Human perception of audio deepfakes: the role of language and speaking style
Source record
san-segundo-2026-audio-language-style· 2026 · Speech Communication
Evidence tier: B (peer-reviewed-journal) · Verification: verified-abstract (compiler confidence: medium)
Bottom line: Spanish and Japanese listeners averaged 59.11% when classifying natural vs synthetic speech, doing better on real voices, and relied mainly on prosodic and extralinguistic cues (intonation, rhythm, fluency, pauses, breathing, laughter).
Citation
Eugenia San Segundo, Aurora López-Jareño, Xin Wang, Junichi Yamagishi (2026). Human perception of audio deepfakes: the role of language and speaking style. Speech Communication, 185, 103487. https://doi.org/10.1016/j.specom.2026.103487
Verification entry (identifiers, resolvers, claims with source locations): /data/citations.json#san-segundo-2026-audio-language-style
Study design
| Field | Value |
|---|---|
| Design | Perception experiment: listeners classified voices as natural or synthetic and justified each choice; stimuli varied by language (Spanish, Japanese), speaking style (audiobook, interview) and voice familiarity (familiar, unfamiliar) |
| Modality | audio |
| Stimuli | 80 stimuli, 50% artificial (synthesis system not in abstract) |
| Participants (N) | 54 |
| N detail | 28 native Spanish speakers, 26 Japanese speakers |
| Topics | human-detection, audio, cross-linguistic, cues |
Results
| Metric | Value | Note |
|---|---|---|
| Average accuracy | 59.11% | better on authentic samples |
| Cues used | mainly suprasegmental / extralinguistic | intonation, rhythm, fluency, pauses, speed, breathing, laughter rather than segmental features |
| Cross-linguistic | shared cues plus notable differences in how 'humanness' is conceptualised | Japanese vs Spanish listeners |
Moderators
- Listener language
- Speaking style
- Voice familiarity
Limitations
- Values not checked against full text
- Small N
- Condition-level accuracies not in abstract
Cited in
- MR-0001: Human Detection of Deepfakes: Accuracy, Bias, Confidence, and Training
- MR-0003: Voice Clones and Synthetic Speech: How Well Do People Detect Them?
Verification
| Field | Value |
|---|---|
| Status | verified-abstract |
| Compiler confidence | medium |
| Checked on | 2026-10-18 |
| Via | OpenAlex abstract record for doi:10.1016/j.specom.2026.103487 |
Values on this page have not all been checked against the full text. Confirm them against the primary source before high-stakes use.