Translated Is Not Localized: 56% of Multilingual Shoppers Have Walked Away From a Purchase

Most content teams treat a second language as a box to tick. Run the site through translation, check that the buttons say the right thing, ship it. New survey data suggests that box is costing sales even when the translation itself is perfectly good. More than half of multilingual shoppers have walked away from a purchase because a site did not feel built for them, and a majority of those who praised the translation walked away anyway.

The Translation Passed. The Sale Still Failed.

Liferay's 2026 Website Localization Report, released on October 13, surveyed 800 US adults through Pollfish in September, all of whom had used a website, app or digital service in a language other than English in the previous 12 months. 56% said they had decided not to buy from a company because its site or app felt poorly adapted to their language, and 48% said they had chosen one company over a competitor because its experience fit their language and culture better.

The more uncomfortable number sits underneath. Of the respondents who rated a brand's non-English translations as good or excellent, 54% still abandoned a purchase over poor localization. Only 23% said non-English sites usually feel like they were written for them, and 90% had noticed at least one localization problem on a site or app. The words were right. Something else was wrong.

Liferay's co-founder and CMO, Bryan Cheung, drew the line in the release: translation moves content into another language, while "localization is what makes it feel natural to the people using it." That is a vendor's framing, and the survey covers US adults only, so the figures are best read as indicative. But the distinction maps neatly onto what content teams actually ship: copy that has been converted, sitting inside imagery, examples, offers and checkout flows that were never reconsidered for the person reading them.

Machine Translation Is the Easiest Tell to Spot

Generative AI has made translation nearly free, which is precisely why it has stopped being a differentiator. In the Liferay data, 82% of respondents said they are confident they can tell when content was machine-translated rather than written or reviewed by a fluent speaker. That is self-reported confidence rather than a blind test, but it is the perception that shapes trust. Only 20% credited AI alone when a brand's localization was strong, while 43% credited human involvement.

None of this is a new preference. CSA Research's 2020 "Can't Read, Won't Buy" study, run with Kantar across 8,709 verified consumers in 29 countries, found 76% prefer product information in their own language and 40% will not buy from websites in other languages at all. 73% wanted product reviews in their language, "if nothing else," and CSA estimated that companies which fail to localize the buying experience risk losing 40% or more of their addressable market. What has changed since then is the volume of cheap, unreviewed translated content now competing for the same reader.

This Is a Domestic Audience, Not an Export Market

It is tempting to file localization under international expansion. The Liferay sample is entirely US based, and the domestic numbers are large. The Census Bureau's most recent detailed language tables, covering 2017 to 2021, found more than one in five US residents aged five and older, 22%, spoke a language other than English at home, across more than 500 languages and language groups. About 61% of those speakers used Spanish. And 62% also spoke English "very well," which matters: this is not an audience that needs translation to understand the site. It is an audience choosing where it feels addressed.

The spending power behind that choice keeps growing. The 2026 U.S. Latino GDP Report from UCLA and California Lutheran University put US Latino economic output at $4.4 trillion in 2024, the fourth-largest GDP in the world if it stood alone and nearly 9% larger than Japan's economy. Real Latino GDP grew 6.4% that year, against 2.4% for non-Latino GDP, and the population now tops 68 million, one in five US residents. A content programme that treats Spanish as a translated afterthought is underserving one of the fastest-growing consumer markets anywhere, without leaving the country.

Read together, the data points to a content problem more than a language problem. The 54% who praised the translation and still left were reacting to everything around the words: an example that only works in one culture, a hero image that ignores them, a promotion built around a holiday they do not keep, a review section with nothing in their language, a checkout that does not offer the way they prefer to pay. Each of those is a content and experience decision, and each one is usually made once, in English, and never revisited.

• Stop measuring localization by translation quality alone. More than half of shoppers who rated the translation good still walked away, so test localized journeys on conversion and abandonment, not on linguistic accuracy.
• Put a fluent reviewer between the model and the page. 82% of multilingual consumers believe they can spot machine translation, and only 20% credit AI alone for localization they rate as strong.
• Localize the assets, not just the copy. Imagery, video, examples and offers carry most of the cultural signal; a system that adapts them per market at scale is what separates localization from translation.
• Bring reviews and social proof into the same language. CSA found 73% of consumers want product reviews in their own language, even when nothing else on the page has been adapted.
• Carry localization through to checkout. A localized product page that ends in an unfamiliar payment flow or currency display undoes the work at the point where the sale is decided.
• Size the domestic opportunity before the international one. With 22% of US residents speaking another language at home and Latino GDP at $4.4 trillion, the first localization market may already be in your existing audience.