AEO Content Engine v2.0

Problem

Hiring a content team to publish daily across 6 platforms costs $8–15k/mo. Most AI content is generic, ranks poorly in AI Overviews, and gets no citations. The brand wanted production-grade content on autopilot with citation-first structure.

Approach

Built a two-pipeline system: a Daily 11-node flow for fast social posts and a Weekly 51-node flow for long-form articles + newsletters. Fact-check step added via Serper-backed citation lookup. Error handling with 3 retries at 5s on 19 critical nodes. Output enforced to AEO structural patterns (lead answer, source citations, entity-rich headings).

What shipped

• n8n Daily JSON (11 nodes) + Weekly JSON (51 nodes)
• Gemini header-auth credential (reusable)
• Serper live citation fact-check node
• Merge-before-publish branch consolidation
• SMTP error alerts on failure
• 5 GPT-4.1 → GPT-4o-mini swaps (~60% cost cut)
• Ops runbook + credential setup guide

Outcome

Ships daily and weekly content across platforms without a human in the loop. Research → write → image → fact-check → publish — all runs for roughly $0.30 per full cycle after swapping 5 GPT-4.1 nodes to GPT-4o-mini. Gemini used for research, Serper for live citations, SMTP for team alerts.

Lessons

Serper citations beat web-scraping for fact-check reliability by a wide margin. GPT-4o-mini is production-ready for 90% of content nodes if prompts are constrained. Merge nodes before the last publish step, not after, to avoid race conditions on Friday runs.