ADA Title II just became enforceable — what training teams need to fix this week

What actually changed at 00:00 eastern today

Nothing visibly changed. No federal agency pushed a button. The rule itself has been public since the DOJ issued its final rule text in the Federal Register in April 2024, two years ago. What changed today is the enforceability date for Tier 1 entities — those serving populations of 50,000 or more, which effectively means every state, every state agency, most counties, most cities you have heard of, and most public universities and community colleges. DOJ can now open an investigation. Private parties can now file a complaint under Title II. Inaction after today is no longer "we are working on it by the deadline"; it is non-compliance.

For the full who's-in-scope breakdown and the text of the standard, see our reference page: ADA Title II captions — the 2026-04-24 deadline. That page stays short and authoritative. This one is the operator's playbook.

Why training video is where most of the risk lives

A public entity's web presence contains many asset types — HTML pages, PDFs, forms, images, apps. Most of them can be audited with an automated tool and fixed by a developer within a sprint. Training video is different for three reasons:

1. Volume. The average university LMS holds several thousand hours of recorded lecture, onboarding, and compliance content. A state Department of Health's training library measures in the tens of thousands of assets once you include archived certifications and in-service modules. Fixing a PDF takes an hour. Re-captioning 2,000 hours of video is a different scale of problem.
2. Accuracy threshold. WCAG 2.1 AA does not set a numeric accuracy floor in its text, but the consensus audit benchmark and the threshold cited by most legal guidance is roughly 99%. YouTube's general-purpose auto-captions, measured across academic and technical content, land between 80% and 93% depending on speaker accent, domain vocabulary, and audio quality. That gap is where complaints come from. See our WCAG video captions page for the specific success criteria that apply.
3. Vocabulary. Training content is domain-heavy by definition. A cybersecurity course says "CVE," "kubectl," "HIPAA Safe Harbor," "SAML assertion." A medical training module says "tirzepatide," "cholecystectomy," "T-SPOT.TB." A pharmacy compliance module says the brand names of seven controlled substances. The general-purpose speech model has never seen most of these, and it guesses. Those guesses are exactly the words a deaf or hard-of-hearing learner most needs to get right.

If you are at a public university, the single most likely path to a Title II complaint is a student who cannot follow a lecture because the captions render the professor's technical vocabulary as word salad. That is a low-effort complaint to file and a high-probability finding.

The 7-day sprint: day by day

Day 1 (today) — inventory and triage

Build a spreadsheet. Columns: asset ID, title, hours, LMS location, caption source, monthly views, required by compliance course?, speaker's domain. If you have a lecture-capture platform (Panopto, Kaltura, Mediasite), export its asset list. If you have an LMS (Canvas, Blackboard, Docebo, TalentLMS, Absorb), export that too. Do not try to merge them today; two parallel lists are fine.

Then sort by monthly views. You are looking for the top 50 assets. Those are where a complaint starts. Everything else is background risk.

Day 2 — the fix list

For each of your top 50 assets, answer two yes/no questions:

• Does this video have a caption track at all?
• Was the caption track produced by a human or by an auto-caption tool?

Assets with no captions go into a "fix immediately" bucket. Assets with YouTube-grade auto-captions go into a "fix this month" bucket — they are technically captioned, but unlikely to meet the accuracy threshold. Assets with human-edited captions are in the clear if the editing pass actually verified domain vocabulary (ask your team). Most vendors that offer "human-reviewed" captions do not correct technical terms unless the reviewer happens to know the field, which is rare.

Day 3 — technical-term accuracy check

Pick three assets from the "fix this month" bucket. Watch ten minutes of each with the captions on. Write down every instance where a domain-specific term is wrong. A few typical findings from audits we have seen:

• "Kubernetes" rendered as "cooper Netty's," "kupernetes," "coup burn at ease"
• "tirzepatide" rendered as "tier zip a tide," "tears a pat id"
• Employee name "Suárez" rendered as "swearers"
• "SAML assertion" rendered as "sample as urgent"
• "SCORM-packaged" rendered as "scorn packaged"

If you find more than two of these in ten minutes, you have a compliance problem on that asset, and by extension on every asset produced in the same pipeline.

Day 4 — set the go-forward policy

Everything above is remediation of the past. You also need a rule for the future — one sentence, posted in the faculty or training-producer handbook: No new training video ships without a WCAG 2.1 AA-conformant caption track, and auto-captions alone do not qualify. This is the single most important artifact you produce this week. An investigator will ask "what is your process going forward?" and a one-sentence policy is the answer.

Pair the policy with a mechanism. If you use Panopto or Kaltura, require a caption-review checkbox on the asset before publish. If you use a generic LMS, require the producer to upload an SRT or VTT file alongside the video. The mechanism is what turns the policy into compliance.

Day 5 — document the plan

Write a one-page remediation plan. Three sections: current state (inventory summary, top-50 triage results), fix sequence (which buckets, in what order, by when), go-forward policy (the sentence above, plus the mechanism). Sign it, date it, and send it to your institution's accessibility officer or general counsel. If you do not have one, send it to your director and keep the email.

When DOJ or a plaintiff's counsel writes the inevitable letter, this document is your answer. "Here is what we found, here is what we are fixing, here is when, here is how we prevent recurrence." A written plan that is actively being executed is the defensible posture. No plan is not.

Day 6–7 — batch-fix the top 10

Over the weekend or the next two working days, pick the top 10 most-viewed assets and actually fix their captions. This is the credibility move. If an investigator or a journalist writes about your institution's training accessibility, you want to be able to say "the ten videos that matter most to our learners are already compliant." That is a stronger story than "our remediation plan is comprehensive."

At 30 minutes per hour of content for a glossary-aware captioner, 10 hours is a weekend's work. At the traditional caption-correction rate (an enablement team hand-fixing auto-caption output at roughly 1–2 hours per video-hour), it is a week and a half. The difference is whether you outsource the vocabulary problem to a tool that knows your terms, or whether you do it yourself.

How to talk to your captioning vendor — a script

Whichever vendor you pick, ask the following questions in the first call. The answers separate vendors built for mainstream transcription from vendors built for training content:

1. How do you handle domain-specific vocabulary? Good answer: glossary ingest, glossary-biased decoding, per-customer term model. Bad answer: "our model is very accurate" (it isn't, on your content).
2. What accuracy do you guarantee, and what is the measurement methodology? Good answer: a number (usually 99%+), measured over word error rate on the customer's own content, with the testing script shared. Bad answer: generic marketing.
3. How do you deliver to my LMS? Good answer: direct API or webhook integration with the named platform. Bad answer: "we give you the file, you upload it." For a library of 2,000 assets, manual upload is another half-FTE of work.
4. What is the per-hour price at my monthly volume? Get this in writing. Enterprise vendors quote per minute and the math gets ugly at scale. Flat monthly plans exist and are usually 4–35× cheaper at L&D volume — we walk through the numbers on our Rev alternative page and on our 3Play alternative page.
5. What do you retain and for how long? Training content often contains employee names, student names, clinical scenarios, or internal system details. If the vendor keeps your video for model training, that is a privacy and IP problem separate from the compliance question.

If the vendor cannot answer any of the first three clearly, they are not built for your use case. That does not mean they are bad — it means they are built for something else (mainstream media, legal deposition, podcasts) and will treat your training library as a side case.

Budget politics: how to frame the ask internally

Accessibility budget conversations go badly when framed as "we need money for captions." They go well when framed as "we have a documented federal compliance exposure, and here is the cost of closing it." The difference is the word exposure. Show your director the size of the library, the auto-caption accuracy gap on a sample, and one article about a Title II settlement at a peer institution. Then show the monthly cost of a flat captioning plan.

For a typical mid-market public-sector training team producing 20–40 hours of new video a month, a flat monthly plan lands in the low-to-mid hundreds of dollars per month. A single Title II settlement lands in the five-to-six-figure range plus remediation cost, legal fees, and reputational harm. The ratio makes the budget case for you — the work is presenting the ratio clearly. Our university lecture captions page walks through the specific numbers for public-higher-ed.

Red flags — things to not do this week

• Don't hide auto-captioned video. Unpublishing training content rather than captioning it creates its own Title II exposure (denial of access). Fix, don't hide.
• Don't rely on a single accuracy number from a vendor. "99.9% accurate" on mainstream English tells you nothing about your medical or engineering vocabulary. Ask to test on your own content.
• Don't outsource the policy. A vendor can produce your caption files. Only you can publish the go-forward policy that keeps new content compliant.
• Don't confuse captions with transcripts. WCAG 2.1 AA requires synchronized captions on video; a separate transcript does not satisfy SC 1.2.2. See our captions vs transcripts explainer.
• Don't skip the written plan. "We are working on it" is not a defense. "Here is the plan dated 2026-04-24, executed as follows, signed by the accessibility officer" is.

Where GlossCap fits

We built GlossCap specifically for this week. The one thing we do differently from Rev, 3Play, and Verbit: we read your institution's glossary (Notion page, Confluence space, Google Docs folder, or a pasted list of terms) and bias the Whisper-large transcription model toward those terms on every pass. That is why "kubectl," "tirzepatide," and your professor's surname come out right on the first render. The output is WCAG 2.1 AA-conformant SRT or VTT, and the Team and Org plans deliver directly by webhook to Kaltura, Panopto, TalentLMS, Docebo, Absorb, and the other LMSes that matter to training teams. Pricing is flat monthly — $29, $99, or $299 — because for a team producing regular volume, per-minute billing punishes exactly the behaviour Title II asks for (more captioned content, not less).

If you want the regulatory and operator backstory, the case for why we built this essay goes deeper. If you want the short answer, the Team plan at $99/mo covers 30 hours of video a month — roughly the monthly output of a typical mid-market L&D team — and the trial starts from the form on our homepage.

See pricing   Read the Title II reference page

FAQ

Further reading

• ADA Title II captions — the 2026-04-24 deadline (reference page)
• WCAG 2.1 AA captions — the exact spec
• SC 1.2.2 Captions (Prerecorded) explained
• Every WCAG success criterion that applies to video
• University lecture captions — the back-catalog reality
• Compliance training video captions — captions as audit evidence
• EAA captions — the EU companion rule
• A Rev alternative for L&D training video
• Why we built GlossCap: the regulatory and operator case