Writing alt text for images with AI

Alt text is a short written description attached to an image. Screen readers read it aloud to members who are blind or have low vision, and browsers show it when an image fails to load. Search engines use it to understand what an image shows. Without it, a member using a screen reader meets your carefully chosen lesson image as a file name or silence.

Most membership libraries have hundreds of images, and many of them have no alt text at all, because writing it one by one feels endless. AI can describe images quickly and help you clear that backlog. But an AI description is not the same as good alt text, and the difference matters to the members who rely on it. This guide shows how to use AI well for the job.

What good alt text actually does

Good alt text conveys the purpose of an image in its context, not everything that is in it. The same photo might need different alt text on two different pages. A few principles:

• Say what matters for this page. On a lesson about kneading dough, the texture of the dough matters; the color of the countertop does not.
• Keep it short. A sentence or two is usually enough.
• Skip “image of” and “picture of.” Screen readers already announce that it is an image.
• Include any text in the image, such as a quote graphic or a sign that matters.
• Mark decorative images as decorative. Purely decorative images, such as a background pattern, should have empty alt text so screen readers skip them. Most membership platforms and editors offer a way to do this.
• Complex images need more. Charts and diagrams need a longer description or the data in the page text, with the alt text pointing to it.

How AI helps, and what it misses

General assistants such as ChatGPT, Claude, Gemini and Copilot can look at an uploaded image and describe it accurately in most cases. That makes them useful for drafting alt text in bulk. What they do not know is why the image is on your page. Left to themselves, they describe everything equally: lighting, background, colors, objects. The result is accurate, long and often misses the point.

Take Rustic Loaf Bakers, a bread-baking membership, and a photo in a lesson about proofing:

• No context: “A round loaf of bread cut in half on a wooden board, with a knife and a scattering of flour, in warm light.”
• Good alt text: “Sourdough loaf cut open to show an even, open crumb with large irregular holes, the result of a long cold proof.”

The second tells the learner what they would learn from seeing the photo. That is the job.

Give the assistant the context

The fix is to tell the assistant where the image appears and what it is for:

Write alt text for the attached image. It appears on [page or lesson title], which teaches [describe your members] how to [what the page teaches]. The image is there to show [the point it makes]. Write one or two sentences, under about twenty-five words, that convey what a sighted member would learn from it in this context. Do not start with image of or picture of. Include any text visible in the image that matters. Do not guess anyone's identity, age, ethnicity, gender or emotions. If the image looks purely decorative for this page, say so instead of writing alt text.

For a batch of images, list each with its page and purpose, and review the results together.

Watch for these pitfalls

• Guesses about people. AI may describe someone's ethnicity, age, gender, disability or mood based on appearance. Unless it is relevant and you know it to be accurate, leave it out.
• Naming people. Only name someone if you know who they are and it is relevant, such as a guest expert on their session page.
• Invented detail. Assistants occasionally describe something that is not there, or misread text in the image. Check against the image every time.
• Charts described vaguely. “A bar chart showing growth” does not help. Put the key figures in the text or a table on the page.
• Member photos. Photos of members are personal data. Use a business plan with training switched off, check the vendor's terms and never upload member photos to a consumer account.

A person reviews every description

Alt text is only useful if it is right, so a person should approve each one before it is published. A quick review:

1. Look at the image and read the alt text. Does it capture the point in context?
2. Check any visible text is transcribed correctly.
3. Remove any guesses about people.
4. Cut anything that does not help, such as decorative detail.
5. Now and then, listen to a page with a screen reader. Phones and computers have one built in, and hearing your alt text read aloud quickly shows what works.

For diagrams and infographics, such as those in diagrams and infographics with AI, write a fuller description in the lesson itself as well as alt text.

Your first steps

• List the pages members use most and check which images lack alt text.
• Mark decorative images as decorative.
• Draft alt text for ten images using the prompt above, with page context.
• Review each against the image and fix anything wrong.
• Add alt text to your publishing checklist so new images are never missed.
• Try a screen reader on one lesson page.

Alt text is one part of a wider commitment. For the principles behind it, read AI and accessibility: helping, not excluding, and for file names and social images, see image SEO: file names, alt text and social images.