Closed captions are a text version of a video's soundtrack - dialogue, speaker changes and meaningful sounds such as "[door slams]" or "[upbeat music]" - delivered as a separate data track that the viewer can switch on or off. The word "closed" is the important part: the text is not printed into the picture but travels alongside it, and a decoder in the television, set-top box, app or web player draws it on screen only when asked. The format was built for deaf and hard-of-hearing audiences. It now matters to advertisers for a less noble reason: a large share of video, including video ads, is watched with the sound off.

Under US Federal Communications Commission (FCC) rules, closed captioning is "the visual display of the audio portion of video programming pursuant to the technical specifications set forth in this part." A slogan typed over a product shot is therefore not, in the legal sense, a caption.

How closed captions work

Every captioning workflow has three stages: authoring, carriage and rendering.

Authoring produces a timed text file in which each cue holds a start time, an end time, the text and, in richer formats, position and styling. Recorded content is captioned by people, by automatic speech recognition (ASR), or by ASR corrected by a person. Live television relies on stenographers or "respeakers", with a delay of a few seconds.

Carriage depends on the delivery path. In analogue North American television, captions were encoded into line 21 of the vertical blanking interval, an invisible part of the picture signal. That system, standardised as CEA-608 (now CTA-608), allows roughly 32 characters per row. Digital broadcasting uses CEA-708 (CTA-708), which carries captions in the digital stream and lets viewers change font, size, colour and background. Online video does not use either directly. Streaming services and web players read sidecar files, most commonly WebVTT (Web Video Text Tracks, a W3C format) or TTML (Timed Text Markup Language) and its profiles, while many social platforms still accept the older SubRip (SRT) format.

Rendering happens on the device. Because the captions are separate data, the player decides how they look. In the US, that decision is now regulated: in July 2024 the FCC adopted rules requiring caption display settings to be "readily accessible", judged on four factors - proximity, discoverability, previewability, and consistency and persistence.

Where captions sit in the ad transaction

In programmatic video, the creative is described by a VAST (Video Ad Serving Template) document that the ad server returns to the player. Before VAST 4.1, there was no standard slot for a caption file, so captions were either burned into the video or handled through proprietary extensions. IAB Tech Lab's VAST 4.1, released for public comment in June 2018, added a standard way to deliver closed caption files. In practice, a ClosedCaptionFiles block sits in the media section of the ad response, and each ClosedCaptionFile entry declares a file type (WebVTT or TTML, for example) and a language. A compliant player can then show the ad's captions with the viewer's own settings.

The buy side supplies the file; the sell side has to render it. Players that only parse older VAST versions ignore the element, and many CTV (connected television) environments still rely on captions burned into the video.

Origin and evolution

Captioning began as an accessibility experiment. On 15 February 1972, the National Bureau of Standards demonstrated captions on an episode of "The Mod Squad" at Gallaudet College, according to the National Captioning Institute (NCI). The same year, "The French Chef" with Julia Child became the first programme broadcast with open captions - text visible to every viewer.

The FCC reserved line 21 for captions in 1976. On 16 March 1980, the first closed-captioned programmes aired: "The Wonderful World of Disney", "The ABC Sunday Night Movie" and "Masterpiece Theatre", according to NCI. Viewers needed a $250 decoder sold by Sears. Real-time captioning followed in 1982.

The decoder problem was solved by law. The Television Decoder Circuitry Act of 1990 required sets with screens of 13 inches or larger sold in the US to include caption decoders from 1 July 1993. In February 2014 the FCC adopted caption quality standards covering four elements: accuracy, synchronicity, completeness and placement.

The internet came under the rules through the Twenty-First Century Communications and Video Accessibility Act of 2010 (CVAA). Full-length programmes shown on TV with captions had to keep them online, and in 2014 the FCC extended the duty to clips: straight-lift clips from 1 January 2016, montages from 1 January 2017, and live and near-live clips from 1 July 2017, according to a Lerman Senter memo on the order.

Automation arrived in parallel. Google launched automatic captioning for YouTube on 19 November 2009, initially for a set of educational partners, according to a Google Research post by Christopher Alberti and Michiel Bacchiani. Editable automatic captions reached YouTube Shorts in 2024, when the company said creators would be able to edit the captions YouTube generated.

Why closed captions matter to marketers

The commercial argument rests on silent viewing. A 2019 study by Verizon Media and Publicis Media, an online survey of 5,616 US adults aged 18 to 54, found that 69% had watched video with the sound off in public and 25% did so even in private, according to Streaming Media. Some 80% of caption users had no hearing impairment, and captions lifted mobile ad recall by 8%. These are self-reported survey results, not observed behaviour.

Newer data points the same way. XR Extreme Reach surveyed 3,000 consumers in the UK, US, France, Germany and Spain in 2026 and found that 80% use subtitles at least sometimes, while 81% report no hearing difficulty. In the UK, 34% said subtitles made them pay more attention to ads, rising to 49% among 18 to 24 year olds. Supply is catching up: XR said on 7 April 2026 that the share of UK TV ads delivered with subtitles rose from 27% in 2024 to 54% in 2026 to date, led by beauty and personal care at 81% and with automotive lowest at 28%.

Platform data adds performance claims. Reddit's 2026 ad creative study, based on about 150,000 in-feed ads, found video assets with overlay text, closed captions included, had an 8.2% higher average conversion rate than those without. The figure groups captions with other on-screen text, so it cannot isolate the captioning effect. TikTok Shop's May 2026 creator guide notes that videos are frequently watched without sound. Google treats captions as an accessibility recommendation rather than a technical requirement in its Demand Gen ad specifications.

Accessibility law reaches ads less than many assume. US rules exclude advertisements of five minutes or less from the definition of video programming, so most TV spots carry no federal captioning mandate. Channels earning under $3 million a year are also exempt. The European Accessibility Act, in force since 28 June 2025, obliges services that provide access to audiovisual media - apps, players, programme guides - to transmit subtitles accurately and in sync, but, according to law firm Taylor Wessing, it does not create a direct duty to caption the content itself.

Limitations and disputes

Accuracy is the persistent complaint. Automatic captions can garble brand names, technical terms and accents. Caption vendor 3Play Media has argued that auto-captions "realistically average 60-70% accuracy", a figure from a 2017 post by a company that sells human-verified captioning, so it carries a commercial interest. No independent, current benchmark for platform captions is published.

Control and consent cause friction as captions merge with translation. YouTube users complained in July 2025 that titles and descriptions were auto-translated without an obvious way to see the original, and similar objections followed the default-on rollout of AI dubbing.

Format fragmentation remains. One campaign may need 608/708 data for broadcast, WebVTT for web players, a VAST caption file for programmatic and burned-in text for social placements. Burned-in captions solve delivery but cannot be resized or turned off.

Regulatory timing has produced conflicting reports. TV Technology reported a compliance date of 1 August 2026 for the FCC's display-settings rules; the FCC's own Public Notice DA 25-48, released 15 January 2025, sets 17 August 2026.

  • Open captions are burned into the picture and cannot be turned off. Most "captions" on social video ads are open captions.
  • Subtitles traditionally translate dialogue for viewers who can hear but do not understand the language, and omit sound cues. In the UK and much of Europe the word also covers same-language text for deaf viewers (often labelled SDH, subtitles for the deaf and hard of hearing), which is why survey data on "subtitle" use often includes caption use.
  • Transcripts are full text documents of the audio, not synchronised with the picture.
  • Audio description is the reverse service: narration of on-screen visual information for blind and partially sighted viewers. It is the video counterpart of alt text for images.

Recent developments

The FCC's readily accessible caption-settings requirements took effect on 17 August 2026 for TVs, set-top boxes, computers, phones and tablets, with exemptions for consumer-installed apps and devices made before the deadline.

Streaming services have widened language coverage. Disney+ began a rollout on 15 July 2026 under which subtitles and closed captions are offered in up to 42 languages. Captioning is also converging with AI dubbing: YouTube said in May 2026 that auto dubbing covers 27 languages.

On 12 November 2026, YouTube Studio began turning 16:9 podcast segments into vertical Shorts with captions added automatically from the same transcript used to select the clips. Those captions are restyleable overlays, a long way from the line 21 standard of 1980.

The legal backdrop keeps moving. On 20 April 2026 the US Department of Justice pushed ADA Title II web accessibility deadlines for large state and local governments to 26 April 2027, still under WCAG 2.1 Level AA, which requires captions for prerecorded video.

Timeline

  • 15 February 1972 - National Bureau of Standards demonstrates captions at Gallaudet College; "The French Chef" airs with open captions the same year.
  • 1976 - FCC reserves line 21 of the TV signal for captions.
  • 1979 - National Captioning Institute established.
  • 16 March 1980 - First closed-captioned programmes air on ABC, NBC and PBS.
  • 1982 - First real-time captioning, at the Academy Awards; "World News Tonight" becomes the first regularly scheduled live-captioned programme.
  • 1990 - Television Decoder Circuitry Act passed.
  • 1 July 1993 - Caption decoders required in US TV sets of 13 inches or more.
  • 19 November 2009 - YouTube launches automatic captions.
  • 2010 - CVAA extends captioning duties to internet-delivered TV programming.
  • February 2014 - FCC adopts caption quality standards (accuracy, synchronicity, completeness, placement).
  • 1 January 2016 - FCC captioning duty applies to straight-lift online clips.
  • June 2018 - IAB Tech Lab releases VAST 4.1 for public comment, standardising caption file delivery in ads.
  • 2019 - Verizon Media and Publicis Media study reports 69% sound-off viewing in public.
  • 18 July 2024 - FCC adopts "readily accessible" caption display settings rules.
  • 15 January 2025 - FCC Public Notice DA 25-48 sets the compliance date.
  • 28 June 2025 - European Accessibility Act applies.
  • 7 April 2026 - XR reports UK ad subtitling at 54%, up from 27% in 2024.
  • 20 April 2026 - DOJ delays ADA Title II web accessibility deadlines by one year.
  • 15 July 2026 - Disney+ begins expanding subtitle and caption languages to 42.
  • 17 August 2026 - FCC caption display settings compliance deadline.
  • 12 November 2026 - YouTube Studio adds auto-captioned vertical Shorts from podcast episodes.

Summary

Who: Broadcasters, streaming services, platforms and advertisers create and carry closed captions; the FCC, the European Union and national regulators set the rules; IAB Tech Lab defines how captions travel with programmatic video ads; deaf and hard-of-hearing viewers, plus a much larger group watching without sound, use them.

What: A timed text track of a video's dialogue and sounds, carried separately from the picture so viewers can switch it on or off, encoded as CTA-608/708 for TV, WebVTT or TTML online, and as a caption file in VAST 4.1 and later for ads.

When: First demonstrated in 1972, first broadcast on 16 March 1980, mandated in US TV sets from 1993, extended online from 2010, standardised in video ad serving in 2018, with new US display-settings rules in force from 17 August 2026.

Where: Broadcast and cable TV, streaming and CTV apps, web players and social platforms, with legal obligations strongest in the US and the EU and with US rules excluding ads of five minutes or less.

Why: Captions began as an accessibility service and remain one, but survey and platform data showing widespread sound-off viewing have made them a standard part of video ad creative, while accuracy, format fragmentation and viewer control remain unresolved.