Describe the purpose, not just the picture
Alternative text replaces the image for someone who can't see it. The question to ask is: if I removed this image and put text in its place, what would that text need to say? The same photograph needs different alt text in different documents:
- In a staff newsletter: "Maria Lopez receiving the 2026 Service Award from Director Chen."
- In a facilities report: "The east entrance ramp, showing the cracked concrete at the base described in section 3."
Both describe a person by a building. Each says what that document needs the image to say.
General rules
- Be concise. A sentence or two is usually right. If you need a paragraph, the information belongs in the body text or a long description (see below).
- Don't start with "Image of" or "Picture of". Screen readers already announce that it is a graphic. Do say "Photo:", "Map:" or "Screenshot:" if the type of image matters.
- Don't repeat the caption. If a visible caption already says what the image shows, the alt text can add what the caption leaves out, or stay short.
- Include any text in the image. If an image contains words, such as a sign, a slide or a scanned letterhead, the alt text must include them.
- Write in sentences, with punctuation. Screen readers pause at full stops, which makes alt text easier to follow.
- Never use the file name. "IMG_4471.jpg" and "chart1" are worse than nothing.
When an image should have no alt text
Images that add nothing to the content should be marked decorative (in Word) or as an artifact (in a PDF), so screen readers skip them entirely. That covers:
- Ornamental borders, dividers, background textures and flourishes.
- Stock photos used only for mood, such as a generic handshake on a cover page.
- Icons that sit next to text saying the same thing, like a phone icon beside "Phone: 555-0100".
- A logo repeated on every page in the header.
Marking an image decorative is better than giving it filler alt text like "decorative image", which a screen reader still reads out.
Common image types
| Image | What to write | Example |
|---|---|---|
| Logo | The organisation's name, plus "logo" if that matters. If the logo is a link, say where it goes. | "Department of Example" |
| Photo of a person | Who, and what they are doing if it is relevant. | "Dr. Ana Ruiz presenting the survey results." |
| Screenshot | The point being illustrated, plus any text the reader needs. | "Screenshot: the Save As dialog with the Options button highlighted." |
| Signature | Whose signature it is. | "Signed, J. Smith" |
| Icon used as a button or link | The action, not the picture. | "Download the report", not "arrow" |
| Image of a table or text | Replace it with real text where you can. If you can't, the alt text must carry the content. | — |
Charts, maps and diagrams
A chart can hold more information than a sentence of alt text can carry. Use two layers:
- Short alt text that says what the chart is and its main takeaway: "Bar chart: applications rose every year from 2021 to 2025, from 1,200 to 3,400."
- The full data or explanation in text, either a table next to the chart or a paragraph in the body, so every reader can get the detail. Mention in the alt text where to find it if it isn't immediately next to the chart.
For maps, describe what the map is meant to show, such as "Map: the six counties in the service area, all in the northwest of the state", rather than every road. For flowcharts and process diagrams, a numbered list of the steps in the text is usually the clearest alternative.
Reviewing generated alt text
Word, Acrobat and remediation tools, EZ508 included, can now suggest alt text automatically. These suggestions are a good starting point for common images, but a tool can't know why you included an image, and it sometimes gets facts wrong. Before publishing:
- Read every generated description against its image.
- Add names, places and figures that matter in context; a tool won't know who is in a photo.
- Mark purely decorative images as decorative instead of keeping a description.
- Check charts carry their takeaway, not just "a bar chart with blue bars".
EZ508 builds its suggestions from each image's caption and the surrounding text, and lists them for review before you download the file.