Blog · Accessibility
How to Write Alt Text for Screenshots with AI
Descriptive alt text takes time to write and easy to skip. Give the image to an AI model and get a solid first draft in seconds.
Screen-reader users experience your screenshots through the alt text you write for them. Good alt text says what the image actually shows — and for screenshots, that’s a meaningful description of the UI, the data, or the point you were making. Writing it well is tedious, so most people skip it or paste in something useless like “screenshot”. An AI model that can see the image fixes that.
Why alt text matters for screenshots
- Screen readers announce the image, and users rely on your description.
- Search engines index alt text — descriptive alt text doubles as image SEO.
- A screenshot of a chart or UI is meaningless without a real caption.
Method 1: paste the image into an AI chat
Open the Claude app (or claude.ai), attach your screenshot, and prompt “write alt text for this image in under 120 characters.” You get a draft you can refine. Good, and it works anywhere.
Method 2: generate alt text while you capture, inside iShot
If you’re documenting with screenshots regularly, the two-step capture-then-paste gets old. In iShot, take the capture and use the AI action to generate alt text from the shot itself — no switching apps. It’s powered by your own Claude API key or an OpenAI-compatible endpoint, so the image goes only to the provider you choose.
Prompt for a better draft
- Concise — “one sentence, under 120 characters.”
- Descriptive of the UI — “describe the screen and its main elements.”
- Data-accurate — “summarize what the chart shows, including the key figures.”
- Tone-aware — “match a friendly, plain tone.”
Frequently asked questions
Can AI write alt text for an image?
Yes. A vision model like Claude reads the image and writes a description to your spec — length, tone, and whether to include figures. Draft it with AI, then trim to fit your accessibility guidelines.
What counts as good alt text for a screenshot?
It says what the screenshot conveys rather than just its type — for example, “settings screen showing font size and theme options” instead of “screenshot”. Aim for the meaning, not a description of the pixels.