Caption accuracy requirements
No US rule requires a caption accuracy percentage. The FCC's standards, 47 CFR 79.1(j)(2), say captions must match the spoken words; WCAG 1.2.2 asks for captions of all prerecorded audio; the DCMP Key aims for errorless captions. A caption file cannot show accuracy. On the audit, Program and Contract, a person samples it and states what the sample supports.
Status as of .
Is 99% caption accuracy a legal requirement?
No. When the FCC adopted its caption quality standards in FCC 14-12, it described them as qualitative rather than quantitative. They say what accurate captions do, not what share of words they get right.
The 99% figure most people have seen is a worked example in 47 CFR 79.1(k)(2)(iv). That paragraph is one of the FCC’s best practices for live captioning vendors, which programmers adopt by choice. It shows how to calculate accuracy as a percentage of correct words and uses 99% as the arithmetic example. It sets no threshold. A contract can set one, but that is the contract speaking, not the law.
What does the FCC require?
Under 47 CFR 79.1(j)(2), accuracy is one of four standards for captions on US television:
Captioning shall match the spoken words in their original language, and convey background noises and other sounds, to the greatest extent possible.
The rule goes on to cover spelling, punctuation, numbers, and sounds and speakers that cannot be seen. It applies to television and its clips online; the FCC’s four caption quality standards covers all four.
What does WCAG require?
WCAG 1.2.2 asks that “Captions are provided for all prerecorded audio content in synchronized media” (Understanding 1.2.2). It sets no accuracy figure. The W3C’s guidance treats captions that leave out dialogue or important sounds as failing it, so missing words count against you even without a percentage. What WCAG requires of captions covers the rest.
What does the DCMP Key say?
The DCMP Captioning Key names accuracy as the first element of quality captioning, and says errorless captions are the goal. It gives no percentage either. See the DCMP Captioning Key.
Why can’t a caption file show accuracy?
Accuracy is a comparison between the captions and what was said, and a caption file holds only one side of it. Deciding whether the words are right means comparing them to the audio. This engine never sees the audio and will not assert accuracy from the text alone. Nothing in a caption file can settle this. It is reported as not evaluated and is never marked passed.
So a tool that reads only the file, ours included, can find captions that are malformed, late or missing for a stretch, but cannot tell you they are right.
How does CaptionPass measure accuracy?
On the audit, Program and Contract, a person samples it. They compare clips of the captions with the institution’s own video, by hand, and the report states the sample size and what it supports.
- About 10 clips of 2 minutes each, drawn at random and weighted by running time, aiming for about 3,000 words.
- 5 kinds of error, each counted the same: substituted, omitted, inserted, speaker and missing sound.
- The result is a lower bound: what the sample supports, rounded down. A small sample supports less however clean it looks: 300 words with no errors supports only 99.0%.
That is also why a 99% target in a contract is harder to show than it sounds. About 3,000 words can show at least 99.0% only with 20 errors or fewer (observed 99.33% or better); errors bunched in a few clips lower what it shows. Captions that run close to 99% cannot be shown to meet 99% with one batch. The method, with every number it uses, is at how we measure caption accuracy.
CaptionPass for compliance shows what else we check and what we cannot.
Where does this come from?
- 47 CFR 79.1(j)(2): FCC captioning quality standards: accurate, synchronous, complete and appropriately placed. Checked .
- FCC 14-12: FCC Report and Order adopting the caption quality standards, CG Docket No. 05-231, dated 24 February 2014. Checked .
- 47 CFR 79.1(k)(2)(iv): FCC best practices for real-time (live) captioning vendors, on calculating accuracy. Checked .
- WCAG 2.1, success criterion 1.2.2: W3C, Understanding Success Criterion 1.2.2: Captions (Prerecorded). Checked .
- DCMP Captioning Key, Elements of Quality Captioning: The Key's five elements: accurate, consistent, clear, readable and equal. Checked .
This page describes the rules as published and where they stood on 16 November 2026. It is not legal advice.