How to evidence ADA Title II captions
The DOJ’s Title II web rule names WCAG 2.1 Level AA as the technical standard for state and local government web content. Most teams reading it discover that the hard part is not captioning the videos. It is being able to show, per asset and a year later, what each one was checked against and what the result was.
The dates, and who they bind
Takes effect 26 April 2027 — 6 months away
ADA Title II web rule — large public entities
US state and local government entities serving a population of 50,000 or more, including public universities and school districts.
Standard: WCAG 2.1 Level AA. 28 CFR Part 35, DOJ final rule, as extended by the April 2026 interim final rule
Takes effect 26 April 2028 — 18 months away
ADA Title II web rule — small entities
US state and local government entities serving fewer than 50,000 people, and special district governments.
Standard: WCAG 2.1 Level AA. 28 CFR Part 35, DOJ final rule, as extended by the April 2026 interim final rule
The timing above is computed from the effective date each time this page renders, so it cannot go on describing a deadline as upcoming after it has passed.
Which criteria the caption question turns on
Title II points at WCAG rather than at a caption-specific regulation, so “are our captions compliant” resolves to a small number of success criteria — and to how much of each one a caption file can actually settle.
| Criterion | What it requires | What a caption file can settle |
|---|---|---|
| 1.2.1Audio-only and Video-only (Prerecorded) | Prerecorded audio-only media has a transcript, and prerecorded video-only media has a transcript or an audio track that presents the same information. | Not evaluatedNothing in a caption file can settle this. It is reported as not evaluated and is never marked passed.Not evaluated. A caption file cannot show whether the media is audio-only or video-only, or whether a transcript or audio alternative exists for it. |
| 1.2.2Captions (Prerecorded) | Captions are provided for all prerecorded audio content in synchronized media. | Partly checkedThe file can show this failing, but cannot show it passing. No findings means nothing detectable is wrong — not that the criterion is met.A caption file proves captions exist, and can show they start late or skip stretches of the programme. It cannot prove they cover all audio, because that comparison needs the audio. |
| 1.2.3Audio Description or Media Alternative (Prerecorded) | Prerecorded video in synchronized media has audio description or a full text alternative. | Not evaluatedNothing in a caption file can settle this. It is reported as not evaluated and is never marked passed.Not evaluated. Audio description and text alternatives are separate from the caption file, and whether the video needs them depends on what it shows. |
| 1.2.4Captions (Live) | Captions are provided for all live audio content in synchronized media. | Not evaluatedNothing in a caption file can settle this. It is reported as not evaluated and is never marked passed.Live captioning is a property of the broadcast, not of a delivered file. Nothing in a sidecar can settle it. |
| 1.2.5Audio Description (Prerecorded) | Prerecorded video in synchronized media has audio description. | Not evaluatedNothing in a caption file can settle this. It is reported as not evaluated and is never marked passed.Not evaluated. Audio description is a separate track, and whether visual content needs describing depends on the video, which this engine never sees. |
Cited from W3C Web Content Accessibility Guidelines 2.1, Level AA (W3C Recommendation, 5 June 2018) — read the standard.
Assurance vs evidence
An assurance is a sentence: our videos are captioned. Evidence is a record that survives the person who wrote it. The difference shows up the moment somebody outside your team asks the question — a procurement reviewer, a grant condition, an auditor, or a complainant — because none of them can act on the sentence.
A per-asset record that answers them names, at minimum:
- the asset, and the exact caption file that was checked;
- the standard and its version, not just “WCAG”;
- the result per criterion, including the criteria that were not evaluated and why;
- when the check ran, and by what engine;
- something that lets a third party confirm the record has not been edited since.
That last one is what makes it evidence rather than a spreadsheet. A CaptionPass receipt is signed, and anyone can verify one without trusting whoever sent it — the check runs on our server, against our signing key, and needs no account.
Rolling it up for a whole library
One receipt answers “was this delivery conformant”. Procurement usually asks the larger question — across everything you publish, which criteria do you meet? — and that is an Accessibility Conformance Report. The caption rows of a VPAT are where the same evidence lands, and a sample report over a small library shows what that looks like with its failures included.
What this is not. A conformance receipt is evidence of what was checked, by which engine, against which version of which standard, and what it found. It is not a legal opinion, an audit, or a safe harbour. The engine that issues it reads the caption file and never the audio, so anything that needs the audio — above all whether the captions match what was said — is reported as not evaluated rather than assumed.
Common questions
- What does ADA Title II require for captions?
- The DOJ web rule names WCAG 2.1 Level AA as the technical standard for state and local government web content and mobile apps. For captions that means the WCAG success criteria on synchronized media rather than a caption-specific regulation. Takes effect 26 April 2027 — 6 months away for entities serving 50,000 or more people; takes effect 26 april 2028 — 18 months away for smaller ones. Source: 28 CFR Part 35.
- Is having captions enough to evidence compliance?
- Having captions is the requirement; evidencing it is a separate problem. An auditor, a procurement officer or a complainant asks what a given video was checked against and what the result was. A caption file on its own does not answer that, and neither does a vendor invoice. What answers it is a per-asset record naming the standard, its version, the date, and the result per criterion.
- Can a tool prove captions are accurate?
- Not from the caption file. Deciding whether captions match the spoken words means comparing them to the audio. The engine that grades them reads the caption file and never the audio, so accuracy is reported as not evaluated on every receipt it issues — permanently, by design. Any vendor claiming to certify caption accuracy from a sidecar file is claiming something they cannot know.
- Is a conformance receipt a legal guarantee?
- No. A receipt is evidence of what was checked, by which engine, against which version of which standard, and what it found. It is not a legal opinion, an audit, or a safe harbour. It is the artifact you hand to the person asking, so the conversation starts from a record rather than an assurance.