Content production.
5 workflows
Programmatic page generation at scale
500+ indexable city × service pages from one data file. A Claude Agents SDK loop (writer + fact-check + voice) tailors each page so it reads human and dodges the scaled-content-abuse penalty. Ships with schema, hreflang, and OG.
- Cadence
- Weekly batch
- Output
- 20–40 indexable pages / week
Citation-first blog pipeline
Multi-agent draft via Claude Agents SDK + Vercel AI SDK + Browserbase. Every post is researched, fact-checked, voice-matched, schema-tagged, then pre-tested against ChatGPT, Perplexity, and AI Overviews before it ships.
- Cadence
- 2–4 posts / week
- Output
- 1,800–2,400 word posts
AI content humanization
Per-brand voice.json fingerprint + deterministic AI-tell detector + Claude chunked rewrite. Strips em dashes, "leverage", hedging language, template-stamped openers. Drafts pass both editor review and AI detectors without losing speed.
- Cadence
- Every published draft
- Output
- Voice-matched humanized drafts
On-brand image generation at scale
Locked brand prompt → Nano Banana 2 generation → vision-verify QA → alt-text + ImageObject JSON-LD → CDN push. Every page ships with on-brand 4K visuals that LLM crawlers can actually read.
- Cadence
- Per page + weekly
- Output
- Brand-locked images + schema
Localisation pipeline (EN → FR / DE / AR / ES / PT / JA)
Flat key extraction → translation → in-context preview → ship. Brand literals (ChatGPT, AY Rank, GEO) stay Latin in every locale, hreflang and canonical generated automatically.
- Cadence
- Per page
- Output
- 7 locales per page
