Localization, as we know it, will die by 2033

The experience is moving. Our work needs to move with it.

TL;DR

By 2033, I expect preparing and translating a fixed source experience into a planned set of languages to end as the standard approach within digital software and information delivery. Not gone, but no longer the default.

Three changes drive the shift:

• interfaces generated around intent,
• agents that reach capabilities without the provider's UI,
• people will bring their own translation.

Authored localization still matters when exact wording and review have material and financial consequences.

Internationalization and cultural expertise must move into the systems creating and delivering those experiences.

---

I have spent more than thirty years helping organizations communicate and do business across languages and cultures. I think the model that has supported much of our industry will lose its place as the default within seven years.

By 2033, I expect preparing and translating a fixed source experience into a planned set of languages to be a specialist approach within digital software and information delivery. It will still exist, but a larger share of people will get what they need through experiences assembled for them, at the moment they need them.

Three changes are converging:

• Interfaces can be generated around intent.
• Agents can reach capabilities without navigating the provider's interface.
• People can bring their own translation and interpretation to the encounter.

Each challenges a different assumption about what we localize and who needs us to do it.

The interface arrives with the task

Where does your localization workflow fit into an experience that exists only long enough to get something done?

Imagine a person asking for help comparing subscriptions. Their agent could assemble a comparison in their language, emphasize the conditions they care about, and display a confirmation only when a decision is needed. The interface might exist only for that interaction and never be seen again.

Let's move beyond imagination here for a moment.

Try the simulation. Choose a place and a goal, use the experience that appears, and finish the task.

The example is an illustrative design, not rules about everyone in a country. This is a scripted simulation, with no live model. It demonstrates the lifecycle: intent arrives, an experience forms, the goal is reached, and the interface evaporates.

Where would the localization handoff have happened?

The simulation walks that loop without a live model. One tap telegraphs place and need together - Japan to buy, Germany to learn, and so on.

What assembles is not only another language. It is an experience mapped to cultural expectations. Change its language or level of detail while keeping your progress, finish the task, and watch the interface disappear. The handoff question stays.

The shift is not only that an interface can be generated, it is the whole cycle.

• Intent is expressed.
• An experience forms around that intent and around personalization for the person - language, region, what they care about.
• The goal is reached.
• The interface evaporates.
• The surface waits for the next signal.

Google Research and others are treating generative UI as a first-class interaction pattern, not a novelty panel.[1][5] MCP's Apps extension moves interactive surfaces into the conversation itself.[2] The product metaphor is changing from a place you visit to a moment that appears.

There may be reusable components, approved text, and fixed controls inside generated experiences. Hybrid designs will matter. But some of the language will never have had an English source string waiting in a translation queue.

That is a substantial change for an industry organized around identifiable source content.

Capabilities that bypass your interface

An agent can use your service without navigating your product. Through MCP or another tool connection, it can reach a capability and make it useful inside someone else's experience. MCP can also deliver interactive interfaces through its Apps extension.[3]

The important shift is that your own interface is no longer a necessary point of contact.

Your carefully localized navigation and help center may never enter the exchange. Your capability could appear in a personal assistant's comparison, a generated form, or a task involving several providers. You would never see that combined experience.

The agent still needs to understand what your service does, its prerequisites, and its limits. We have spent decades helping people understand products. We need that knowledge to travel with the capabilities agents use.

That is why we internationalized our website for agents. Our first adaptation served visitors that needed facts and usable product knowledge, while the human interface remained English-only.

It also represents where we expect a future language experience to be produced.

A person brings their own language layer

The third change is particularly disruptive to the economics of language coverage.

A company traditionally decides which languages to invest in. People outside that selection either work through the available experience or go elsewhere.

Personal translation gives them another answer.

Apple's Live Translation with AirPods demonstrates that. After the required language models are downloaded, processing takes place on the paired iPhone.[4]

Phones, earbuds, and glasses bring interpretation closer to the individual. Hardware will vary. The meaningful change is that the person can arrive with a language capability you did not supply.

Their assistant may translate your information, summarize it, or build an explanation around what matters to them.

In my Three-Player Model, I called personal helpers Digital Yōkai. The receiver becomes an active participant in adapting the communication - on their own terms.

For some tasks, that person will no longer need a provider to publish a version in their language. They just need the underlying service to work for them.

The work moves upstream and into the experience

Understanding a service is only part of using it. An assistant can explain an English payment screen in Japanese. It cannot make an unsupported currency work or make an ineligible customer eligible.

Internationalization becomes more consequential. The underlying system must handle global data, writing directions, dates, currencies, and the markets it promises to serve. Those requirements reach into the tools and information agents use.

Cultural judgment also needs a place in generation and evaluation. Future i18n must support this. Fluency can conceal a misleading promise. Personalization can omit an important condition. A generated experience can reproduce assumptions that fail for the person using it.

Authored localization will continue where deliberate wording and review matter - regulated communications, contracts, etc.

Those distinctions are why I expect internationalization to become more important as conventional localization production contracts. Systems still need to handle global data, language, dates, currencies, writing directions, market requirements - and cultural expectations. Those responsibilities extend into the tools and information agents use.

We will need internationalized ways to define intended meaning, test behavior, and understand failures across many potential experiences.

Maybe we need people with degrees in cultural engineering and cultural design to achieve this.

What I expect to survive

Authored localization will remain valuable where exact wording, deliberate creative choices, stable experiences, or explicit review matter.

Handcrafted translation may become more like calligraphy or bookbinding in some settings: a deliberate choice with its own value. That analogy has limits. Large areas of authored content will continue to need industrial-scale language work.

Adoption will also vary with connectivity, device access, trust, and language quality. If agents cannot earn enough trust or generated experiences cannot achieve reliable quality, the transition will take longer.

For those of us in localization and internationalization, there is an opportunity to shape the transition. We understand how language, culture, systems, and business requirements interact. That knowledge can inform the rules, data, and evaluations behind experiences we will never review one screen at a time.

I would rather put our expertise to work shaping the systems, knowledge, and tests behind the new experiences.[6] The customer may bring their own interpreter, interface, and assistant. We should be preparing for that customer now.

That is where I would put the next seven years of work.

---

Sources

1. Google: Gemini 3 brings upgraded smarts and new capabilities to the Gemini app, November 18, 2025.
2. MCP: Server primitives overview.
3. MCP: Bringing UI Capabilities to MCP Clients, January 26, 2026.
4. Apple Support: Use Live Translation with your AirPods.
5. Google Research: Generative UI.
6. Julia Díez: The Interface That Writes Itself, October 7, 2026.

-Ben Cornelius