Alt text, captions and transcripts are often bundled together as if they were interchangeable accessibility add-ons. They are not. Each carries information through a different channel, and each is useful to people who may not be able to access an image, hear a recording or follow a video in real time. Knowing which tool does what helps editors avoid a common mistake: adding one text feature and assuming every missing detail is now covered.

Alt text explains the image’s purpose in this story

Alt text is a short text alternative attached to an image. Screen readers can announce it in speech or braille, helping blind and some low-vision readers understand relevant visual content. According to the W3C Web Accessibility Initiative’s Images Tutorial, what the alternative says depends on how the image is used and what the surrounding page already explains. The goal is not to catalogue every visible detail; it is to convey the image’s information or function in context.

For an informative photograph, describe what matters to the article. A story about a flooded street might need “Residents carry sandbags past water covering a city intersection,” rather than “A street.” A profile about an identified public figure should name that person if identity is clear and relevant, while avoiding guesses about someone’s identity or feelings. A chart, map or diagram may need its key finding in nearby text as well as a concise alt description, because a short field cannot always carry a complex set of data.

Not every image needs a spoken description. When an image is only decoration, or when adjacent text already gives its information, mark it with an empty alt attribute. That tells assistive technology to skip it. The W3C’s decorative-image guidance cautions against simply leaving the alt attribute out; some screen readers may announce the filename instead. The editorial question is whether the image adds meaning here, not whether it is technically a photograph or illustration.

Function matters, too. If a linked image is the only content of a button, its alternative should identify the action, such as “Search,” rather than say “magnifying glass.” If the link already reads “Search,” the icon can be decorative so it is not announced twice. This is why meaningful alt text is written after considering the headline, caption, nearby copy and link destination, not in isolation from them.

Captions follow sound; transcripts make a readable record

Captions appear in sync with a recording. They give Deaf and hard-of-hearing viewers access to speech and relevant non-speech audio, such as a siren, a door slamming or a speaker’s laughter when it changes the meaning. W3C’s captions guidance also notes that captions help people who process written information more easily than audio. They are useful beyond disability contexts, too: a viewer may be in a noisy station, a quiet classroom or simply watching with the sound off.

A transcript is a text version people can read apart from the media. For a podcast, that usually means the speech and important sounds, with speakers identified where needed. It can be read at the reader’s pace, revisited or searched, and a well-structured transcript can provide headings and links. According to W3C’s transcript guidance, a descriptive transcript for video also includes visual information needed to understand what is happening. That extra layer matters for people who are Deaf-blind and may access text through braille, as well as readers who find audio or changing visuals difficult to process.

Here is the first comparison: imagine a short video of a community meeting. A speaker says, “We have reached an agreement,” while a chart on screen shows the new bus route. Captions can convey the sentence and an audible chair scrape; they do not automatically describe the chart. Alt text might explain that chart if it is published as a separate image, but a single still image description will not capture changes throughout the video. A transcript can combine the speech with the route details and key visual changes, while captions remain timed to the spoken words.

Or consider a social post with a photo of a protest and a short video clip from the same event. The photo’s alt text can say what distinguishes the scene—such as a line of marchers holding a banner with a legible, relevant message. Captions for the clip should identify speech and significant sounds. A transcript can set out the clip’s words and describe meaningful visual details, such as a sign changing or a speaker pointing to a map. The three supports may share information, but they are not duplicates: one describes a still image, one tracks audio in time, and one offers a fuller text account.

When that visual context is missing, readers can miss the point rather than merely the picture. An article may say “the sign sparked debate” while the sign’s wording appears only inside an unlabelled image. A screen-reader user may reach the sentence without ever hearing what the sign said. The same gap can affect someone who has images turned off, uses a small screen, or relies on a transcript instead of a video. If a visual detail changes the meaning, identify it in an accessible place: concise alt text for an image, descriptive transcript or audio description for video, or ordinary article text that everyone can read.

A practical workflow for writers and editors

Start by asking what information each asset contributes. For every image, identify its role in the article, then choose informative alt text, a functional label, or an empty alt value when it adds nothing beyond adjacent text. Check whether the description is accurate, useful in the article’s context and free of unnecessary visual inventory. If a chart carries essential evidence, put its main takeaway and usable data in the story too rather than expecting a one-line alternative to do everything.

For recorded audio and video, plan captions and transcripts as separate deliverables, even if both begin with the same speech transcription. Captions need timing and relevant sound cues; transcripts should be readable as a standalone record and include visual information when the video depends on it. W3C recommends providing both captions and a separate transcript. Automatic speech recognition can speed up a first draft, but its output needs human review: a missed name, negation or sound cue can alter the meaning. Editors should also check that the caption control works and that the transcript is easy to find near the player.

Finally, review the story without relying on the image or video. Can a reader still identify the people, understand the chart’s point and follow the event’s sequence? Then review the media itself with captions on and compare the transcript against both sound and visuals. This process will not make every experience identical, and checking a few text features cannot guarantee that a site or story is fully accessible. It does, however, help teams catch specific information gaps before publication—and makes the story more usable for a wider range of readers.