92% of AI Nodes in Public n8n Workflows Were Silently Replaced

Description and use

92% of AI nodes in public n8n workflows were silently replaced with no-op placeholders, obscuring their presence and skewing AI usage statistics.

Problem

Public n8n workflow repositories may contain AI agent nodes that are silently replaced with no-op placeholders, leading to inaccurate counts of AI usage and potential reliability issues.

Outcome

A single squashed PR merge in the Zie619/n8n-workflows repository replaced 92.6% of AI/LLM node types with no-op placeholders, making AI usage invisible to type-based analysis while preserving node names. This misleads automated counts and may affect workflow reliability.

Workflow

• 1. Clone the Zie619/n8n-workflows repository at a pinned commit for reproducibility.
• 2. Filter workflow files for AI node types (@n8n/n8n-nodes-langchain.agent etc.) and count matches.
• 3. Observe zero matches by type but 711 matches by name.
• 4. Count nodes with AI/LLM names and verify 92.6% are typed as noOp.
• 5. Inspect git history to locate the commit that performed the overwrite (3c0a92c4, Sept 29 2025).
• 6. Confirm the commit changed 2,057 workflow files, replacing all LangChain-typed nodes with noOp while keeping names.
• 7. Validate that the issue is isolated to this repo by repeating the check on a different repo (enescingoz/awesome-n8n-templates) where no replacements occur.
• 8. Document findings and provide reproducible analysis scripts.
• 9. Discuss implications for automated AI node counting and workflow reliability.
• 10. Share the analysis code and corpus fetch script on GitHub (https://github.com/sevenedge-pl/n8n-reliability).

Evidence

92.6% of AI/LLM names were typed n8n-nodes-base.noOp

Source

The commit 3c0a92c4 replaced 2057 workflow files, turning all LangChain-typed nodes into noOp

Source