How to Make Your Social Posts Accessible.

An accessible social post needs three things: a text alternative for every image, captions for anything with speech, and enough contrast that the words can be read.

That is the standard, and it is not complicated. The Web Content Accessibility Guidelines set it out, and RNIB turns it into practical instructions for social teams.

Almost every brand agrees with all of it. Almost none does it consistently.

The reason is not conviction. Accessibility fails as a workflow problem: a checklist item with no owner, remembered at the end by whoever happens to be uploading, at the moment they are most rushed. Move the step, name the owner, and the standard mostly takes care of itself.

What accessible actually means on a social post

Three WCAG success criteria cover almost everything a social team publishes.

• 1.1.1 Non-text Content, Level A: all non-text content has a text alternative that serves the equivalent purpose. Purely decorative content should be marked so assistive technology can ignore it, because alt text on a decorative item just distracts a screen reader user from the content.
• 1.2.2 Captions (Prerecorded), Level A: captions are provided for all prerecorded audio in synchronised media. 1.2.4 extends that to live audio at Level AA.
• 1.4.3 Contrast (Minimum), Level AA: text needs a contrast ratio of at least 4.5:1, or 3:1 for large text. 1.4.11 asks for 3:1 on graphical objects and interface components.

WCAG was written for web pages, not for a Reel. You do not control the markup of a social post, so treat these as the bar to clear rather than a spec to implement. The bar is still the bar.

Why almost every brand fails at this

Ask a social team whether accessibility matters and you will get an unhesitating yes. Ask who checked the alt text on last Tuesday’s carousel and the room goes quiet.

That gap is about sequencing, not belief. Everything else in the process has an owner, a stage and an approver. Accessibility has a good intention and a scheduling tool that will publish happily without it.

So it becomes the thing you do if there is time, which there never is at the point of upload. It is the same failure pattern as every other quality problem in a team that runs social without a system: the standard exists, but nothing forces it to be met.

Alt text: describe what the image is doing

RNIB is blunt about the mechanics. Screen readers cannot make sense of information in an image without help, so a photo with no description is a gap in the post, not a smaller version of it.

Their instruction is to write who is in the picture, what they are doing and what the key takeaways are. Not every detail, the point of the image.

The rule most teams miss: if there is text in the image, write it out in full. A quote card with no alt text is a blank rectangle to a screen reader, however good the quote is.

Alt text is not the caption repeated, and it is not a keyword line, which helps nobody and reads as noise when spoken aloud.

Text baked into a graphic is the most common failure

Stat cards, quote graphics, announcement tiles and price lists carry their entire message as pixels. If the alt text does not repeat that message, the post says nothing.

RNIB’s guidance on graphics is specific enough to design against: large text on solid backgrounds, no all capitals, good colour contrast, a font size of at least 32px (24pt), and a maximum of about ten words.

Ten words is the line most brand templates fail. A graphic carrying three sentences is a slide that escaped from a deck, and it is illegible on a phone before it is inaccessible.

Captions are not subtitles, and auto captions are a first draft

WCAG draws a distinction worth knowing. Subtitles carry the dialogue. Captions carry the dialogue plus the non-dialogue audio a viewer needs to follow the content: who is speaking, meaningful sound effects, music, laughter.

Auto captions get you the dialogue, imperfectly. Brand names, product names and place names are where they break, and those are exactly the words your video exists to say.

So the definition of done is edited captions, not generated ones. TikTok lets creators edit the generated text, LinkedIn lets you review captions before viewers see them, and YouTube lets you upload or type your own track. Where that control exists, an uncorrected transcript is a decision rather than a limitation.

The work pays twice, too: edited captions and a clear opening frame are also what make a post readable with the sound off.

What each platform actually gives you

Check these in the app you publish from, on the account you publish from. Availability and language coverage move, and language coverage moves fastest.

• Instagram: an alt text field behind Accessibility when you post, and on an already published post through the options icon, then Edit, then Accessibility. Reel captions are auto-generated by speech recognition and toggled in the reel’s advanced settings, including after publishing. Those caption controls are mobile only. Instagram’s own help page for the desktop version says the feature is not available on computers.
• Facebook: alt text is generated automatically and can be replaced with your own, before or after publishing, with a Clear option that reverts to the automatic version. Treat the automatic text as a fallback, never as your description.
• TikTok: auto captions are selected on the editing page after you record or upload, and the transcript can be edited once generated. Alt text is available on photo posts. Auto captions launched in American English and Japanese with more languages added since, so confirm your working language rather than assuming it.
• LinkedIn: auto captions are a toggle when you attach the video, with a second toggle to review them before viewers see them. LinkedIn lists support for English, Spanish, Portuguese, French, German, Italian, Indonesian, Turkish, Arabic and Dutch. Images take alt text at upload.
• YouTube: three routes into a caption track. Upload a file with or without timing, paste a transcript and let it auto-sync, or type captions manually against playback.
• X: RNIB documents an add description button on an attached image, and an Alt badge shown on images that carry a description.

RNIB adds two platform notes worth building into your templates. Instagram Stories have no image description field, so an image carrying real information should go out as a video clip with a voiceover instead. And YouTube subtitles should not be burned in: supply them as an SRT file so the player can handle them.

Contrast, and where it bites

4.5:1 for body text, 3:1 for large text, 3:1 for graphical objects. Those numbers decide whether people can read your graphic on a phone in daylight.

Brand palettes are where this goes wrong, because the failing combination is usually the one in the brand guidelines. Pale grey on white and mid-tone brand colour on brand colour both look refined in a design file and disappear at arm’s length.

For burned-in subtitles on social video, RNIB is prescriptive: solid colours, preferably white text on a solid black background, legible on all devices. That is a template decision made once, not a judgement call per video.

Make it a production stage, not a final check

The fix is structural rather than motivational.

Accessibility belongs in the process as a named stage that happens before approval, not as a habit you hope survives the last five minutes before publishing. An asset that reaches the approver without alt text and without checked captions is not finished, in the same way an asset with the wrong logo is not finished.

That means the approver rejects on it. If your reviewer will wave through a video with an unread auto transcript, the stage does not exist, whatever the process document says. Building that into the gate itself is exactly what a written approval process is for.

Give the step a named owner per asset type

One person cannot own all of this at the moment of upload, which is precisely why the current arrangement fails. Split it by who is already holding the file.

• Whoever writes the caption writes the alt text. They know what the post is for, so they can describe what the image is doing rather than what it contains.
• Whoever edits the video owns the captions and any burned-in text. Only they can fix timing, spelling and legibility.
• Whoever designs the graphic owns contrast, type size and word count. That is set in the design file, not rescued at upload.
• Whoever approves checks all three and sends it back if any are missing.

A definition of done, per format

Vague standards produce vague compliance. Write the specific version down and put it where the work happens.

• Static image: alt text written, and any text in the image written out in full. Hashtags in CamelCase at the end of the copy so screen readers read the words separately.
• Video with speech: captions edited word by word, speaker changes and meaningful sounds marked, and the video understandable with the sound off.
• Video without speech: on-screen information voiced or described, because RNIB’s test is that a video should be understandable just by listening.
• Audio-only content: a transcript, or edit it into a short subtitled video.
• Graphics: contrast checked, minimum 32px (24pt) type, no all capitals, around ten words maximum.

What good looks like

Nobody on the team talks about accessibility much, because it stopped being a debate and became a field on a form.

The alt text box is filled in the same tool where the caption is written, not in the app at upload. Captions are corrected in the edit, so the export is already right. Contrast was solved once in the templates, so a designer cannot easily produce a failing graphic.

The approver has one extra thing to check, and checks it. The published post is accessible because the process could not have produced anything else.

How NBK thinks about this

NBK treats social as an operating system, and accessibility is a stage in that system rather than a values statement in a deck. Standards do not fail because teams disagree with them. They fail because nothing in the workflow owns them.

So the questions are operational. Where does this step sit, who does it, what does done mean, and what happens when it is skipped. If your accessibility standard lives in a document rather than in your publishing steps, the document is not the problem, and rebuilding the workflow and ops around it is the work.

Next step

If your team knows what accessible content looks like but ships it inconsistently, the constraint is in the process, not the people. NBK can help rebuild the workflow behind the content so the standard holds without anyone having to remember it.