Fact · Transcription
1percent
In a 2024 study, roughly 1 percent of Whisper audio transcriptions contained entire hallucinated phrases or sentences that did not exist in the audio.
Checked against its source on October 16, 2026
In the source's own words
“roughly 1\% of audio transcriptions contained entire hallucinated phrases or sentences which did not exist in any form in the underlying audio”
Careless Whisper: Speech-to-Text Hallucination Harms, Allison Koenecke et al., arXiv (ACM FAccT 2024), 2024, Abstract.
Open the original- What it measures
- Whisper was the 2023 OpenAI service, tested on recordings from speakers with and without aphasia.
- What it does not show
- One dataset and one 2023 version; later versions may behave differently.
Source
Cited on
- Dictation for Work: Doctors, Lawyers, Writers, StudentsTopic
- Speech Recognition: What It Is and How Well It WorksTopic
- Transcription: How to Turn Audio and Video Into TextTopic
- Speech to Text Privacy: Where Your Voice GoesTopic
- Speech to Text AI: Dictation, Assistants and Voice AgentsTopic
- How Speech Recognition Works, From Sound to TextGuide
- How to Transcribe Audio to TextGuide
- Meeting Transcription: How It Works and What to UseGuide
- What Is Whisper AI? OpenAI's Speech Model ExplainedGuide
- Speech-to-Text APIs Compared: Price, Accuracy, SpeedGuide
- Is Whisper AI Safe? Privacy and Security ExplainedGuide
- Video to Text: Transcribe Videos and YouTubeGuide
- How Accurate Is Speech Recognition? The Real NumbersGuide
- Speech to Text and LLMs: What Language Models ChangeGuide
- Best Dictation Apps for iPhone and AndroidGuide
- Speech Recognition Models: How Neural ASR WorksGuide
- Speech to Text Without AI: Dictation With No AI RewriteGuide
- Transcription Software: How to Choose OneGuide
- Legal Dictation Software for Lawyers and Law FirmsGuide
- Open-Source Speech Recognition: The Best ModelsGuide
- How to Run Whisper Locally on Your Own ComputerGuide
- Legal Transcription: The Job, Formats and PayGuide
- Speech to Text on Linux: Dictation on Ubuntu and MintGuide
- Transcribe Audio With Whisper, ChatGPT or GeminiGuide