Fact · Voice and AI
44percent average transcription error rate
In a 2026 study, 15 speech models from OpenAI, Deepgram, Google and Microsoft had an average transcription error rate of 44 percent on US street names spoken by US participants.
Checked against its source on October 17, 2026
In the source's own words
“We evaluate 15 models from OpenAI, Deepgram, Google, and Microsoft on recordings from linguistically diverse U.S. speakers and find an average transcription error rate of 44%.”
Sorry, I Didn’t Catch That: How Speech Models Miss What Matters Most, Kaitlyn Zhou et al., arXiv, 2026, Abstract.
Open the original- What it measures
- The authors say the errors matter because a wrong street name can send an error into a routing decision. Their study is about short, high-stakes utterances.
- What it does not show
- A preprint on street names only. It shows that a low overall word error rate can hide failures on the words an agent needs.