Beyond Subtitles: Why Multilingual Passenger Experience Is Now an Operations Problem

Here is what airlines and in-flight entertainment providers are actually asking about language, and what it takes to solve it at scale.

A passenger books a flight in Portuguese. She checks in through an app that switches to English halfway down the page. She boards, opens the seatback screen, and finds a catalog where the titles are localized but the synopses are not. Mid-flight, the schedule changes. The disruption notification arrives in English. She opens the airline's chatbot to ask what happens to her connection, and the bot offers her four languages, none of them hers.

Nothing in that journey failed. The aircraft flew. The app worked. The entertainment played. And yet the experience quietly told her she was a secondary customer.

This is the gap most airlines are living with right now. It is not a translation problem. It is an operations problem, and it shows up in exactly the places aviation CX and digital teams are being measured: self-service, satisfaction, disruption handling, and the perceived quality of the on-board product.

What airline and IFE teams are actually asking

Over the past year we have spent a lot of time in rooms with airlines, in-flight entertainment providers, content programmers and aviation technology companies. The questions repeat with striking consistency, and they are not the questions a translation vendor is usually asked.

The most common concern, by a wide margin, is quality and governance. Not "can you translate this," but "how do I know it is right in a language nobody on my team reads?" Teams want to know how a native speaker's corrections get captured, whether an approved fix will still be there in the next release, and how old content stops overwriting new content.

Second is workflow and integration. Aviation content is highly structured. Subtitle timing, character limits on an interface string, metadata fields, tags, and non-standard templates all mean a linguistically perfect translation can still be unusable. Teams want APIs and connectors, not email attachments and spreadsheets.

Then comes speed. Airline deadlines are externally imposed and frequently short. A carrier adds a route, a content window opens, a campaign goes live, and the request lands with days to spare, not weeks.

Where language breaks in the passenger journey

1. In-flight entertainment

This is the most aviation-specific use case and the one with the most hidden complexity. Subtitles, dubbing, audio description, program titles, synopses, genre descriptions and promotional content all have to work within character limits, timing constraints and metadata structures. Get any of those wrong and the content is technically translated and practically broken.

Volume is the other half of the problem. Catalogs refresh on a cycle, each airline client has its own language list, and a single new carrier can add a dozen languages overnight.

2. Passenger-facing interfaces

Seatback interfaces, mobile apps, web check-in, content catalogs, menus, navigational labels and service instructions all demand short, clear copy that fits a fixed space. One in-flight entertainment provider we work with recently localized a passenger interface across 17 languages, and the single biggest lever on quality turned out to be context. Isolated strings do not tell a linguist whether "Return" means going back a screen or the second leg of a journey. Give the same expert screenshots and surrounding context and the output changes completely.

3. Support, service and disruption communications

This is where CX leaders feel language most directly. Help center articles, agent macros, canned responses, inbound tickets and chatbot flows are high volume and high sensitivity at the same time. A cold or literal translation in a disruption moment does more damage than no translation at all, because it lands precisely when the customer is anxious.

Chatbots and virtual agents are often built and tuned in a handful of languages, and coverage thins out beyond them. Wherever that happens, self-service is capped and the contact lands in a queue that then needs a multilingual agent.

4. Marketing, loyalty and retail

Campaigns, offers, ancillary merchandising and loyalty communications have to feel written for the market, not translated into it. This is transcreation territory, and it is where literal accuracy is the lowest bar, not the goal.

5. Safety, accessibility and regulated passenger communications

Not all passenger content carries the same risk, and treating it as though it does is how programs get slow and expensive at the same time. Safety and service information, accessibility features, and anything an auditor or an airline client's legal team may eventually read belong in a different tier from a seasonal campaign.

The European Accessibility Act* has been applicable since 28 June 2025, and passenger transport services sit explicitly within its scope, alongside e-commerce, consumer banking and audiovisual media. For aviation that reaches booking flows, apps, self-service kiosks and the audiovisual content served on board. Its audiovisual provisions require captions, audio description and spoken subtitles to be transmitted in full, synchronised and at adequate quality.

It is worth being precise about what that does and does not mean, because the distinction is where most confusion sits. The directive requires those accessibility components to be preserved and delivered properly. It does not require them in every language you serve. Penalties are set nationally under Article 30 and vary widely, from roughly EUR 60,000 in Ireland to SEK 10 million, about EUR 900,000, in Sweden, with no EU-level cap.

Then there is the risk that arrives from inside the building. When translation is slow or simply unavailable, people route around it. A service notice gets pasted into a consumer AI tool. An agent runs a passenger message through a free translator. Each of those is a small and entirely reasonable decision, and together they produce data exposure, terminology drift and passenger-facing content nobody approved, in languages nobody on the team can read back.

Why the traditional model runs out of road

Most airlines got here honestly. Language got solved once, per team, per need. Marketing has an agency. IFE has a media vendor. Support has a translation queue. The digital team has whatever is built into the CMS. Each solution works in isolation and none of them share anything.

The result is predictable. Terminology drifts between the app and the seatback screen. The same phrase gets paid for four times. Nobody can answer a simple question like "how much of our passenger-facing content is actually live in Arabic, and how good is it?" And every new language means another vendor conversation and another headcount discussion.

The fix is not a better vendor. It is treating multilingual content as infrastructure: one connected system where language assets compound, terminology is governed centrally, content flows through APIs rather than inboxes, and human expertise is applied where risk actually sits instead of uniformly across everything.

That is the shift from AI as a translation shortcut to AI as an operating layer. Agentic workflows handle intake, routing, formatting and first-pass review automatically. AI review agents fix errors before a human sees them, and every fix trains the model. When content is high-risk, sensitive or brand-critical, agents call in the right human expert by policy rather than by exception. Analytics show what is live, in which language, at what quality level, which is the thing leadership actually asks for.

What this looks like in practice

Anuvu, a global provider of in-flight entertainment and connectivity to airlines worldwide, faced exactly this balance of accuracy, turnaround and scale. Traditional methods were slow and labor intensive, inconsistencies posed a brand risk in international markets, and demand for multilingual content including closed captions kept growing.

"We were looking for an exclusive partnership to ensure that our customers were going to receive the best quality in any kind of localization that we do."

Jim Reed, VP, Media Operations, Anuvu

Working with LILT, Anuvu cut localization time while holding accuracy and consistency, built a centralized hub that integrated with existing systems, and gained the scalability to absorb higher volumes without a proportional increase in effort. Translation memory delivered significant time and cost savings.

"It's allowed us to not only work faster, but also improve the overall quality of our translations, which has helped us to better deliver on our brand promise of dreaming and delivering innovation that moves you."

Jim Reed, VP, Media Operations, Anuvu

The outcome most relevant to CX and digital leaders is not the speed number. It is what the speed bought.

"Ultimately, what we've seen AI help us with is the ability to increase our capacity and scalability. But it's also freed up resources to focus on higher level tasks, such as innovation and creativity. So I think at this point, AI has been an integral part to our growth and success."

Jim Reed, VP, Media Operations, Anuvu

That is the real prize. The team stopped managing translation and started improving the passenger experience.

Six questions worth asking your own team

1. How much of our passenger-facing content is live in every language we claim to serve, and how would we prove it?
2. When a native speaker corrects a translation, where does that correction live, and does it survive the next content refresh?
3. What percentage of our translation requests are still submitted by email or spreadsheet?
4. How many languages does our chatbot genuinely handle, and what happens to self-service rates outside them?
5. If we added a new carrier or market tomorrow with a new language list, how long until content is live?
6. Which of our passenger communications count as safety-critical or regulated, what routes those to a stricter level of review, and could we show an auditor who approved each language version and when?

If any of those answers are uncomfortable, the constraint is almost certainly the operating model rather than the people.

Frequently Asked Questions

Meet us in Lisbon

LILT will be at World Aviation Festival 2026 in Lisbon, 13 to 15 October. If you are working on multilingual passenger experience, IFE localization, multilingual self-service, or simply trying to figure out how AI fits into your content operation without putting quality at risk, we would like to talk.

Book a meeting with the LILT team

Related: How Anuvu amplifies passenger experience with LILT's AI-powered localization and the Cabin Chronicles episode Beyond Subtitles: Using AI to Speak Every Passenger's Language.

*Informational only. Not legal advice. Regulatory facts current as of September 2026.