Translated text does not fit the box it came from. IBM’s published guidance, cited by the W3C, puts expansion for European languages at 200 to 300% for source strings up to 10 characters, and 130% for strings over 70 (w3.org, accessed September 2026). Short strings expand most, and ad copy is almost all short strings. A layout built to fit English breaks in German.
The campaign is approved in English. Twelve markets need it. The designer rebuilds the layout twelve times.
This covers why expansion breaks fixed layouts, the four things that must stay variable, and how to review twelve languages when you read one.
Key Takeaways
- Short strings expand most, and ads are short strings. Up to 200 to 300% for source text under 10 characters, tapering to 130% above 70 characters, per IBM guidance cited by the W3C.
- A button label is the worst case, not the best. “Buy” and “Views” are exactly the strings that double or triple.
- Design the layout for the longest language, not for English. If it fits German, English has room. The reverse never holds.
- Four things must stay variable: the text as live text, the text box with room to grow, the layout direction, and the fonts with full character coverage.
- CJK languages contract rather than expand, which creates the opposite problem of a box that now looks empty.
- Never bake text into an image. It cannot be translated without rebuilding the asset, which is the entire cost this article is about.
- Review needs a native speaker checking fit, not only meaning. A correct translation that overflows is still a broken ad.
Why does translated text break a layout?
Because expansion is largest exactly where ads have the least room.
The published guidance is counterintuitive at first reading. Expansion is not a flat percentage. It scales inversely with source length:
| Source length | Expansion |
|---|---|
| Up to 10 characters | 200 to 300% |
| 11 to 20 characters | 180 to 200% |
| 21 to 30 characters | 160 to 180% |
| 31 to 50 characters | 140 to 160% |
| 51 to 70 characters | 151 to 170% |
| Over 70 characters | 130% |
From IBM’s guidelines for designing global solutions, as published by the W3C (w3.org, accessed September 2026). The W3C notes the principle plainly: “the smaller the source message, the higher the likely translation length”, and gives a worked example where European translations of the word “views” ran from 2.6 to 3 times the English length.
Now map that onto an ad. A headline is often under 30 characters. A button is under 10. Meta’s own guide gives a 27-character headline for feed image ads (facebook.com/business/ads-guide, accessed September 2026). Those are precisely the strings sitting in the 160 to 300% band, and they are the ones placed in the tightest boxes in the design.
The result is predictable. The body copy usually survives. The headline and the button, which carry the message and the action, are what overflow.
What has to stay variable?
Four things. Each one, if fixed, forces a rebuild per language.
Text as live text, never baked into the image. This is the one that decides everything else. Text rendered into a JPEG cannot be translated without remaking the asset, which means the twelve-market campaign is twelve design jobs rather than twelve string swaps. Video has the same problem with supers burned into the frame, and video localization for fashion ads covers that layer along with captions and voice.
Text boxes with room to grow. Size the box for the longest expected language and let English sit with space around it. If the layout only works when the headline is exactly two lines in English, it does not work.
Layout direction. Arabic and Hebrew read right to left, which mirrors the composition rather than reflowing it. If those markets are on the roadmap, decide early, because retrofitting is a redesign.
Fonts with full character coverage. A typeface without the accented characters, or without Cyrillic or CJK glyphs, produces missing-character boxes or a silent substitution to a different font. Check coverage per market before the campaign, not at export.
What about languages that get shorter?
They create the opposite problem, and it is easier to miss.
Chinese, Japanese and Korean generally render the same meaning in fewer characters than English. A headline sized for a German worst case looks sparse in Japanese, and a layout balanced around a two-line headline becomes visually unbalanced with one short line.
Rather than a separate layout, design the composition so the text block can shrink without the whole thing collapsing: anchoring the text to a consistent edge instead of centring it in a fixed box, and letting the surrounding space absorb the difference.
CJK also brings requirements beyond length. Line-breaking rules differ, font sizes often need to be larger for legibility at the same visual weight, and the fonts themselves are large files. None of that is solved by a translation step.
How do you review twelve languages when you read one?
Split the review into two questions and give them to different people.
Does it fit? Anyone can check this, in any language. Does the text overflow, get cut, collapse to an unreadable size, or leave the layout unbalanced. Does the button label still fit the button. Does anything overlap the safe zone. This is a visual check and it catches most failures.
Does it read? This needs a native speaker, and it is a different question from whether the translation is accurate. Ad copy carries tone, idiom and a call to action, and a literal translation of a phrase built on English wordplay is usually correct and useless.
A workable process:
- Export every language as an image set and view them together in a grid rather than one at a time. Fit problems are obvious side by side and invisible in isolation.
- Check the shortest strings first. Buttons and headlines are where expansion concentrates.
- Send to a native speaker with a specific question, not “does this look right”. Ask whether the call to action sounds like something a person would say.
- Keep a per-market string length limit once you learn it, and write it into the brief so the English copy is written to survive translation.
That last point saves the most time. If you know the headline box holds 27 characters in English and German expands to roughly double, the English headline needs to be written short enough that its German version fits. That is a copywriting constraint, decided before design, and it belongs in the creative brief.
How do you produce twelve markets without twelve rebuilds?
Separate the layout from its contents, then swap the contents.
The structure that works is the same one that solves aspect ratios: the product as its own layer, the background extended past the frame, the logo placeable, and the text live. Once those four are separate, a market is a string swap and a box resize rather than a design job.
Then make the parts again rather than rebuilding the whole. In DesignerBox, the ad localizer rebuilds one ad in another language. It redraws the layout. A market becomes a run instead of a design job. Once the first market is right, a saved workflow runs the same way on the next one, and a template for ad variations gives you a starting point for the set. The layout stays constant, the copy layer changes per market, and the cost is shown before the run. The commercial licence starts on Pro, at $35 a month billed monthly.
Two honest limits. Machine translation alone is not a localisation strategy: a native review pass is the part you cannot automate. And a market with a right-to-left script needs the layout decision made upfront, not retrofitted.
Do the work once and it stops being a per-market job. The box sizes, the fonts, the safe zones and the review checklist you settle on for the first approved concept are the same ones the next campaign needs, and the one after that. Save them as a workflow. You set the brand once, and the workflow reads it on every run, so the next campaign gets the same standard. A per-asset rebuild never gives you that.
The same master-asset discipline covers the other multiplier, which is one ad across every platform size. Solve them together, because they are the same structural problem pointing in two directions. The volume that results is covered in ad variants at scale.
FAQ
How much longer does translated text get?
For European languages, IBM’s guidance cited by the W3C gives 200 to 300% expansion for source strings up to 10 characters, 160 to 180% for 21 to 30 characters, and around 130% for strings over 70. Expansion is largest on the shortest strings, which is why buttons and headlines break before body copy.
Which languages expand the most from English?
German is the commonly cited worst case among European languages, along with Finnish, Polish and Russian for many strings. The bigger factor is source length rather than target language: a three-word button label can double or triple in almost any European language, while a long paragraph expands by around a third.
Why do my ads break when translated?
Because the headline and button sit in the tightest boxes and expand the most. A layout that fits an English headline exactly has no room for a version that is 60 to 200% longer, so the text overflows, gets cut, or shrinks below legibility. Body copy usually survives because it expands least.
Should I design ads for English first?
Design the layout for the longest language you will ship, then let English sit with space around it. A layout that fits German will fit English. A layout that fits English exactly will not fit German, and that direction cannot be fixed without redesigning.
Do Chinese and Japanese ads have the same problem?
The opposite one. CJK languages generally use fewer characters for the same meaning, so a box sized for German expansion looks empty. They also bring different line-breaking rules, larger legible font sizes, and large font files. Anchoring text to an edge rather than centring it in a fixed box handles both directions.
Can I use machine translation for ad copy?
For a first pass and for fit-checking, yes. As the final step, no. Ad copy carries tone and a call to action, and a literally accurate translation of an English phrase often reads as something no native speaker would say. The review that matters asks whether the call to action sounds natural, not whether the words are correct.
How do I avoid rebuilding the layout for every market?
Keep four things variable: the text as live text rather than baked into the image, text boxes sized for the longest language, the layout direction where right-to-left markets are in scope, and fonts with full character coverage. With those separated, a new market is a string swap rather than a design job.
Sources
- W3C, Text size in translation, citing IBM Guidelines to design global solutions (w3.org, accessed September 2026)
- Meta ads guide, Facebook feed image ads (facebook.com/business/ads-guide, accessed September 2026)
Text expansion figures verified from W3C published guidance as of September 2026. Expansion varies by language, string and context. Individual results vary.
