Why do dictation apps get words wrong?

Dictation apps get words wrong mainly because the engine under the app is imperfect, and users tell us it shows most with unusual words, accents and sloppy punctuation. Our data measures how often people report these problems, not why they happen. Recognition accuracy is a problem in 10.9 percent of the App Store reviews we coded ¹, and in 21.0 percent of reviews of 1 to 3 stars ².

Accuracy is a common complaint, but not the biggest one

Reviewers are more likely to complain about reliability than about wrong words. Among critical reviews, reliability appears in 42.9 percent ³, double the accuracy figure. Still, accuracy is raised in every source we coded:

• Hacker News comments: 12.1 percent ⁴
• Bluesky posts: 14.7 percent ⁵
• Transcription apps on the App Store: 12.9 percent ⁶
• Professional dictation apps on the App Store: 10.1 percent ⁷

Accuracy is also the most praised theme, so the same quality that frustrates some users delights others. The rows below put problem and praise side by side for every source and group we coded.

Showing 8 of 14 rows. See every row in the dataset.

Names, special words and punctuation are named less often

Users who name a cause usually point to words the engine has not seen often. Names and special words are a problem in 4.3 percent of critical reviews ¹² and 3.9 percent of professional dictation reviews ¹³. Jargon is a likely cause, though reviewers rarely explain why a word went wrong.

Punctuation and formatting is a problem in 4.4 percent of transcription app reviews ¹⁴ and 3.9 percent of critical reviews ¹⁵. Right words in the wrong places still count as a mistake to the person reading.

Accents and languages affect a minority of reviews

Languages and accents are a problem in 6.4 percent of critical reviews ¹⁶, 6.1 percent of transcription app reviews ¹⁷ and 3.6 percent of all reviews ¹⁸. The small shares may partly reflect who writes reviews on English-language storefronts. They are no proof that accents are handled well.

Background noise appears only as praise in these rows

Among the rows we hold, background noise appears as praise, not as a problem: 4.5 percent of meeting note taker reviews ¹⁹ and 3.1 percent of transcription app reviews ²⁰ praise how an app copes with it. We do not have a measured share of reviews that blame noise for errors, so we do not rank it as a cause.

The chart below shows what Bluesky posters report as problems, where accuracy is the one we can measure most directly.

The engine matters more than the app's name

A 2025 test by a Mac news site compared engines on one recording. Apple's speech API had a word error rate of 10.3 percent ²¹, against 1.5 percent for Whisper Large V3 Turbo ²². It was one episode and one speaker, so treat it as an example of how far engines can differ, not as a ranking.

What the numbers do and do not measure

• Reports, not tests. Our rows count how many documents mention a problem. They do not measure error rates or find causes.
• Not a random sample. App Store pages show reviews Apple selects, often the most helpful. Hacker News comments come from developers and founders.
• Stated problems only. A theme counts only when the document states it.
• Small groups, small shares. Several themes sit at a few percent, so differences between them are weak evidence.

Documents that used the words in another sense were set aside, and groups of fewer than 30 documents are left out. See the full dataset.

What this means for getting fewer mistakes

1. Try your own vocabulary. Names, product terms and jargon are where users report trouble, so test the words you use most.
2. Test with your accent and your room. Reports on these are sparse, so check yourself.
3. Compare engines, not logos. Measured error rates can differ widely between engines. See the market overview for how products compare.
4. Expect to correct punctuation. Check how an app handles it before you rely on it.

More on what users report is in the most common dictation app problems. For background, see our overview of dictation software and the guide to the best dictation software.