Skip to content
SilktideHelp

PDF image alt text

Silktide checks that the images in a PDF carry a text description. Charts, diagrams, photographs, logos, and infographics all say something, and a description is how that meaning reaches someone who cannot see it.

The description is , stored on the image inside the document's tags. It is never shown on the page, so it costs a sighted reader nothing. Images that are purely decorative need no description, but they do need to be marked as so that a reader knows to skip them rather than wondering what was there.

Why this matters

An image with no description is an absence, not a gap a reader can work around. Some readers announce nothing at all, so the content is silently missing; others announce "graphic" or read out the file name, which tells a reader something was there and refuses to say what. In a report where the only statement of a trend is a chart, this is the difference between having the finding and not having it.

The stakes rise with how much work the image is doing. A decorative border carries nothing. An organizational chart, a process diagram, or a map may be the entire point of the page, and a document can lose most of its value while still technically containing all its text.

Descriptions are also read by anything else that processes the document, so they improve how a document is indexed and searched, and they are what lets a reader on a slow connection or a text-only view still follow the argument.

How to fix it

  1. Add the description in the source document, then re-export. In Word, right-click the image, choose "View Alt Text", and write the description there. In InDesign, set it through Object Export Options. Keeping the description in the source means it survives the next revision.
  2. Or add it in a PDF editor. In Acrobat Pro, use the Tags panel or the Set Alternate Text tool to describe each figure.
  3. Mark decorative images as artifacts instead of describing them. A description of a background flourish is noise, and "decorative image" is not a useful thing to hear. See Decorative images.
  4. Or let Silktide write them. Silktide can generate descriptions for the images in a document, and shows you each one for review. See Image descriptions.

Write the description the image is doing, not the image. Say what a reader would miss:

  • A chart: the finding, not the shape. "Sales rose 40% from 2.1 to 2.9 million between 2024 and 2025" rather than "bar chart of sales".
  • A photograph: what matters here. The same photograph on a news page and a staff directory needs different descriptions.
  • A logo: the organization's name.
  • A diagram: the relationships it shows, in prose. Where that takes a paragraph, put the paragraph in the document text where everybody benefits, and keep the description short.

Avoid starting with "image of" or "graphic of", which readers announce already, and avoid describing a chart's colors when the point is its numbers.

How Silktide tests this

  1. Find the PDF documents linked from your website, and analyze each unique file once however many pages link to it.
  2. Find the figures in the document's tags.
  3. Report the document when a figure has neither alternative text nor replacement text.
  4. Report each document once, however many images are affected.

Images marked as artifacts are treated as decorative and are not reported, which is why marking decoration properly matters. This check tests whether a description exists, not whether it is any good; whether a description actually conveys the image is examined by AI review.

Troubleshooting

The image is decorative

Then mark it as an artifact rather than giving it an empty description, and it will not be reported. The document needs to state that the image is decorative; leaving it undeclared is what gets flagged.

The caption underneath already describes it

A caption is good practice and does not satisfy this check, because nothing connects the caption to the image for software. Where the caption genuinely says everything, the description can be brief and point at it rather than repeating it in full.

The image is a screenshot of text

Text in an image is invisible to a reader and is a poor way to publish it. Ideally put the text in the document. Where you cannot, the description has to carry the text itself, which is a strong argument for the first option. See Text drawn as shapes.

The document is a scan

Scanned pages are images of text, and describing them is not the fix; recognizing the text is. See Scanned documents.

Learn more

Last updated

Was this page helpful?