Most airlines have figured out the product side of ancillary revenue. Seat upgrades, extra legroom, priority boarding, checked bags, lounge access, travel insurance, hotel and car partnerships - the menu of ancillary offerings at a typical carrier is extensive and, in many cases, genuinely valuable to the passengers being offered them.
The products are not the gap. The gap is in how, when, and to whom those products are offered. Ancillary revenue per passenger varies enormously across carriers - not primarily because of differences in the products available, but because of differences in the precision with which those products are matched to the passengers most likely to want them, at the moment they are most likely to buy. The airlines capturing the most ancillary revenue per passenger are not the ones with the longest product list. They are the ones who have gotten closest to treating each passenger as an individual with specific preferences, specific willingness to pay, and a specific point in their journey at which a given offer will resonate.
AI does not create new ancillary products. It changes the precision with which existing products reach the right passengers at the right moment.
The timing problem
Consider how ancillary offers typically reach passengers. A seat upgrade offer arrives in a booking confirmation email, alongside fourteen other pieces of information about the itinerary. A checked bag offer appears during the online check-in flow, when the passenger is focused on getting through the process quickly. A lounge access offer is presented at the gate, ninety seconds before boarding begins.
Each of these moments is suboptimal in a specific way. The booking confirmation email arrives when the passenger's attention is distributed across everything in the message. The seat upgrade is competing with the flight details, the cancellation policy, the hotel cross-sell, and the app download prompt. The checked bag offer during check-in arrives after many passengers have already packed and made their bag decisions. The lounge access offer at the gate arrives too late for most passengers to use the lounge meaningfully even if they buy.
The right moment for an ancillary offer is not determined by where it fits in the booking and travel workflow. It is determined by when the specific passenger is most likely to be receptive to that specific offer. Those two things are rarely the same.
A business traveler who has a four-hour connection and a delayed inbound flight is highly receptive to a lounge offer - not at the gate of their original flight, but the moment their connection time expands and they realize they are going to be waiting. A family traveling with young children who has just cleared security and is looking at a long walk to a gate at the far end of the terminal is more receptive to a stroller or wheelchair assist offer than they were at any point during the booking process. A passenger whose preferred seat was taken and who has been assigned a middle seat is more receptive to a seat upgrade offer in the 24 hours before departure than they were six weeks earlier when they booked.
These windows exist. Most airlines are not offering into them because the offer infrastructure is built around the workflow - booking, check-in, gate - rather than around the passenger's actual state of mind and circumstances at each point in their journey.
The personalization problem
Layered on top of the timing problem is a personalization problem that is equally significant. The same ancillary product has dramatically different value to different passengers. An extra legroom seat is worth a great deal to a six-foot-four frequent traveler on a four-hour flight and almost nothing to a leisure traveler on a forty-minute hop. Priority boarding matters enormously to a passenger who is connecting tight and needs to guarantee overhead bin space, and not at all to a passenger checking bags and traveling with no time pressure. A travel insurance offer is relevant to a passenger booking nine months out on a refundable fare, and largely irrelevant to a passenger booking two days before departure on a non-refundable ticket.
When offers are made uniformly - the same seat upgrade email to every passenger on a flight, the same bag offer to everyone during check-in - the conversion rate on the offer reflects the average relevance across a heterogeneous group. Some passengers for whom the offer is highly relevant buy. Many passengers for whom the offer is irrelevant ignore it. A few passengers for whom the offer is actively annoying form a slightly negative impression of being marketed to.
The revenue captured is a fraction of what would be captured if each passenger received only the offers most relevant to them, at the moment they were most likely to act on them. This is not a hypothesis. Airlines that have moved from uniform ancillary offers to personalized ones have documented meaningful improvements in conversion and revenue per passenger. The improvement is not because the products changed. It is because the match between the offer and the recipient improved.
What personalization actually requires
Getting personalization right in ancillary requires three things that most airlines have imperfectly developed. The first is passenger-level data that goes beyond the booking. A passenger's booking history tells you what they have purchased before. Their check-in behavior tells you something about their travel patterns. Their service recovery history tells you something about what they value. Their loyalty program activity tells you something about their frequency and preferences. A passenger who has purchased extra legroom on 80% of flights longer than two hours is a high-probability buyer of that offer on their next qualifying flight - but only if that data is accessible at the moment the offer is being made.
Most airlines have this data in some form. The problem is that it lives in systems that were not designed to be connected at the point of offer generation. The booking system knows the fare. The loyalty system knows the tier and points balance. The service history system knows the complaints. But assembling all of that into a single view of the passenger at the moment an offer is being generated requires integration work that many carriers have not completed.
The second is a model that translates passenger data into offer relevance estimates. Not just "this passenger has bought extra legroom before" but "given this passenger's history, this flight's characteristics, and the current timing in the booking curve, the probability that they will purchase extra legroom at this price point is X." That probability estimate, generated individually for each passenger for each potential offer, is what makes it possible to sequence offers intelligently rather than broadcasting them uniformly.
The third is the ability to deliver offers through the right channel at the right moment. An offer that would be highly relevant to a passenger is worthless if it arrives through a channel the passenger does not check, or at a moment when they are not receptive. This requires integration between the personalization model and the communication infrastructure - email, app notifications, in-airport digital touchpoints, gate agent prompts - in a way that allows the model to choose not just what to offer but when and how.
Where AI changes the equation
The constraint that has historically limited ancillary personalization is not the concept. Airlines have understood for years that personalized, well-timed offers perform better than uniform ones. The constraint has been the analytical infrastructure required to do it at scale.
Generating individualized offer recommendations for every passenger on every flight - evaluating each potential offer, estimating conversion probability for that specific passenger at that specific moment, sequencing the offers to maximize total revenue without overwhelming the passenger - is a computational problem that is intractable to do manually and difficult to do with rule-based systems. Rule-based personalization is better than no personalization. If a passenger has purchased extra legroom before, show them the extra legroom offer. If they are a top-tier loyalty member, show them the upgrade offer. But rules are blunt instruments. They cannot capture the interaction between multiple variables - the passenger's history, the flight characteristics, the timing, the price point, the channel - in a way that produces genuinely individualized recommendations.
Machine learning models can. A model trained on historical ancillary purchase data, with passenger attributes, flight characteristics, offer timing, and price points as features, learns the complex interactions between these variables that rules cannot capture. It finds that a specific combination of loyalty tier, flight duration, booking lead time, and seat assignment history is highly predictive of extra legroom purchase - a pattern that no rule designer would have identified but that the model surfaces from the data.
Applied at scale, across every passenger on every flight, this kind of model generates a meaningfully different offer strategy than uniform or rule-based approaches. Not because it is predicting the future - no model knows with certainty whether a specific passenger will buy. Because it is ranking the probability of conversion more accurately than any alternative, which means that the offers being made are better matched to the passengers receiving them.
The revenue math
The improvement from better ancillary personalization compounds in a specific way that is worth making explicit. Better timing and personalization does not just increase conversion rates on individual offers. It also reduces offer fatigue - the phenomenon where passengers who receive too many irrelevant offers stop engaging with ancillary communications entirely. An airline that sends personalized, well-timed offers maintains a higher baseline engagement rate across the passenger base, which compounds over multiple trips.
It also affects the relationship between ancillary revenue and ticket pricing strategy. Airlines that generate strong ancillary revenue per passenger have more flexibility in their base fare pricing - they can afford to be more competitive on the ticket price because they are capturing revenue on the back end that competitors with weaker ancillary performance cannot. This is a structural advantage that grows over time as the personalization model improves and the ancillary revenue per passenger increases.
The product gap between major carriers on ancillary is small. The execution gap - in timing, in personalization, in the precision with which the right offer reaches the right passenger at the right moment - is large. That execution gap is where the revenue opportunity lives, and it is the gap that AI is best positioned to close.