AI vs Human Localization
Can consumers tell the difference between AI-localized and human-localized advertising? A bilingual perception study across English and Turkish, conducted as part of my MSc Digital Marketing dissertation at BSBI/UCA Berlin.
A bilingual perception test
This study was conducted as part of my MSc Digital Marketing dissertation at BSBI/UCA Berlin (2026). The core question: when a premium automotive ad is localized by AI versus a human, can consumers actually perceive a difference? And does it affect how they evaluate the message?
Two versions of a Porsche advertisement were created for each language (English and Turkish). One was localized by a professional human translator. The other was localized using a generative AI large language model with specific cultural and tonal prompting. 76 participants evaluated both versions on five-point Likert scales across five perceptual dimensions: naturalness, native-speaker quality, authenticity, emotional engagement, and overall appeal. Participants did not know which version was AI-generated.
Five constructs across four conditions
Where perception diverged
English Condition
Human and AI versions scored almost identically. Perceived naturalness was tied at 3.94 out of 5. Human copy scored slightly higher on authenticity (3.94 vs 3.66), while AI copy scored marginally higher on native-speaker quality and overall appeal. Respondents showed ambiguity in authorship attribution, unable to reliably identify which ad was AI-generated.
Turkish Condition
AI-localized copy outperformed the human version across every single dimension. The largest gap appeared in emotional engagement (2.89 vs 2.48). Turkish respondents showed higher overall sensitivity to localization quality. Notably, participants more often attributed human authorship to the AI version and AI authorship to the human version.
Practical implications
- Perceived authenticity depends on language quality and cultural fit, not on who or what wrote the copy.
- AI localization can match or exceed human output when given proper prompting and cultural context, especially for structured advertising content.
- Human expertise remains essential for high-stakes creative work involving cultural metaphors, indirect meaning, and nuanced tone shifts.
- A hybrid workflow (AI for drafts and volume, human review for emotionally sensitive content) is both viable and efficient.
- The question is no longer "who wrote it?" but "does the message land?"
Scope and constraints
Self-reported perceptions, not behavioral conversion data. Convenience sample of 76 respondents via voluntary participation. Single product category (premium automotive). Results indicate perceptual tendencies rather than statistically generalizable conclusions. The study measures perception, not purchase intent.