Multi-agent content engine for SEO and AI search

A multi-agent writing pipeline that took a brand-only website from about 50 to 2,000 AI-search impressions a day in one quarter.

In July 2026 the site got 233 Google clicks per 90 days, almost all from people searching the brand name, and it never appeared in AI answers. I designed and built a content pipeline with Claude Code: DataForSEO picks topics, one agent writes a data-backed brief, another drafts in Dutch and English, and a separate model checks every figure against the source, because AI drafts drift on numbers. A person approves at 4 gates before anything goes live.

51 posts in two languages in 3 months

Quality gates in Python: answer-first openings, banned phrases, coverage against the top 8 results

A Markdown copy of every post, llms.txt and schema for AI crawlers

Clicks 233 → 1,280 (+450%), impressions 20.4k → 307k per 90 days

Impressions in Google's AI features: ~50 → ~2,000 a day, 80k in the quarter