The doubled zero-width space in Android translations, and why it clusters after "EXIF"

Description and use

Invisible U+200B characters appear in Android translations, especially after "EXIF", breaking search and comparison. 703 strings across 122 repos and 56 languages were found. l10n-guard detects and removes them.

Problem

Invisible zero‑width spaces (U+200B) appear in Android translation strings, breaking search, comparison and wrapping.

Outcome

Zero‑width spaces cluster after "EXIF" in translations, occurring 1.59% of the time versus 0.0004% after other capital words; 3,900× more common. The issue is isolated to specific translation batches, not entire projects.

Workflow

• Count invisible characters in all translation strings
• Identify strings with only U+200B
• Analyze pattern of space + two U+200B + word
• Measure frequency after "EXIF" versus other capital words
• Publish predictions and verify
• Provide l10n‑guard integration for CI detection
• Recommend removal of U+200B except in scripts that need them

Evidence

Seven in ten of those 703 strings have the same pattern: an ordinary space, then exactly two U+200B, then a word

Source

"EXIF" + space + word 2,517 units, 40 with the pair, 1.59%

Source

The pair is about 3,900 times as common after "EXIF" as after any other short capital word

Source

l10n-guard has flagged them in CI since 0.15.0 (invisible-char warning)

Source