For developers

For developers who inherit builder-made sites

A client or a founder hands you a Lovable or v0 project and asks why nothing ranks. The audit tells you exactly which files carry the defect, and the repair opens the diff so you review instead of hunt.

Metadata lives in five places

Next metadata exports, Vite index.html, react-helmet, builder injections and hosting rewrites disagree with each other.

robots.txt was generated, not written

Blanket Disallow rules, wildcard groups that catch search crawlers, and a preview-host robots file shipped to production.

Regressions after every prompt

The next builder session rewrites the file you just fixed. Without a check on every push, you find out from the client.

The checks that matter here

  • robots.txt per named crawler, with the deciding line
  • Canonical and preview-host leaks
  • Raw versus rendered content delta
  • JSON-LD parse and entity types
  • Sitemap accessibility
  • Regression watch on push

What the repair does

Typed transforms with ts-morph, matched to the file quote style and indentation, on a new branch. The PR body quotes the finding and the evidence; the target repository CI runs on it. Training-crawler tokens are never modified and no Disallow is opened on a preview host.

Audit my site

Questions

What stacks are supported for repairs?
Next.js App Router, TanStack Start and React with Vite, plus static robots.txt, llms.txt and sitemap files in any stack. Other stacks get the report and the exact recommendation.
Is the build verified before the PR opens?
Not on our side. The pull request runs the repository CI; a failing check is the signal to close it. The transforms are unit tested against fixtures.
Can I trigger checks from CI?
A watcher listens for pushes through the GitHub App and re-checks the live site after a grace period. An API for triggering runs is on the list.

Also: agencies, founders, marketers, online stores, saas.