GitHub built an LLM-powered alt text plugin that checks quality, not just presence
martinwoodward · x · 2026-09-17
GitHub's engineering blog explains that most alt text checkers only verify an accessible name exists, not whether it says anything useful — alt="IMG2847.png" passes, and so does the same rating text pasted onto five different star icons. Per WebAIM's 2026 Million report, more than 1 in 4 images on the top million home pages have alt text that's missing, vague, or duplicated from a neighbor (16.2% missing; another 10.8% undescriptive among those that have it).
GitHub built an alt text plugin for its Accessibility Scanner that brings a model into the loop to judge whether alt text actually describes the image. The post covers where they drew the line between what a checker can prove versus only suspect, why their worst bug turned out to be a layout problem rather than a parsing one, and what changed once an LLM joined the pipeline. The tradeoffs transfer to building any automated checks, accessibility or otherwise.
More from coding & agent
- DeepSeek V4.1 Flash gotcha: Pi agents need explicit "input": ["text", "image"] config — solyarisoftware · 2026-09-17
- Open-source Skill Index compares MCP config drift across Claude, Codex and Cursor — CombinationOk2374 · 2026-09-17
- One-command MCP connector wires new Jev model into Claude Code and Codex — purealgo · 2026-09-17
- OpenMuse: open-source sandboxed computer lets AI agents work without vendor lock-in — aniketmaurya · 2026-09-17
- OpenMuse: open-source sandboxed computer lets AI agents work without vendor lock-in — aniketmaurya · 2026-09-17
- Open-source Muse app gives AI agents a local sandboxed computer with human takeover — aniketmaurya · 2026-09-17