Study data
The 123 frontends we audited
Every repository in the study, listed alphabetically with its findings. These are public projects that identify as AI-generated. Open any of them on GitHub and check for yourself.
Errors are serious: they should stop code from shipping. Warnings and info are lower-severity. Because the audit is deterministic, these numbers are stable, not a snapshot of one run.
How this list was built
We searched public GitHub for repositories that self-identify as AI-generated (v0, Lovable, bolt, and similar), across 150+ accounts. Of 165 cloned, the 123 shown here had an analysable frontend and form the study’s sample; the rest were empty or backend-only and were excluded.
Every one was checked with the same deterministic tool: no model, no network, no judgement, so the same repository yields the same numbers every time. Repositories can change after publication, so a row may drift if its owner edits or deletes it.
Technical, and want to re-run it? In any project root: npx @imrandwc/dwic audit