Captioned Doesn't Always Mean Accessible

A caption file exists. The tool processed the audio. The compliance checkbox got checked. And yet, for someone who is Deaf or hard-of-hearing, the video might as well have no captions at all.

This is the central problem Christopher Phillips identifies in a new analysis from WebAIM (opens in new window): the accessibility field has conflated caption presence with caption quality, and that conflation has real consequences for real people. The piece arrives at a moment when automated captioning tools have improved dramatically — good enough, in many cases, to fool a surface-level review. That improvement is precisely what makes the argument more urgent, not less.

What WCAG 1.2.2 Actually Requires

WCAG Success Criterion 1.2.2 (opens in new window) sets the floor: synchronized captions for all prerecorded audio content. But the Understanding document goes further, specifying that captions must identify who is speaking and represent meaningful non-speech information — sound effects, tone, environmental audio that carries meaning. Phillips points out that the standard establishes what is required without prescribing how to implement it. That gap is where quality collapses.

Consider the implementation questions the WebAIM piece surfaces:

Quality Dimension	Compliance Minimum	Accessible Practice
Dialogue accuracy	Synchronized text present	Verbatim vs. cleaned-up speech — context-dependent
Speaker identification	Not explicitly required per-change	Identify speaker at every change, especially off-screen
Non-speech information	Meaningful sounds represented	Standardized formatting (e.g., [applause], [alarm])
Line length / display	No character limit specified	Industry standard: ~32 characters per line, 2 lines max
Caption timing	Synchronized to audio	Timed to reading speed, not just audio onset
Caption placement	Not specified	Avoid obscuring on-screen text or speaker faces

Each row represents a decision point where an organization can produce a technically compliant caption file that still fails the people it's supposed to serve.

The Automation Trap

For years, the shorthand was simple: automatic captions are bad, human captions are good. Phillips explicitly retires that binary. Today's automated tools can produce accurate transcripts — but accuracy of dialogue is only one dimension of quality. A tool that transcribes every word correctly while never identifying speakers, never representing meaningful sound effects, and displaying text in 60-character blocks that disappear before a reader can process them has technically done part of the job.

This mirrors a broader pattern in accessibility evaluation. Research on automated testing methodology shows that even the most sophisticated automated tools detect at most 37% of accessibility barriers when measured against comprehensive manual review. The same structural limitation applies to automated captioning: the tool can process what it can measure (audio-to-text), but it cannot evaluate the full user experience that depends on context, formatting, and human judgment.

Phillips adds a counterintuitive warning: paid captioning services aren't a guarantee either. Some vendors market professional captions where the end product is worse than free automated alternatives. The production method — human or machine — is not a proxy for quality.

What the Compliance Checkbox Misses

The deeper issue is organizational. Many institutions treat caption compliance as a binary state: either a caption file is attached to the video or it isn't. This is the same failure mode that research on settlement agreements identifies in legal contexts — organizations achieve a technical milestone and stop there, never building the internal capacity to sustain quality over time.

For higher education, this matters acutely. Universities produce enormous volumes of video content — lectures, recorded seminars, promotional materials, student services information. A lecture with inaccurate speaker identification or missing non-speech cues isn't a minor inconvenience for a Deaf student; it's a barrier to the educational content itself. Title II obligations (opens in new window) require effective communication, not merely the presence of a caption file.

Public transit systems face a parallel challenge with video-based service announcements and rider information. A captioned safety video that fails to represent warning tones or identify which speaker is giving which instruction has not achieved equal access — it has achieved the appearance of it.

Building a Caption Quality Review Workflow

Phillips offers a practical evaluation framework that organizations can implement now, without waiting for vendor standards to mature. Before publishing any video, review against these specific criteria:

• Dialogue accuracy: Is the transcribed text accurate, including corrections for speaker self-corrections where they change meaning?
• Speaker identification: Are speakers identified at every change, particularly when speakers are off-screen or visually indistinguishable?
• Non-speech representation: Are meaningful sounds represented — not just music cues, but warning tones, applause, environmental sounds that carry informational weight?
• Timing: Do captions appear and disappear at a pace a reader can follow, or are they racing ahead of or behind the audio?
• Readability: Are line lengths reasonable? Is text placement avoiding critical on-screen information?
• Player testing: Have you reviewed the caption experience in the actual player your audience will use, not just in your editing environment?

The Described and Captioned Media Program's Captioning Key (opens in new window) provides detailed, example-driven guidance on each of these dimensions. The BBC Subtitle Guidelines (opens in new window) go further into timing and presentation standards. These are operational references, not aspirational documents.

The Organizational Capacity Question

Most organizations don't have a captioning quality problem. They have a captioning review problem. The capacity to generate captions — through automated tools or vendor services — now exists almost universally. The capacity to evaluate caption quality against established standards is far less common.

Building that capacity doesn't require hiring a captioning specialist for every team. It requires:

1. Designating someone responsible for caption quality review before publication
2. Training that person against a specific quality checklist (the resources above provide the raw material)
3. Establishing a feedback mechanism so users who encounter poor captions can report them
4. Auditing a sample of existing published videos against quality standards — not just checking for caption file presence

This is the kind of operational investment that compliance frameworks often undercount: the human review capacity that sits between automated output and genuine accessibility.

The Standard That Matters

Phillips closes with the right measure: Can viewers rely on the captions to understand the message of a video? Not: does a caption file exist? Not: did an automated tool process the audio? Not: did a vendor invoice get paid?

For someone who is Deaf or hard-of-hearing, a caption file that fails on speaker identification, non-speech representation, or timing isn't a degraded experience — it's a barrier. The curb-cut effect only works if the curb cut is actually navigable. A caption track that looks complete from the outside but fails in practice is the digital equivalent of a curb cut that ends at a raised median.

The practical takeaway for accessibility practitioners: build caption quality review into your publication workflow, not as a one-time audit but as a standard gate. Review in the player your users will actually encounter. Test against the specific dimensions the WCAG Understanding document and DCMP Captioning Key identify. Stop treating caption presence as the finish line — it's the starting point.