Blog

A Caption Is Inherited, Not Created

November 20, 2026  ·  Tablingo

The US internet caption rule only applies to programming that already aired on television with captions. Follow the chain back and it ends before it reaches most of what you watch.

Open a foreign news channel in a browser tab and look for the CC button. On plenty of streams it is not there, and the usual reading is that the platform is saving money on transcription.

That reading is generous. In the United States, four separate documents decide who has to caption what, and each one was written for somebody else: a broadcaster, a distributor, a government office. None of them was written for the video in that tab. The one rule that does reach the internet does not create a duty of its own either — it inherits one, and only if a television broadcast already happened.

Four rules, four circles

Here is the map, in the order the obligations actually flow.

RuleCoversReaches you?
§79.1TelevisionIf it aired
§79.4Internet videoIf it aired captioned
WCAG 1.2.4Web, liveNever a call
ADA Title IILocal governmentIf it is public

The first two are the interesting pair, because the second one is not independent. It borrows.

The television rule exempts the language you want

§79.1 is the caption rule people mean when they say captions are required by law. It puts the duty on television distributors — stations and cable operators — and asks for 100% of new non-exempt programming, per channel, per quarter.

Then it lists what is exempt, and the third item is the one that matters here: “All programming for which the audio is in a language other than English or Spanish, except that scripted programming that can be captioned using the ‘electronic news room’ technique is not exempt.”

Read the carve-back carefully, because it is narrower than it sounds. Everything whose audio is not English or Spanish is out — a Japanese drama, a Korean talk show, a French press conference, a Portuguese match feed. The one thing pulled back in is scripted programming that can be captioned from a prepared news script, which is a way of generating captions from a teleprompter rather than from a person listening to the audio.

So the rule does not require foreign-language captioning and then carve out exceptions. It exempts the category and claws back a technique. If the audio is in the language you actually need help with, the American television caption rule has nothing to say about it.

The internet rule borrows from television

Most readings stop there. The structure gets more interesting one section later. §79.4 covers internet delivery, and it does reach streaming. But look at the condition on the obligation: “All nonexempt full-length video programming delivered using Internet protocol must be provided with closed captions if the programming is published or exhibited on television in the United States with captions on or after the following dates.”

That “if” is doing all the work. The duty does not attach to a video because it is online and someone might need captions. It attaches because the same program already went out on American television, with captions, after a date the rule sets. Internet captioning is inherited from television captioning, not created beside it.

Follow the inheritance back one step and it runs into the exemption above. A Japanese drama that never aired on American television never satisfies the condition. A Japanese drama that did air also fails it — it was exempt under §79.1(d)(3), so it aired without captions, and the condition is “with captions”. The first link of the chain is where the chain ends.

Two more limits sit on top. The definition of video programming ends with a carve-out, “but not including consumer-generated media”, which is most of what a browser tab shows on any given evening. And clips are handled separately from full-length programs.

None of this is a loophole somebody found. It is what a rule written to protect a broadcast audience looks like when it is stretched over a network that did not exist when the obligation was designed.

The web standard covers the stream, not the call

The W3C rule is the one most often quoted at platforms, and it is one sentence. Success Criterion 1.2.4, Level AA: “Captions are provided for all live audio content in synchronized media.”

The Understanding document — the working group’s own explanation — then draws a boundary that almost nobody quotes: “This success criterion was intended to apply to broadcast of synchronized media and is not intended to require that two-way multimedia calls between two or more individuals through web apps must be captioned regardless of the needs of users. Responsibility for providing captions would fall to the content providers (the callers) or the ‘host’ caller, and not the application.”

So the standard cited in every accessibility audit does not require a meeting product to caption your call. It requires the host of the call to arrange it. That is a defensible division — the application cannot know who needs what — but it means the CC button in a meeting is a product feature, not a compliance artefact. When it is missing, no audit finding follows.

The one rule with a date is the smallest circle

The ADA rule is the only one of the four with a deadline, and the deadline has just moved. In April 2024 the Department of Justice made WCAG 2.1 AA the technical standard for state and local government web content and apps. On April 20, 2026, an interim final rule pushed both dates back by a year: “The compliance date for State and local government entities with a total population of 50,000 or more is extended from April 24, 2026, to April 26, 2027. The compliance date for public entities with a total population of less than 50,000, or any special district government, is extended from April 26, 2027, to April 26, 2028.”

PopulationWasNow
50,000 or more2026-04-242027-04-26
Under 50,000, districts2027-04-262028-04-26

The delay matters for one reason: a checklist written before April 2026 will quote the old dates, because they were correct at the time, and a city that planned around April 2026 has been handed twelve months nobody went back and announced. If you are reading a calendar, check which version it was written under.

Then notice the scope. This is the one circle with both a date and an enforcement path behind it, and it covers a council webcast, a school board meeting, a public hearing. Not a channel you subscribed to.

So what is left

Put the four together and the tab falls in the gap.

LayerWho owes itYou can
StatutoryBroadcasters, citiesWait for a date
VoluntaryThe platformFind the switch
UnlegislatedYouAdd a layer

The honest conclusion is not that nobody cares. It is that the obligation was designed as a floor under broadcast, and the floor was never extended to where people actually watch. When a stream has no captions, that is usually not a company dodging a duty. It is a duty that was never written for that stream.

Three checks, one minute

First, find a council or school board meeting streamed live and look for captions. If they are there, you are looking at the one circle with a deadline attached — and its deadline is now 2027 or 2028, depending on how many people the body serves.

Second, open a meeting product and look for its caption switch. Its absence is not a compliance failure, because 1.2.4 does not cover calls. It is a product decision, and product decisions can be reversed by asking.

Third, find a foreign programme with official subtitles. That track is voluntary — either the distributor chose to make it, or the rights holder shipped one. §79.1 never asked for it, which is why the ones that exist are inconsistent: a hit series gets subtitles and the rest of the catalogue does not.

None of the three requires reading a regulation. Each tells you which layer you are standing in, and the layer decides whether waiting makes sense.

The layer nobody legislated

What is left is the layer that cannot be inherited, because there is nothing upstream to inherit from. The audio is in front of you, and no broadcaster has prepared a script for it, no television airing has triggered an obligation, and no public body is on the hook for it. That layer can only be built.

Since there is no standard to appeal to, judge it on one thing: when a speaker finishes a phrase, is the caption already on screen, or does it arrive a beat later? Run that test on any tool, ours included. If the one you try is slow, the honest answer is that this layer has no rule to appeal to either — which is exactly why the test is the only thing worth trusting.

If you want to try ours, it’s called Tablingo.