Translation Can Support ROI. But Only When the Output Actually Performs.

DeepL's latest framing is hard to miss: generic AI is risky, agencies are slow and expensive, and enterprise teams should move to a dedicated AI translation platform like DeepL.

But that framing skips a more important question: does translation actually improve ROI?

Not automatically.

A translated version may increase traffic for a while. It gives people something to find, click, and land on. But traffic is not ROI.

Translation can improve ROI if the localized content increases trust, comprehension, conversion, retention, or user action. But if a company spends money to produce a Japanese version that does not work for Japanese users, translation has not created value. It has added cost.

Making that cost smaller does not turn it into ROI.

This is why localization ROI should not be measured only by translation speed, cost per word, traffic, or consolidation of the translation stack. The real question is simpler: does the localized content do the job in the target market?

For Japanese, and for many other markets, that often means more than translating the English source. It means deciding what should be preserved, what should be changed, what should be removed, and what needs to be generated differently.

Translation can support ROI. But only when the output actually performs.