Draft image alt text and edit it before publishing
At a glance
A blogger uses Claude to propose image descriptions, then checks and revises each one for readers. This workflow is listed under Accessibility, Content & Creative.
- Agents / tools
- Claude
SourceBuilding a tool for these workflows? See weekly rates.
What they did
1) Open an image-capable Claude conversation and prepare the image being published. The author used Claude 3.7 Sonnet. 2) Give reusable instructions to describe pasted images, place the alt text on a single line in a code block, and begin with the image type. 3) Paste the image and generate a draft. 4) Read it against the image and the purpose of the surrounding post. Edit what matters for a screen-reader user before placing the reviewed description in the image's alt text.
How it went
In the author's table-screenshot example, Claude described the interface, column types and represented states instead of repeating all 88 numbers. The author reports better descriptions after human revision.
Worth knowing
Claude did not follow the instruction to include all image text exactly. The author treats output as an editable draft.
