American Puts Screens Back in Every Narrowbody Seat. An Ad-Funded Game Still Earns Nothing Up There.

Licensing HTML5 games for in-flight entertainment means giving up ad revenue. A seatback game cannot reach an ad server, so IFE deals are flat-fee licences — and every licensor who quotes an airline off a portal rate card learns this the expensive way, usually after the integration is half-built.

This matters more this month than it did last month. On 18 August 2026, American Airlines announced it is putting 4K seatback screens at every seat across its narrowbody fleet, reversing the decision it made in 2017 to strip them out on the theory that passengers preferred their own phones. Trade coverage of the announcement put that fleet at more than 870 aircraft, with new deliveries carrying screens from 2028 and retrofits running into the early 2030s. The airline's own release lists the content categories: movies, TV, music, games.

That last word is a purchase order eventually. But it is not the purchase order most HTML5 licensors think it is.

🛬 A Reversal Worth Reading Carefully

The seatback-versus-bring-your-own-device argument ran for a decade and never resolved, because both sides were right about different flights. American went screenless in 2017. United and Delta kept investing in hardware. Now American has switched back while simultaneously fitting Starlink across its narrowbody fleet starting in 2027, which means the same aircraft will carry a fixed screen and real broadband to every passenger's phone.

The industry did not pick a winner. It ended up with both, on the same airframe, and they are not the same product for anyone selling games into them.

📴 Three Surfaces That Look Like One

Say "in-flight games" to a licensing team and they picture one integration. There are three, and they differ on the only axis that decides your revenue model: whether the running game can open a socket to the public internet.

• The seatback unit. An embedded system — Thales's AVANT platform, for instance, is Android-based — pulling content from an onboard server. From the game's point of view there is no internet. There is a filesystem.
• BYOD over wireless IFE. The passenger's phone joins a closed onboard Wi-Fi network served by a local box, and browses a portal. Feels like the web. Is not the web. Same walled garden, nicer screen.
• BYOD over satellite broadband. Starlink or Amazon Leo, free to the passenger, genuinely routable to the open internet. This is the only one of the three where an ad request completes.

Two of those three are offline environments wearing a browser costume. Your game renders in a WebView, your CSS works, your touch input works — and your first fetch() to anything off-aircraft hangs until it times out.

💸 Nothing in the Air Fires an Ad Request

Walk through what a standard licensed HTML5 casual title does in its first eight seconds and count what breaks on a seatback unit.

1. Loads a third-party analytics script from a CDN. Fails.
2. Initialises an ad SDK, which phones home for config. Fails.
3. Requests an interstitial before the menu. Fails, and depending on the SDK, blocks the loading screen behind a spinner that never resolves.
4. Fetches a webfont, a leaderboard, a remote config JSON. All fail.
5. Fires a rewarded-video call at the first continue prompt. Fails, and now the game has a dead button in the middle of the core loop.

Nobody ships that on purpose. It ships anyway, constantly, because the build being handed to the airline is the same build that runs on a web portal, and on a web portal all five of those calls succeed silently. The failure mode only appears at 35,000 feet, which is the single worst place to discover it.

Even on a Starlink aircraft where the pipe is real, the economics do not rescue you. In-flight digital advertising has been a stubbornly small business for structural reasons. Valour Consultancy put the category at $266 million in 2019 and forecast $3.3 billion by 2030 — a forecast now five years old and worth treating as a direction rather than a number. What has aged well from that analysis is the list of barriers: fill rates then sat around 30% for video and 20% for display, targeting could not segment below the city-pair level, airlines impose brand-safety restrictions that rule out most programmatic demand, and the industry has no first-party data standard. Those are not bugs a better SDK fixes. They are properties of selling ads to a sealed metal tube.

So the honest position: on a seatback screen, ad revenue is zero. On a connected BYOD portal, it is a rounding error attached to a compliance headache. Build the deal on neither.

🧾 Price It Like a Film Licence, Because That Is What It Is

Airlines already know how to buy content that cannot monetise itself in the seat. They have done it with film for fifty years: a fee, a term, a defined set of rights, no revenue share with the passenger's wallet. Games slot into the same procurement muscle memory, and licensors do better when they stop fighting it.

Practical shape of a workable IFE game deal:

• Flat fee, defined term. Annual or multi-year, quoted per fleet or per tail count, not per impression.
• Territory and route rights written down. Film licensing in aviation is carved by territory and sometimes by the airspace an itinerary crosses. Do not assume a worldwide grant is free to give — check what your own upstream rights allow before you promise a global carrier anything.
• Localisation in scope, priced. A carrier's cabin is multilingual by default. Language count drives your cost, so it belongs in the quote, not in a change request.
• A maintenance window, not a support SLA. You cannot hotfix an aircraft. What you can commit to is a build delivered against a content cycle. Promise that instead.
• Ad and IAP code stripped, contractually. Make the offline build a named deliverable so nobody ships the portal bundle by accident.

What the airline is buying is dwell time on a screen it paid to install, and a line on a competitive comparison chart. It is not buying eCPM, and it will not accept a revenue-share structure that requires it to hand you passenger data. Quote accordingly. Our licence and pricing structures separate flat-fee grants from portal deals for exactly this reason.

🗓️ Your Build Ships Once and Then Sits

The constraint that surprises software people most is the update cadence. IFE content moves on a loading cycle, and historically that cycle has been brutal. APEX documented the traditional pipeline in detail: a roughly 90-day cycle in which a file is passed between the producer, the licensor, the subtitling and metadata houses, a transcoding step, a content integration facility, a data centre, an airport-secure facility, and finally hand-carried onto the aircraft. Michael Childers, then chair of the APEX Technology Committee, described per-file delivery costs in the hundreds of dollars under that model.

Cloud delivery has been chipping away at this since. The pressure is still live in 2026 — in August, AirFi launched AirFunnl, a media delivery stack it says cuts average file sizes by up to 45% specifically to enable smaller, faster-refreshing onboard libraries. The direction of travel is good. It is still not a deploy pipeline.

Consequences for how you build:

• QA happens before delivery, in full, on target hardware. There is no "we'll patch it next sprint."
• No server-driven content. No remote config, no live events, no seasonal skins pushed from your backend.
• Difficulty tuning is frozen at ship. If your retention curve depends on tuning against live telemetry, that mechanism does not exist here.
• Every asset is in the bundle. File size is a hard commercial input, not a performance nicety — you are competing for space on a server that also holds a film library.

📐 The Spec That Actually Survives a Cabin

Assume the worst plausible target and you will pass integration testing the first time.

Input

Seatback touchscreens are at arm's length, at an awkward angle, on a surface that flexes when the person in front reclines. Some units are driven by a handset or a D-pad rather than touch. Design touch targets large, avoid anything requiring precision drag or multi-touch, and make sure the game is completable with four directions and a select button.

Audio

Most passengers are not wearing headphones when they open the games menu. Audio must be strictly optional and the game must be fully playable, and legible, in silence. No audio cue may carry information that is not also carried visually.

Session length

Sessions get interrupted by service, turbulence, announcements and sleep. Anything that punishes a two-minute walk-away is wrong for this environment. Save state locally and resume cleanly.

Network

Zero external calls. No CDN fonts, no analytics, no ad SDK, no leaderboard, no telemetry beacon. If the code contains a network call at all, it must degrade instantly and invisibly rather than retry. Build offline-first and test with the device airplane-moded on a bench.

Performance

Embedded IFE hardware is chosen for durability and certification, not for benchmarks, and it stays in service for many years after installation. American's screens will start arriving in 2028 and still be flying in the late 2030s. Target the low end and stay there. The HTML5 catalogue titles that port cleanest into embedded environments are the ones that were already disciplined about asset budgets on low-end Android.

🚪 You Are Not Selling to the Airline

Very few game licensors sell direct to a carrier. The buyer is usually the IFE system integrator or the content service provider sitting between you and the airline, and they have their own catalogue to protect. Panasonic Avionics advertises 300+ games playable via seatback touchscreen, handset or personal device, and names developer partners including Bandai Namco, Envee and Gameloft. Thales ships a games and apps catalogue on AVANT. You are pitching into a slot in someone else's line-up.

What that changes about the pitch:

• Content rating is a gate, not a checkbox. The cabin is a mixed-age audience with no parental controls and no way to leave. Anything with gambling mechanics, simulated violence, or an unmoderated social surface is out before the conversation starts.
• Chain of title gets scrutinised. Integrators carry the airline's risk. Expect to prove you can grant what you are granting, especially for anything with licensed characters or music.
• Volume is the argument. A single title is not worth the integration overhead to anyone. A curated, pre-cleared, uniformly-specced block of titles is. This is one of the few channels where catalogue depth genuinely beats a hero title.

📉 Games Are the Tail, Not the Headline

Keep the ambition proportionate. A Moment survey presented at APEX EXPO 2025 and reported by APEX found 72% of passengers name video-on-demand as their top content preference, ahead of both games and music, with 92% preferring films and series as a format. The same research found 97% preferring their personal device, 95% of them a phone, and 59% reporting no access to charging — a detail that quietly caps how long anyone plays anything. On a five-hour flight, passengers spent an average of 58.4 minutes engaged with the IFE platform, about 20% of the flight. Note that the underlying fieldwork dates largely to early 2024, so read it as structure rather than as this year's number.

Two things follow. Games are not going to out-draw the film library, and pretending otherwise in a pitch marks you as someone who has not flown. But 58.4 minutes is a real budget, video does not fill all of it, and the passenger who has finished their film at hour four is exactly the person a well-specced puzzle game catches. Sell the tail honestly. It is a good tail.

The BYOD side has a second angle worth naming. Free satellite Wi-Fi is increasingly gated behind loyalty signup — United's Starlink service is free to MileagePlus members, and Delta has said its forthcoming Amazon Leo service will be free to SkyMiles members when it begins rolling out in 2028. That makes the onboard portal an enrolment funnel, and a game is a legitimate reason to spend time on a page whose real job is to convert a signup. That argument lands with a commercial team far better than an engagement stat does.

🚫 Four Ways Licensors Lose This Deal

• Quoting the portal price. A per-domain web licence priced against ad revenue makes no sense to a buyer who cannot serve ads. It also anchors you low, because the airline deal should be worth more per title, not less.
• Assuming Starlink solves it. Broadband to the passenger's phone does not put an ad server behind the seatback unit, and the ad market up there is thin regardless. Connectivity changes the plumbing, not the business model.
• Shipping the portal build. One un-stripped ad SDK produces a hang on a device nobody can reach for three months. This is the failure that ends the relationship.
• Treating certification as paperwork. Rating, rights, localisation and hardware QA are the deal. The game being fun is assumed.

✈️ What to Do Before You Pitch

Take five titles you already licence. Strip every external call, force them offline, and run them in a WebView on the cheapest Android device in the office with airplane mode on. Play each one for ten minutes with the sound off, at arm's length, using only a D-pad. Whatever still works is your aviation reel. Whatever breaks tells you exactly what your standard build assumes about the internet — which is useful well beyond aviation, because the same assumptions break on embedded and low-connectivity surfaces generally, from set-top boxes to kiosks to offline Android fleets.

Forestry Games has licensed HTML5 and Android titles since 2017 and specs builds for offline and embedded targets as a matter of course, which is the part of this that is tedious rather than clever. If you are assembling a games block for an IFE integrator, the useful first conversation is about rating, rights and file size — not about which titles are the most fun. The fun ones survive the constraints or they do not, and you want to know which before the aircraft door closes.