Captions Were Never Just for Deaf Users

Closed captions exist because of a fight for access. Deaf and hard-of-hearing viewers and advocates spent decades pushing broadcasters and eventually streaming platforms to make video legible without sound, and the technology that resulted — timed text tracks, standards for how captions should be formatted and paced — was built specifically to solve that problem. That history matters and it hasn't expired. For a genuinely deaf or hard-of-hearing viewer, captions aren't a convenience or a stylistic choice; they're the only way the content is available at all, and any conversation about how broadly useful captions turned out to be for other people should start by being honest about that.

What's striking is what happened after that fight was won. Once captions became a standard, switchable feature rather than a rare accommodation, platform after platform found that the people actually turning them on skewed overwhelmingly toward viewers with no hearing impairment at all. The audience the feature was built for kept using it, as expected. But they became a minority of total caption usage, dwarfed by hearing viewers who simply preferred, or needed in the moment, to follow video without sound.

The explanation isn't complicated once you look at where video actually gets watched. An open-plan office where playing audio out loud is a minor act of aggression toward everyone nearby. A commute on public transit, phone in hand, one earbud in or headphones left at home. A shared bedroom where a partner is asleep and the phone stays on silent as a matter of course. A gym floor lined with televisions that have never had working audio in the entire history of the building. None of these are edge cases — they describe an enormous share of where video consumption actually happens, and in every one of them, captions aren't a nice-to-have, they're the only way the content gets consumed at all.

Picture a product team building a video-heavy feed for a social app, arguing over whether captions should default on or stay tucked in a settings toggle most people never find. If they had only the original accessibility case to draw on, defaulting them off might feel like a reasonable trade-off, since the addressable audience for the accommodation is a minority of total users. But once usage data on any comparable platform gets pulled up, the picture flips entirely: a large share of total watch time, across the whole user base, happens with sound off and captions on by choice. Defaulting captions off doesn't just underserve a minority — it makes the product harder to use for a huge portion of everyone.

This is a specific instance of a pattern that shows up again and again in accessibility work: a feature designed to solve a narrow, well-defined need for a specific population turns out to be quietly load-bearing for a much larger group once it's actually available by default. Curb cuts are the canonical physical-world example — built for wheelchair users, and now depended on by parents pushing strollers, delivery workers wheeling hand trucks, travelers dragging luggage, and anyone momentarily distracted while crossing a street. Captions are the digital-world equivalent, at a scale that's easier to measure precisely because every major platform logs exactly who turns them on and when.

A reasonable worry here is that emphasizing the broad, mainstream usage case dilutes the original accessibility argument — that it lets organizations fund captions as a 'growth feature' while quietly losing sight of the population the requirement exists for in the first place. That's a real risk, but it cuts the other way more often than it seems. A feature with a large, visible, mainstream constituency gets sustained investment, better tooling, and product attention in a way that a feature framed purely as a compliance obligation for a small population usually doesn't. The disability-first case and the mainstream-usage case aren't competing for the same budget — the second one tends to protect the first by making the feature too valuable to deprioritize.

What this looks like in practice starts with treating captions as a default, not a preference buried three menus deep. Any video product deciding where captions live in its settings hierarchy is implicitly deciding how many people will ever discover them, and burying a feature that a large share of the audience actually wants isn't a neutral choice — it's a tax paid disproportionately by whichever users are least likely to go looking for a toggle, which very often includes the exact population the feature exists for. Defaulting captions on, or at minimum surfacing the choice clearly the first time someone watches a video, treats the feature as core rather than incidental.

Quality matters just as much as visibility, and this is where the two audiences genuinely diverge. A hearing viewer with captions on has the audio as a fallback — if an auto-generated caption garbles a word or mistimes a line, they can usually still follow along by ear. A deaf or hard-of-hearing viewer has no such fallback; a captioning error isn't a minor annoyance, it's the entire piece of information they were relying on, gone or wrong. That means the acceptable error rate for captions has to be set by the accessibility case, not by the more forgiving mainstream case, even though the mainstream case is now the larger share of usage. Optimizing caption quality for the audience that can tolerate errors, because they're the majority, gets the priority exactly backwards.

There's a related trap in treating captions as an afterthought bolted onto a finished video, the same way accessibility fixes get bolted onto a finished interface elsewhere. Auto-generated captions added after the fact, with no human review of names, technical terms, or timing around fast cuts, produce a visibly lower-quality experience than captions considered during production — timed with the edit, checked for accuracy, styled so they don't obscure the parts of the frame that matter. A team that treats captioning as a production step, not a post-processing afterthought, ends up with a feature that works for both audiences instead of one that technically exists but frustrates everyone who actually relies on reading it.

None of this requires abandoning the language of accommodation in favor of pure growth-metric thinking. It requires holding both facts at once: captions remain essential, non-negotiable access infrastructure for deaf and hard-of-hearing viewers, and they are also, independently, one of the most widely used features in modern video by an audience that has nothing to do with hearing ability. Products that only understand the first fact under-resource captions relative to how much they're actually used. Products that only understand the second lose sight of the accuracy bar that the original audience actually needs, chasing convenience instead of legibility.

The lesson generalizes past captions specifically. Every accessibility feature worth building is worth asking, honestly, who else it might end up serving once it's just there, available, and good — not as a way to justify the investment on other grounds, but because knowing the answer changes how the feature should be built, where it should live in the product, and what quality bar it needs to clear. Captions happen to be the clearest, best-measured example of this pattern in the entire discipline. Treating them as a niche accommodation, rather than core infrastructure that happens to have started as one, misunderstands both halves of what they've become.