One airline describes its test method. Virgin Atlantic’s pricing executive told a podcast hosted by Fetcherr that the airline gave the system half of a route, kept its existing processes on the other half and measured one against the other. The figure was not disclosed, and the account has not been independently audited.
These claims concern different systems, baselines, pricing methods and possibly different measures of revenue. A 3 percent average, a 3.5 percent ceiling and a 3 to 5 percent airline result do not describe the same thing.
Sabre’s 3.5 percent is a modelled figure
Sabre’s figure shows why the distinction matters. The company’s release of 14 October 2025 describes the revenue generated across its Mosaic products as “estimated”. The footnote on its product page attributes the 3.5 percent to the IATA case for change, a McKinsey study and Sabre’s own analysis. That makes it a modelled ceiling rather than a measured result. TDN carried the figure on 29 September as Sabre’s claim in an interview, and it should be read that way.
To be exact about what is verified, every figure above is verified as a statement the company or airline has made. Whether the figures are accurate is not verified, because TDN found no independent audit of any of them.
Delta still calls its tool a pilot
On 9 September, Delta’s president Peter Carter wrote to US Representative Frank Pallone, who is investigating surveillance pricing across the airline industry. Delta published the letter on 18 September. It describes the airline’s AI pricing tool as a piloted use in domestic and international test markets, supporting analysts rather than replacing them. It says only route-level demand and market data feed a dedicated private deployment, and that no customer personal data is used. The letter reports no measured uplift.
In July 2025, Hauenstein said about 3 percent of domestic fares were priced with the tool and that the goal was 20 percent by the end of that year, as reported at the time. TDN found no later public figure. On Delta’s third-quarter call on 9 October, executives attributed results to demand, premium and loyalty revenue and to a fare increase of about 20 percent, which CEO Ed Bastian tied to fuel costs. The transcript does not mention AI pricing.
A year in which fuel costs pushed fares up across the industry is a hard one in which to isolate what any pricing system contributed. That is TDN’s analysis, and it is why the method behind a quoted figure matters as much as the figure.
Airlines may hold evidence the public cannot see
TDN reviewed only public material. Airlines may run holdout tests, receive detailed internal performance reporting or hold audit rights in their contracts, and none of that would appear in a press release. No contract terms or fee structures are public, and TDN has not assumed any. The finding is limited to what outsiders can verify, not to what an airline’s own commercial team knows.
For airlines in emerging markets the gap is practical. Kenya Airways is adopting Sabre Mosaic Offer Optimization, and Ethiopian Airlines has selected Sabre’s Fares Optimizer, Air Price and Ancillary tools, according to what Sabre told TDN. Royal Air Maroc is named as a Fetcherr customer. Selection is not deployment, and none of the three has published measured results.
Three questions for a renewal meeting
Three questions follow from the gaps. How was the claimed uplift measured, who built the baseline, and can the airline reproduce it without the vendor? Is any part of the fee tied to measured uplift? And if the airline switched suppliers, would it keep the data and history needed to test a successor against the incumbent?
A percentage needs a method behind it
AI pricing may well be delivering the gains its sellers describe. The public record cannot show it, and that is a weakness for an industry asking boards to commit to multi-year contracts on the strength of a percentage. A percentage without a baseline, a period and a method is a marketing figure, however sincere. The vendors and airlines that publish how they measured will be believed, and those that publish only the number are asking the market to take their word for it.
