Video Is Your Cheapest Accessibility Fix. Most Sites Skip It

ClearPath Team · 2026-08-08 · 7 min read

Walk through almost any marketing site's homepage and you'll find a hero video. Autoplaying, muted or not, looping in the background, or sitting in a custom player under a "Watch Our Story" button. It's one of the most common accessibility failures auditors find, and it's also one of the cheapest problems on this list to actually fix. Yet it keeps shipping broken, over and over, on sites that otherwise look polished.

The gap isn't a mystery. Video accessibility touches production, not just code, so it falls into a weird no man's land between the marketing team, the agency, and whoever built the site. Nobody owns it, so nobody fixes it. Let's go through what actually needs to happen. Worth flagging early: a widget like ClearPath can help visitors manage things like motion and flashing on the page around a video, but it can't caption, transcribe, or rebuild the video itself, that part still comes down to the steps below.

Which WCAG rules apply to video

For prerecorded video with sound, there are several success criteria stacked on top of each other, and most sites only address one of them (if that):

• 1.2.1 Audio-only and Video-only (Prerecorded): silent video (like a background loop with no audio) needs a text alternative describing what's shown, unless it's purely decorative.
• 1.2.2 Captions (Prerecorded): captions are provided for all prerecorded audio content in synchronized media, except when the media is a media alternative for text and is clearly labeled as such. This is Level A, the baseline.
• 1.2.3 / 1.2.5 Audio Description: if the video shows visual information that isn't spoken aloud (a chart flashing on screen, someone pointing at a product, on-screen text), audio description is provided for all prerecorded video content in synchronized media. This trips people up constantly because they think captions alone cover it. They don't.
• 1.2.4 Captions (Live): for a live-streamed webinar or event, captions are provided for all live audio content in synchronized media at Level AA.
• 1.4.2 Audio Control: audio that plays automatically for more than 3 seconds must be controllable.
• 2.2.2 Pause, Stop, Hide: moving or auto-updating content must be pausable or stoppable.
• 2.3.1 Three Flashes: intro animations and video transitions can't flash more than three times per second.

There's also a Section 508 layer worth knowing if you work with any government or education clients: where a product displays video with synchronized audio, it must provide user controls for closed captions and audio descriptions, at the same menu level as the controls for volume or program selection. A caption toggle buried three menus deep in a custom player doesn't satisfy that.

Why the platform's auto captions usually don't count

YouTube, Vimeo, and most hosting platforms will generate captions automatically now. It feels like the box is checked. It isn't, and this is the single most common mistake teams make. Guidance from the W3C is direct on this point: W3C's current informal guidance states that automatic captions are not sufficient. Interestingly, when the DOJ wrote its digital accessibility rule, it deliberately avoided naming a specific accuracy threshold or method, in part because "captioning technology is rapidly evolving and any additional specifications regarding how to meet WCAG 2.1's live-audio captioning requirements could quickly become outdated." That's not a green light for auto captions. It just means the standard is judged on the outcome (accurate, complete captions) rather than the tool that produced them.

In practice, auto-generated captions miss homophones, botch names and product terms, drop punctuation in ways that change meaning, and almost never label speakers or meaningful sound effects. Captions not only include dialogue, but identify who is speaking and include non-speech information that matters to the story, like a phone ringing or a door slamming. The industry benchmark most caption vendors work toward is a 99% accuracy rate, and raw auto captions typically land well below that on anything with background music, accents, or industry jargon. Fixing them after the fact usually just means paying someone to edit a bad first draft, which often costs nearly as much as starting clean.

What a transcript adds that captions don't

Captions and transcripts solve different problems, and a lot of sites treat them as interchangeable. They aren't. Captions are synced to the video timeline for someone watching in real time. A transcript is a standalone, readable page that a screen reader user can navigate at their own pace, that search engines can index, and that someone on a slow connection or in a sound-sensitive environment can read instead of streaming video at all. A full transcript also has to account for anything shown on screen that isn't spoken, similar to what audio description covers, which is why a prerecorded video with meaningful visual action still needs a real audio description track at Level AA even when a transcript is available. Post the transcript on the page near the video, not as a buried PDF download. That alone makes it more useful to more people.

Background video and animation that won't stop

The hero loop with no controls is its own category of problem. If it has no audio, it still needs to be pausable once it's been playing more than five seconds, and it needs a text alternative if it's conveying anything beyond mood. If it has audio and autoplays, that's a Level A failure the moment it crosses three seconds, full stop. The fix is genuinely simple: mute it by default, cap the loop, and add a visible pause control. If you're unsure whether to autoplay content, don't. Giving users control over when and how content plays improves accessibility, usability, and overall user experience for everyone. Photosensitive seizure risk is the other reason this matters. Fast cuts, strobing transitions, and flashing intro graphics need to stay under the flash threshold regardless of whether there's sound. This is one spot where ClearPath's seizure-safe profile can genuinely help visitors on the page level, since it lets someone dial back flashing and motion sitewide, though it's a supplement to fixing the video itself, not a substitute.

Custom players and the keyboard trap

A lot of "accessible" video setups fall apart the moment someone tries to operate the player without a mouse. Most issues come from custom-built players that replicate native controls without the underlying semantics. The native HTML5 video element solves many issues by default, and custom players have to earn that same behavior with intentional engineering. If your marketing team picked a slick-looking custom player because the native browser controls didn't match the brand, there's a decent chance nobody checked whether play, pause, volume, seek, and the caption toggle actually work by keyboard alone, or whether a screen reader announces the button states correctly. That's a development fix, not a video production fix, and it's usually a matter of days of engineering time, not a rebuild. Something like ClearPath won't rewrite that player's code, but it can offer keyboard and screen-reader-friendly navigation aids elsewhere on the page while the real fix gets scheduled.

A realistic order of work and rough costs

1. Kill autoplay with sound first. Free. This is a settings change, not a production job.
2. Caption your highest-traffic videos properly. Human-reviewed captioning runs roughly $1 to $3 per minute through most professional services, sometimes more for rush turnaround. A five-minute homepage video lands somewhere around $10 to $15 with budget AI-plus-review services, or closer to $50 to $100 with a fully human vendor. Either way, this is not an expensive line item compared to almost anything else on an accessibility punch list.
3. Write or commission a transcript for anything with substantive spoken content, and publish it on the page.
4. Add audio description for videos where the visuals carry information the narration doesn't cover. This is the step most teams skip entirely, and it's genuinely the most involved one, but it's usually only needed for product demos and training content, not a 30-second brand reel.
5. Test the player by keyboard, tab through it, try to pause, seek, and toggle captions with no mouse. If it fails, that's a dev ticket.

Do this in order and most sites can clear their video backlog for a few hundred dollars and a couple of sprints, not a redesign.