When an ecommerce team prepares product video for EU-facing pages, the first accessibility request is often simple: “Add captions before launch.” Captions matter, but they solve only part of the problem. A caption file cannot explain a fitting action that appears only on screen, expose a player control to the keyboard, or stop essential product information from being communicated by colour alone.
The procurement question is therefore larger than whether a video has subtitles. The useful deliverable is a release package that carries essential meaning through sound, picture, text and the playback interface. Brands can scope that package without treating every six-second mood clip like a three-minute installation guide.
Start with the legal scope, not a universal checklist
The European Accessibility Act includes ecommerce among the services it covers. Directive (EU) 2019/882 applies to covered services provided to consumers after 28 June 2025, while EU Member States implement and enforce the directive through national law. That means a brand needs to confirm the selling entity, market, service model and national implementation before turning a production checklist into a legal conclusion.
The directive also contains qualifications that matter. Its definition of a microenterprise is an enterprise with fewer than ten employees and annual turnover of no more than EUR 2 million or an annual balance-sheet total of no more than EUR 2 million; microenterprises providing services are exempt from the directive’s service accessibility obligations. Separate mechanisms address fundamental alteration and disproportionate burden. Those provisions should be assessed for the real legal entity and activity—not assumed from the size of a creative team or storefront.
One particularly useful detail is also easy to overstate. The directive gives adding subtitles to instructional video as a non-binding example of a possible solution. It does not say that subtitles automatically make every ecommerce video compliant, or that every marketing asset requires the same described-audio treatment.
This article therefore uses WCAG and W3C guidance as a production reference. It is not legal advice and does not replace local accessibility review.
Captions, descriptive transcripts and audio description do different jobs
Captions convert dialogue, narration and meaningful sounds into synchronized text. They answer a hearing-channel question: can someone understand the audio when they cannot hear it?
Product video often carries important information through the picture instead. A status light flashes red. A filter is inserted in one orientation. A latch springs back. An interface requires a three-second press. If none of this is spoken, the information does not appear in ordinary captions.
The W3C planning guide for audio and video distinguishes between captions, a descriptive transcript and audio description. A descriptive transcript adds essential visual information to the spoken transcript. Audio description places concise narration around important actions, scene changes, on-screen text or visual states. At WCAG Level A, prerecorded video with audio requires captions and either audio description or a media alternative for important visual content; Level AA includes audio description for prerecorded synchronized media when the visual information is not already in the audio.
This distinction has a direct production consequence. If the final edit has no space for a useful description, the team may need to reserve pauses in the script, create a slightly extended described edit, or release a separate described version. W3C’s production guidance recommends planning that time before the master is locked.
Automatic captions are useful, but they are not a final approval gate. YouTube’s own guidance warns that accents, dialects, background noise and overlapping speakers can produce errors, and asks creators to review and correct the result. For ecommerce, a mistaken model name, capacity, warning or compatibility term can change the product claim itself.
The ONCE Hear–See–Operate framework
Instead of beginning with “Do we need an accessible version?”, run three controlled checks on the exact release.
Hear: remove the picture
Play the soundtrack with the screen out of view. Can a listener identify the product, the main claim, the limitation, a warning and the next action? If important information lives in sound, specify the caption languages, speaker identification, non-speech cues, timing format and terminology approver.
The approved caption text should be reconciled with the final performance and edit, not copied from a pre-shoot script. Keep product names, measurements and warnings in a controlled glossary so that machine-assisted transcription remains reviewable.
See: remove the sound
Watch the cut on mute and list what exists only in the picture: setup steps, connector locations, status changes, gestures, diagram relationships and on-screen copy. If losing one of those details would change a purchase, installation or safety decision, choose a deliberate alternative: integrate it into the main narration, add it to a descriptive transcript, create audio description, or build a separate described version.
The objective is not to narrate every decorative shot. Atmospheric lighting and actions already explained in dialogue may not need repetition. Describe the visual information whose absence would change the task.
Operate: test the release container
A caption file, transcript and described track only work if the publishing environment exposes them. Test the actual page and player: keyboard access, labelled controls, caption selection, language matching, transcript placement, mobile overlays, autoplay audio and the relationship between the video and the purchase action.
This is shared ownership. The production partner supplies approved media assets; the web or platform team implements them; the brand names one release owner who verifies that the pieces work together.
Scope accessibility by the job the video performs
A three-tier classification helps prevent both under-scoping and blanket overspending.
- Decorative video supports atmosphere while the page provides all material information. Focus on motion controls, flashing risk and autoplay behaviour; a complete described version may not be the first intervention.
- Explanatory video communicates product advantages or comparisons. It normally needs accurate captions, readable on-screen text and a text route to the essential audio and visual information. If a differentiator exists only in the picture, decide whether a descriptive transcript or audio description is required for the intended standard and use.
- Task video carries installation, operation, compatibility, safety or purchase-critical evidence. Audit the hear, see and operate channels beat by beat, and reserve time for description and reading during script and storyboard approval.
This classification is a production priority, not a legal exemption. The responsible team still needs to confirm which rules, contracts and standards apply to each release.
Put the deliverables in the brief and contract
For product content expected to travel across markets, request four connected asset groups.
- Picture and sound masters: clean master, required aspect ratios, editable caption-safe graphics, and separate dialogue, music and effects stems where the project allows.
- Text assets: approved dialogue transcript, SRT or VTT captions, speaker labels, meaningful sound cues, terminology glossary and a descriptive transcript when required.
- Expanded audio assets: audio-description script, described audio track or separate described video, with language and version identifiers.
- Release acceptance record: target page and player, keyboard path, caption controls, mobile collisions, language selection, final file names and named approvers.
The cost model should follow the work rather than a vague “accessibility fee”:
total accessibility readiness cost = master changes + caption and language review + description production + player implementation and testing + version maintenance.
Each term can have its own owner and acceptance gate. This also makes drift visible: if a product claim changes after final edit, the captions, transcript and description must be updated from the same approved version.
Teams building multiple market editions may also use the modular multilingual master workflow. Brands coordinating accessibility work with EU product-data programmes can keep model identities and claim sources aligned using the identity–evidence–presentation model described for Digital Product Passports.
Pilot the hardest representative sequence
Do not begin by remediating the entire archive. Pick one commercially important SKU, then select the hardest 20–30 seconds: dialogue, product action, on-screen text, a meaningful sound and a vertical crop in the same sequence. Deliver and test the captions, descriptive transcript, any required description, and the live player implementation.
The pilot should answer four questions:
- Does every purchase-critical fact remain available through at least one usable channel?
- Does a named reviewer approve terminology, numbers and product claims?
- Can users actually activate the delivered alternatives in the target player?
- When the edit changes, who updates every dependent asset?
If the answers are unclear, scaling the workflow will scale the defect.
The strategic change is not an unlimited proliferation of alternate cuts. It is moving responsibility upstream. Confirm the legal context first. Then use Hear–See–Operate to locate the real information gaps. Add text where text is enough; change the script, timing or delivery package where it is not. Accessibility then becomes a maintainable content capability, rather than a last-minute caption patch.
