What is speech recognition used for?

Speech recognition is used wherever turning speech into text saves time or opens access to what is said. The main uses are writing documents by voice, following live speech, controlling a device, making calls for people who cannot hear well, and creating records in medicine and law. Each use has a different tolerance for errors, which decides how much you can rely on it.

Dictation and transcription turn everyday speech into text

People dictate documents and email instead of typing them, and recordings can be transcribed afterward. Law is a clear example: the 2026 draft of Jurafsky and Martin's textbook says transcription is important in fields like law, where dictation plays an important role ¹. Yap, which we make, is an on-device dictation app for Mac and Windows, and its recognition runs on your computer. For dictation on a Mac, see offline dictation on Mac.

Live captions and relay calls serve people who cannot hear well

Apple says its Live Captions makes real-time transcriptions of conversations on the device, so you can follow them on the device itself ². In internet protocol captioned telephone service (IP CTS), a communications assistant repeats what the other party says, and voice recognition software transcribes those words ³.

Assistants and devices use recognition to act on speech

The same textbook says speech is a natural interface for appliances, digital assistants and chatbots, especially on cellphones, where keyboards are less convenient ⁴. It can also serve people whose speech is affected by a condition, as the study below shows.

Medical transcription is shifting toward reviewing machine drafts

The US Bureau of Labor Statistics says advances in speech recognition and natural language processing let physicians document patient encounters in real time, which reduces the need for medical transcriptionists ⁵. The same outlook lists reviewing and editing drafts prepared by speech recognition software among a medical transcriptionist's usual duties ⁶. Read together, the two statements describe a role that shifts toward reviewing machine drafts. A 2024 nursing review suggests that speech recognition, especially combined with artificial intelligence, can speed up documentation by up to 15 percent ⁷. That is an upper figure, not an average.

It depends: when speech recognition is weaker

Accuracy can drop for some accents. In the EdAcc test of accented English, every English speech recognition model tested showed a drop in performance on speakers of Indian, Jamaican and Nigerian English ⁸. Speakers matter too. In Google's Project Euphonia study of 432 people with disordered speech, personalized recognition models reached a word error rate of 4.6 percent, against 31 percent for general models ⁹. A general model had not learned those voices, so the same system can be useful to one speaker and poor for another.

What to do

1. Pick the use first, then the tool. For a document, favor a dictation app. For a live conversation, favor captions. For a medical or legal record, plan for a human review step.
2. Measure accuracy on your own audio with word error rate, using a passage you know by heart, before you rely on a system.
3. For medical or legal use, read the vendor's privacy and compliance terms before you send any audio to it.

Related: Is speech recognition AI?, Is speech recognition part of NLP? and Speech vs voice recognition: what is the difference?. For the basics, read our guide to how speech recognition works and the overview of speech recognition.