The Translated Claim That Reads Perfectly and Is Still Wrong
Translators say beauty brands increasingly buy translation as a quick check on a machine draft. In a roundtable, eight translators who work on cosmetics copy describe a check that takes more skill than it is priced for, and errors that hide in the copy that reads best.
"The most difficult mistakes are often the ones that sound perfectly natural," says Serbian translator Ivana Lojanica. On a cosmetic pack, a natural-sounding mistake is still the brand's claim, and under Commission Regulation (EU) No 655/2013 it still has to be truthful, honest and supported by evidence.
In September The Beauty Insider sent the same nine questions to twelve translators. Eight answered in writing, covering Spanish, European Portuguese, Italian, German and Serbian. All eight work with Lingua Beauty, a beauty translation agency for cosmetics brands and distributors.
The draft is cheaper. The check is not
"Around 90% of the projects I receive involve post-editing," says Spanish translator Estíbaliz Alba. Joana Teixeira, a European Portuguese translator, says the work has changed shape: "The biggest difference is not that AI translates for me, but that it has changed where I spend my time and attention."
The price is under pressure. Teixeira: "there is a growing assumption that a translator's job is now simply to 'check' it quickly, and therefore should cost less." Ivana Bojkić, also a Serbian translator: "Clients expect us to be low-cost proofreaders." Alba says what the check often becomes: "And in many cases, that 'review' means rewriting most of the text anyway."
Where fluent goes wrong
The errors are small and they sit exactly where a brand carries risk.
In Spanish, Alba says the machine "often translates 'serum' as 'suero' or 'fine lines' as 'líneas finas', rather than using the established Spanish beauty terms 'sérum' and 'líneas de expresión'." Sofía Di Gialleonardo, a Spanish translator in Argentina and founder of Verbalia Solutions LLC: "Expressions such as 'glass skin' are another good example: translating them word for word as piel de vidrio does not necessarily communicate the same beauty concept to a Spanish-speaking audience."
Katharina Selich, a German translator, points at the least glamorous text on the pack: "It can get tricky when it comes to the exact instructions for makeup or creams."
The claim is where it costs most. "A literal translation can change how a carefully worded claim sounds, perhaps making it seem stronger or even like a medical claim," says Lojanica. Bojkić is shorter: "A literal translation can turn a cosmetic claim into misleading advertising."
Sofia Bicchiri, an Italian translator, explains why fluent output makes the risk easy to miss: "when a machine produces a fluent, convincing translation, it can be easy to overlook a small but significant shift in meaning." Di Gialleonardo goes further: "if anything, automation can amplify it because content is produced and distributed much faster."
One language, two markets
Teixeira names a failure that is easy to miss from head office: "One thing AI has made even clearer to me is that sharing a language does not mean sharing a market." Working across European and Brazilian Portuguese, she sees how easily machine Portuguese can be understood by a reader in one market and still feel foreign to them.
Where they part ways
The eight broadly agree on the risk. They disagree on the tool.
Teixeira and Bicchiri use it to speed up research and first drafts. Selich is pulling back: "Recently, though, I've started using it less because I don't want to lose my own voice." Chiara Vecchi, also an Italian translator, sees clients going both ways: "I have also noticed a completely opposite tendency in some customers who have explicitly told me not to use any form of machine translation or AI."
Looking ahead, Lojanica sees "AI as another step forward in that development, rather than a threat to us as translators." Bojkić expects that "the number of translators may decrease", though "human translators will remain an essential part of the translation process." Vecchi expects brands to split: some "will realise that they get sort of muted when they speak another language" and invest, while "Other brands will embrace AI and machine translation even more than now and this will turn into a loss for translators."
"Sharing a language does not mean sharing a market." Joana Teixeira
What this does not prove
This is a roundtable of eight working translators, not a test. None put a figure on how often machine output changes a claim, and four of the twelve we asked did not answer.
Why it matters
For a brand owner, the saving is in the draft, not the sign-off. Budget the review as translation, because that is often what it becomes.
For a claims or regulatory lead, the question is who reads the translated claim before it reaches the pack or the product page, and whether that person is accountable for it.
"For brands, the advantage will come from knowing where automation is enough and where human judgment is worth investing in," says Di Gialleonardo. Lojanica has the last word, and it is the one to remember when signing off a label: "a good translation requires someone to take responsibility for the final words. AI can be a very useful tool, but I wouldn't want clients to mistake a tool for the person accountable for the result."
The contributors
Estíbaliz Alba, English to Spanish, Spain. Sofia Bicchiri, English, Spanish and Russian to Italian, Sardinia, Italy. Ivana Bojkić, English to Serbian, Serbia. Sofía Di Gialleonardo, founder of Verbalia Solutions LLC, English and Spanish, Argentina. Ivana Lojanica, English to Serbian, Serbia. Katharina Selich, English and Norwegian to German, Konstanz, Germany. Joana Teixeira, English, Spanish and Brazilian Portuguese to European Portuguese, Porto, Portugal. Chiara Vecchi, English and German to Italian, Spilamberto, Italy.
Every quotation is verbatim from the contributor's written answers to The Beauty Insider, September 2026.







