My AI agent copy-pasted SEO tags across 20 files instead of generating them A developer built a dependency-free Node.js SEO generator to stop an AI coding agent from hand-duplicating metadata across a 20-file multilingual static site for a North Cyprus property listing site. The tool's --check verification flag failed to distinguish intentional data edits from silent corruption, and adding a fifth language without matching path data produced broken hreflang URLs instead of an error. The developer open-sourced a genericized version of the generator on GitHub. I wrote a small generator to stop an AI coding agent from hand-duplicating SEO metadata across a multilingual static site. It has a --check flag that verifies the generated files still match their data source. The day after I shipped it, I found that when the generator silently corrupted all 20 pages, --check reported it as "out of date" — the exact same message it prints when you legitimately edit the data file and haven't regenerated yet. That's the part worth writing down. A verification flag that can't distinguish "you changed something on purpose" from "the tool just wrote broken URLs into every page you have" isn't verifying much. Here's how I got there. A free, direct-from-owner property listing site for North Cyprus — girne-apartment-sale.com https://girne-apartment-sale.com , no agency, no commission. Vanilla HTML/CSS/JS, zero npm dependencies, deployed to Cloudflare Pages. Four languages EN/RU/TR/PL , each language a real indexable URL rather than a client-side toggle, so 5 page types × 4 languages = 20 HTML files. Written together with Claude Code. I have a professional translation background, so my bar for "done" here was a genuinely correct multilingual site, not a page that renders. That bar is the only reason the rest of this happened — nothing below was surfaced by the agent on its own. The agent put