n8n Workflow: Cited Research Reports via DuckDuckGo + OpenRouter

Get citation-verified research reports inside n8n—using DuckDuckGo web evidence and an OpenRouter LLM

This n8n workflow turns a question submitted via an n8n Form into a structured research report where every claim is grounded in sources pulled from the web with DuckDuckGo—and every quote is checked against the page text before it’s allowed into the final output. Self-hosted n8n only.

What this workflow does

• Collects your question and a depth selection from an authenticated n8n Form.
• Breaks work into sub-questions by using an OpenRouter model to generate 3–5 targeted sub-questions and corresponding DuckDuckGo search queries.
• Searches and fetches evidence with the DuckDuckGo Search community node, retrieving top results with page text, grading usability, and pacing requests to reduce rate limiting.
• Manages sources by deduplicating results, numbering sources as S1, S2…, and stopping early with a notice if there are too few usable sources or too many searches are blocked.
• Extracts findings with exact quotes per sub-question, then programmatically verifies that each quote exists in the corresponding source text—discarding unverified findings.
• Generates a Markdown report that includes citations by source ID only, removes uncited/invalid sentences and links/HTML, adds an evidence-coverage table, and provides a clickable source list.
• Renders the report to HTML and displays it back in the form response (or returns a “no report” explanation).

Use cases

• Building evidence-based research summaries for internal teams without manual source checking.
• Generating cited reports for SaaS due diligence (features, benchmarks, claims) using web evidence validation.
• Automating literature-style briefings where quotes must be verifiable in the source text.

Technical details

• Workflow automation: n8n nodes including if, code, form, wait, markdown, and sticky note
• Web search & evidence: DuckDuckGo Search community node (top results with page text, grading, pacing, deduplication)
• LLM processing: OpenRouter model for sub-question planning, quote extraction, and citation-aware Markdown generation