Why is iPhone dictation so bad?

Dictation on an iPhone is good in some conditions and poor in others, so the conditions are the first thing to check. The main causes are the input: how far the microphone is from your mouth, how much noise is in the room, the language the keyboard is set to, and words a general recognizer has seen few times, such as names and technical terms. Apple's pages do not publish an error rate for the Dictation key, so the share of the problem that comes from the engine itself cannot be measured from those pages. Test it with your own voice before you decide it is bad.

Distance and noise raise the error rate

Speech recognition is far more accurate when the microphone is close. In a 2017 study by Ravanelli and colleagues, close-talking Wall Street Journal sentences recorded in a studio gave a word error rate of 3.7 percent ¹. The same kind of recognizer, working from a single distant microphone in an apartment, gave 12.0 percent ². The error rate went from 3.7 percent to 12.0 percent. The two conditions were recorded in different places, so treat the comparison as a rough guide, not an exact law.

Noise adds to the distance effect. The authors of a 2017 paper on conversational speech recognition name speech fidelity, ambient noise and microphone quality among the factors that affect accuracy ³. A phone held at arm's length in a kitchen with a television on is closer to the distant condition than to the studio one.

Apple's older speech class worked well for short dictation

Apple's 2025 developer session says its older speech class "worked well for short-form dictation" ⁴. That session describes a developer interface, and it does not say which engine the Dictation key uses. It does not say the class was built for short phrases either. A recognizer that works well on short dictation may do less well on a long, unbroken one, though Apple's session does not test that case.

The one independent test puts Apple behind Whisper

In a 2025 9to5Mac test, one 7-minute-31-second episode gave Apple's speech interface a word error rate of 10.3 percent, against 1.5 percent for Whisper Large V3 Turbo ⁵. It is the best public measurement we found, and it has three limits. It measured Apple's developer interface, not the Dictation key. It is one recording by one author. And Apple has changed how dictation works since that test, as the iOS 27 change below shows ⁶. It shows a direction and a rough size, not the accuracy you will see in a text message.

Has it gotten worse? We found no measurement that shows it

Users have complained online that iPhone dictation has declined, and some complaints name particular iOS versions. We found no test that measures the same engine before and after an update, and no Apple statement that accuracy went down. The complaints are anecdotes, so they cannot settle the question either way.

A newer change, which Apple describes, targets spelling, punctuation and capitalization. With Apple Intelligence in iOS 27, "dictation uses an on-device model to improve spelling, punctuation, and capitalization" ⁶. That model is limited to English, and to the iPhone 17 Pro, iPhone 17 Pro Max, iPhone Air or later ⁶. If you use an older iPhone or another language, the change does not apply to you.

It depends on your language, your iPhone and what you say

Language is one of the causes you can check directly. Apple links Dictation to the keyboard language: "when you use a keyboard for another language, you can use Dictation in that language as well" ⁷. A sentence that mixes two languages, or a keyboard set to the wrong one, can produce errors that have nothing to do with the microphone.

The iPhone model matters only for the newest English model described above. For other languages and older phones, the change in iOS 27 does not apply.

Names and technical words are a common weak spot, because a general recognizer has heard few of them. Expect errors in product names, medical or legal terms, and code.

Compared with typing, speech can be much faster. In a controlled study on an iPhone, English entered by speech reached 153 words per minute, against 52 words per minute on the built-in keyboard ⁸. That speed counts the time spent correcting errors. The study used Baidu's Deep Speech 2 recognizer, not Apple's dictation, with short phrases in a quiet room, so a long, noisy session may be slower and less accurate than the test.

What to do: test your own voice, then fix the input

1. Open Notes, read a 100-word passage you would really dictate, and count the words that come out wrong. Divide the count by 100 for a rough error rate. The measure is explained in our glossary entry on word error rate.
2. Hold the phone close to your mouth, speak at a steady volume, and move away from the television or other people who are talking.
3. Set the keyboard to the language you are speaking, and dictate one language at a time.
4. If the error rate is still high in a quiet room, the engine or the language is the likelier cause, so test a different app on the same passage before you decide.
5. If the key does not work at all, rather than producing wrong words, follow the guide to fixing iPhone dictation, which starts with the Enable Dictation switch.

Yap, which we make, is an on-device dictation app for Mac and Windows, so the same passage test works for it as well. Use the same passage and microphone for every app you compare.

Related questions

• Why does iPhone dictation stop after a few words?
• Why did dictation stop working after an iPhone update?
• Is Apple Dictation on Mac any good?
• Does Apple dictation work offline?

For the full set of steps on iPhone, read how to use dictation on iPhone. The overview of this subject is dictation and voice to text on iPhone and Android.