Accessibility has been a talking point since a law on it became applicable in June 2025. The talk is about obligations, deadlines and fines, and that is precisely why so little happens: the obligation does not reach most small providers at all. Captions and transcripts are still worth doing, just for a different reason. Together with alt text, they are the only place where the content of your videos and images exists as text. And text is all a search engine or an answer engine can ever use from you.
Does the obligation reach you at all?
Probably not. The German Barrierefreiheitsstärkungsgesetz lists exhaustively in section 1(3) which services are covered when provided to consumers after 28 June 2025: telecommunications services, certain elements of passenger transport services, consumer banking services, e-books and services in electronic commerce. A website where you describe what you do and offer a contact form is not electronic commerce. A shop that closes a sale with a consumer is.
Even then there is a second door. Section 3(3) sentence 1 reads, in the original: "Absatz 1 gilt nicht für Kleinstunternehmen, die Dienstleistungen anbieten oder erbringen." A microenterprise, under section 2 no. 17, is a business "that employs fewer than ten persons and either has an annual turnover of no more than 2 million euros or an annual balance sheet total of no more than 2 million euros". Nine people and an online shop puts you outside. Twelve people and the same shop puts you inside.
Austria implemented the same EU directive. The Barrierefreiheitsgesetz was published on 19 July 2023 as BGBl. I no. 76/2023 and has been applicable since 28 June 2025. The list in section 2 BaFG is equally exhaustive and the microenterprise exemption is identical in wording. The difference is the penalty range: up to 80,000 euros under section 36 BaFG in Austria, up to 100,000 euros in the gravest cases and up to 10,000 euros in the rest under section 37(2) BFSG in Germany.
One detail matters for video and is rarely mentioned. Section 1(4) no. 1 expressly excludes "aufgezeichnete zeitbasierte Medien, die vor dem 28. Juni 2025 veröffentlicht wurden", recorded time-based media published before 28 June 2025. Your archive stays as it is. Everything you have uploaded since counts. Anyone who is covered therefore does not build the obligation backwards, but from the next video onwards.
One more thing that puts the debate in order: neither the act nor its implementing regulation mentions the WCAG even once. We searched both full texts on 19 August 2026 and found zero hits. Instead of success criteria, section 21 of the implementing regulation sets out functional performance criteria, and the one covering sound reads, in the original: where a service offers "auditive Bedienungsformen", there must be "mindestens eine Bedienungsform ..., die kein Hörvermögen erfordert", at least one mode of operation that does not require hearing. The word captions appears nowhere in it. Nothing else is meant.
So why is the state of things still this bad?
Because an obligation that reaches almost nobody moves almost nobody. The annual WebAIM Million analysis tests one million home pages automatically. In February 2026, 95.9 per cent of them had detectable WCAG failures, up from 94.8 per cent the year before. The average was 56.1 errors per page, 10.1 per cent more than in 2025. The share had fallen for six consecutive years; 2026 is the first year it rose again.
More telling than the headline number is which failures they are. Three of the six most common ones are simply missing text: images without alt text on 53.1 per cent of pages, links without a label on 46.3 per cent, buttons without a label on 30.6 per cent. WebAIM also notes that these have been the same six failures for seven years. That is not a technical problem, it is an editorial one.
One figure from the same analysis fits the topic exactly: 6.2 per cent of home pages carried a YouTube video, and those pages had 9.4 more errors than average. An embedded video therefore makes a page measurably worse, not better, as long as nobody writes the text that goes with it.
What does that text have to do with your visibility?
Precision pays here, because a great deal gets claimed in this field. For images, Google says it itself. Its documentation on Google Images states: "Google uses alt text along with computer vision algorithms and the contents of the page to understand the subject matter of the image." And further: "Google extracts information about the subject matter of the image from the content of the page, including captions and image titles." Alt text is therefore not an add-on for screen readers, it is one of three routes by which Google learns what your image shows at all.
For video the picture is different, and that belongs in the story. We read Google's official video SEO page on 19 August 2026: captions, caption files and transcripts do not appear on it at all. Anyone telling you a transcript is a confirmed ranking factor has no source for it. What is true is something more sober and, in practice, more important: without a transcript, the content of your video appears nowhere in the text of the page. What is not in the text cannot be found, cannot be quoted, and cannot be reproduced by an answer engine. A twelve-minute interview with three strong statements is an empty frame to any machine as long as those statements are only spoken.
That is why a transcript is not an attachment but a text page you have already paid for. You wrote the brief, ran the shoot, commissioned the edit. The step that is missing is the cheapest of them all.
Who is actually reading the captions?
Nearly everyone. For its Global Accessibility Report published on 21 May 2026, XR Extreme Reach surveyed more than 3,000 consumers in France, Germany, Spain, the United Kingdom and the United States. 79 per cent of UK viewers use captions at least sometimes. Among 18 to 24 year olds, 59 per cent say they use them always or often. Across all five markets, 81 per cent switch captions on at least occasionally despite reporting no hearing difficulty. The study comes from a vendor in the advertising technology field, so it is not a neutral institute. The direction matches what you can see on any station platform.
That shifts the question. Captions are no longer the assistive device for a small group they appear to be in legal texts. They are the default setting for people watching on a bus, watching while doing something else, learning a second language, or missing a name. Leaving them out does not save you accessibility work, it costs you the majority of your audience.
How do you write this text without turning it into a project?
Almost all of the effort sits in the first attempt; after that it is one step in the routine. Three things are enough to start.
First, the caption file. Automatic recognition now delivers a usable rough version, and that is exactly where the automation stops. Proper nouns, product names, technical terms and figures are what it reliably gets wrong, and those are precisely the words somebody searches for. Budget roughly one minute of correction per minute of video. A caption file that misspells your company name is worse than none, because it fixes the error permanently.
Second, the transcript on the page. Not as a download, not in a separate panel, but as text below the video, with subheadings and paragraphs. If the full length feels excessive, use the edited version: tightened, without filler, statements in the order of the video. That version reads better than the verbatim record and is worth just as much to the page.
Third, the alt text. The rule is short: describe what somebody who cannot see the image would need to know, in the context of this page. A chart needs its figures in the alt text, a portrait needs the name and the role, a purely decorative image needs an empty alt attribute so screen readers skip it. And more important than any rule: alt text stuffed with search terms does damage, and Google explicitly calls it spam.
The yardstick behind all of this sits in WCAG 2.2. Captions for prerecorded media are success criterion 1.2.2 at the lowest level A, text alternatives for non-text content are 1.1.1, also level A, and the minimum contrast ratio of 4.5:1 is 1.4.3 at level AA. The three items above are not the finishing touch, they are the floor. What accessibility covers as a whole is in the glossary.
Three levers for the next two weeks
1. Start with the next video, not with the archive. Correct the caption file, put the transcript below the video, from now on for every new recording. Go back through the archive only where an older piece still gets traffic.
2. Take your ten most important pages and write the alt text. Not all of them, only the ten that make money. Charts and product images first, decorative images set to empty alt. One hour of work, and you are ahead of the 53.1 per cent of home pages with missing alt text.
3. Check once whether the obligation reaches you. Four questions, ten minutes: do you sell online to consumers, is your service in section 1(3) BFSG or section 2 BaFG, do you employ ten people or more, do you publish after 28 June 2025. If the answer at the end is yes, get legal advice rather than a checklist off the internet.
If you want to know how much text is currently missing from your videos and images, we will look at it with you. 📄