AI vs. Closed Captioning: Understanding Audio Accessibility Compliance

Captions are written text meant to account for audio content such as spoken dialogue, important sounds, and other audio information. They may also be referred to by other names, including subtitles for the Deaf and Hard-of-Hearing (SDH), and they may be open or closed. While AI is a helpful tool for captioning, human review is essential for accessibility and compliance.

Captioning plays an essential role in content accessibility. Many people who are Deaf or hard-of-hearing, as well as those with some types of cognitive and auditory processing disabilities, use captions to understand audiovisual material. Captioned content has become increasingly widespread in recent years as more users realize its benefits, including those without disabilities who want to access content in noisy or quiet environments.

The rise of artificial intelligence (AI)-powered automated captioning tools—called AI captioning—has further driven up adoption by making captions much easier to generate. However, AI captioning has important limitations. If your organization is trying to comply with accessibility laws, and provide truly equitable experiences for users, this approach may not be sufficient.

Read on to learn about the different types of captioning and the legal requirements for captions. You’ll also find a quick checklist with captioning best practices.

Key insights

AI captioning transcribes dialogue but generally omits non-speech audio, so solutions that only cover spoken dialogue are no longer sufficient to satisfy WCAG.
Three WCAG success criteria cover captions and transcripts: 1.2.2 for prerecorded video at Level A, 1.2.4 for live video at Level AA, and 1.2.1 for prerecorded audio-only content, which a transcript satisfies.
Captions are required by U.S. and international laws including the Americans with Disabilities Act (ADA), Section 508 and Section 504 of the Rehabilitation Act of 1973, and the European Accessibility Act (EAA) (for content created after the enforcement deadline). WCAG-conformant captions support compliance with these laws.
Translated subtitles do not satisfy WCAG1.2.2 or 1.2.4 and are not sufficient for EAA compliance, even though European Union (EU) and U.K. sources use the word subtitles for what WCAG calls captions. That’s because translated subtitles do not include other information, such as speaker identities and important non-verbal sounds.
Publishers are responsible for the compliance of captions provided by third-party vendors. Organizations should validate the accessibility of captions from third parties.

What are the different types of video captions?

While you may be familiar with subtitles or closed captions, there are actually several different types of captioning, and each has its own benefits:

AI captions: AI captions are computer-generated, or automated, captions that capture the dialogue in multimedia content. AI captioning tools make adding captions to videos easier for those publishing content, but without human supervision, they are often inaccurate and incomplete.
Closed captions: This type of captioning is often started with software but requires human oversight to complete. Closed captions capture dialogue as well as important sounds, sound effects, or music within media, all of which have to be added by human captioners. They may also be turned on and off and can be configured to be displayed in different ways. These captions are both more accurate and more complete than AI captions, and because they include non-speech audio and are reviewed by a person. Closed captions are the form of captioning that can satisfy Web Content Accessibility Guidelines (WCAG) success criteria 1.2.2 and 1.2.4.
Real-time captions: Also called live captions, captions for live synchronized media, or Computer Assisted Real-time Translation (CART), real-time captions are human-generated captions produced at the same time audio is accessed. These are most commonly provided during live speech as opposed to within media recordings.

It’s worth noting that, in some countries outside of the U.S., the term “closed captions” is not used. For example, in the U.K., the term “subtitles for the Deaf and Hard-of-Hearing (SDH)” is used instead.

How can I tell the difference between types of captions?

To tell the difference between types of captions, review the captions to understand if they include descriptions of surrounding audio. This information is typically provided in brackets: for example, [applause], [laughter], or [music plays]. Music can also be indicated by a music note symbol or italics. If these descriptions are present, you may be accessing closed captions.

There is one key difference between closed and open captions: Closed captions can be turned on and off by the user, and there are options to control their appearance. On the other hand, open captions are “burned” into the video, keeping users from turning them on and off or controlling their display preferences.

Captions can follow a variety of visual formats, but they’re most frequently rendered in white or yellow text on a black background. Though captions sometimes appear without a background, it’s important to ensure that there’s sufficient color contrast between caption text and the content behind it for accessibility.

You can also review the captions for accuracy: A regular pattern of errors or incomplete speech generally indicates the captions are generated by AI. Contrast this with computer assisted live translation (CART), which is a professional service in which a trained operator converts speech into verbatim text as it’s spoken.

When are captions required?

As an accessibility best practice, all video or audio recordings should include closed captions. In fact, the U.S. Federal Communications Commission (FCC) has required that broadcasting companies, cable companies, and satellite services provide closed captioning for all new, non-exempt English language programming since 2006, and, under the 21st Century Communications and Video Accessibility Act (CVAA), for IP-delivered video programming that previously aired on U.S. television with captions.

Additionally, many U.S. and international laws, including the Americans with Disabilities Act (ADA), Sections 508 and 504 of the Rehabilitation Act of 1973, and the European Accessibility Act (EAA) mandate the use of closed captioning in publicly available audiovisual content. To comply with these laws, organizations must conform with the Web Content Accessibility Guidelines (WCAG), although WCAG versions and compliance dates vary:

In 2024, Title II of the Americans with Disabilities Act (ADA) was updated to include WCAG 2.1 Level AA as the standard for digital accessibility.
Recipients of funding from the U.S. Department of Health and Human Services (HHS) must meet the same standard under a parallel rule in Section 504 of the Rehabilitation Act.
Section 508 cites WCAG 2.0 Level AA as the compliance standard.
The presumptive EAA compliance standard, EN 301 549, currently incorporates WCAG 2.1 Level AA. (A new version of EN 301 549, which is expected to become official in November 2026, incorporates WCAG 2.2. Level AA.)

These laws impact a wide range of organizations across the public and private sectors. While Title II of the ADA applies to state and local governments and affects their vendors, Section 508 applies to federal government agencies, and impacts vendors who do business with these organizations. The EAA covers a wide range of products and services circulating in the European Union (EU) and applies to most companies with EU-based customers.

Major accessibility laws that require captions

Law	Who it covers	Standard	Compliance date
ADA Title II	U.S. state and local government	WCAG 2.1 Level AA	April 2027 for entities serving 50,000 or more people, April 2028 for smaller entities and special district governments
Section 508	U.S. federal agencies	WCAG 2.0 Level AA	In force
Section 504	Recipients of federal financial assistance	WCAG 2.1 Level AA (for recipients of HHS funding)	May 2027 for recipients of HHS funding with 15 or more employees, May 2028 for those with fewer
EAA	Most products and services placed on the EU market, including non-EU companies with EU customers	EN 301 549, which incorporates WCAG 2.1 Level AA	Applies since June 28, 2025
CVAA / FCC	IP-delivered video programming that previously aired on US television with captions	FCC caption rules and quality standards	In force

Which rules apply depends on who you are and where your audience is. In most cases, the operative technical requirement traces back to the same WCAG success criteria.

WCAG success criteria for captions and audio

Success criterion	Level	What it requires
1.2.1 Audio-only and Video-only (Prerecorded)	A	A text alternative, typically a transcript, for prerecorded audio-only content
1.2.2 Captions (Prerecorded)	A	Captions for all prerecorded audio in synchronized media (video with sound)
1.2.4 Captions (Live)	AA	Captions for all live audio in synchronized media

These criteria are unchanged between WCAG 2.1 and WCAG 2.2. A common misreading is that 1.2.2 covers all recorded audio. It doesn’t: Audio-only content falls under 1.2.1, which a transcript satisfies.

Do subtitles count as captions under the EAA?

Importantly, translated subtitles are not the same as captions, and do not satisfy EAA compliance requirements. The question causes trouble because European usage differs from American: What WCAG calls captions, EU and U.K. sources often call subtitles for the Deaf and Hard-of-Hearing, or SDH, and the EAA uses that term as well.

The label matters less than the content. Translated subtitles carry spoken dialogue for viewers who can hear the audio but don’t speak the language. Captions serve viewers who cannot hear it, so they must also convey non-speech audio: sound effects, music, and speaker identification.

Because the presumptive standard of conformity with the EAA, EN 301 549, incorporates WCAG, success criteria 1.2.2 and 1.2.4 directly in scope. Translated subtitles alone do not satisfy either one, however accurate the translation. If your content is dubbed or subtitled for EU markets, captions are a separate deliverable, not a byproduct of localization.

What to require from a captioning vendor

Most organizations buy captioning rather than produce it. The compliance obligation, however, is on the publisher. Before you sign, make sure to get clarity on the following:

How accuracy is measured: Word error rate on clean studio audio isn’t the same as caption accuracy including punctuation, speaker identification, and non-speech audio. Ask for a rate measured on content like yours. Neither WCAG nor FCC rules set a single numeric threshold, and widely cited industry benchmarks such as 99% are guidance rather than law.
Non-speech audio by default: Sound effects, music, and speaker changes distinguish captions from a transcript of dialogue. Confirm they’re included as standard.
Human review: Automated output is a starting point, but it’s not a final deliverable. Human review is essential for ensuring caption accuracy. Establish who is responsible for that review, and what specifically they check for.

What are best practices for captioning?

WCAG Success Criterion 1.2.2 requires captions for all prerecorded audio in synchronized media, at Level A, with captioning defined as text alternatives for both speech and non-speech audio information. Prerecorded audio-only content falls under 1.2.1 instead, which a transcript satisfies.

Including auditory information outside of speech is critical for making closed captioning conformant with WCAG. Because AI captioning only captures dialogue, it may not meet WCAG criteria without human intervention.

WCAG’s requirements mirror the caption quality standards the FCC adopted in 2014, accuracy, synchronicity, completeness, and placement. Checking the following four items will help you confirm that your captioning conforms with WCAG and complies with regulations—regardless of the method you use to generate those captions:

Accurate: Spoken words and non-speech sounds should match and should be captioned accurately.
Synchronous: Captions must be in sync with the audio of the program. Text should coincide with corresponding spoken words and sounds and be displayed on screen at a speed that users can read.
Complete: Captions must be present throughout the entirety of a program.
Properly placed: Closed captions should not block any important on-screen visuals, overlap with one another, or run off the edge of the screen.

Closed captions are just one piece of the puzzle when it comes to digital accessibility. At Level Access, we offer the expertise, training, testing, and legal expertise to support you with all facets of digital accessibility—helping you achieve WCAG conformance and meet your compliance goals. Contact us to learn more about our solution.

The post AI vs. Closed Captioning: Understanding Audio Accessibility Compliance appeared first on Level Access.