Multi-Version Advocacy Content Generator with AI Scoring - n8n Workflow

• Content marketers needing to A/B test high-volume messaging quickly.


• Animal advocacy organizations seeking data-driven insights on content effectiveness.


• Users looking for advanced examples of nested n8n workflow integration.


• Developers and automation specialists building custom AI solutions using n8n templates.

Generating effective advocacy content requires testing many variations to find the messaging that resonates most strongly. This complex n8n workflow automates that entire process. It starts by accepting detailed user specifications (topic, tone, style) via an n8n trigger. It then executes a research agent sub-workflow that uses external tools (like Serper and the Open Paws database) to create 10 unique, full-length content variations using large language models like Claude (via OpenRouter).

Crucially, this n8n workflow then passes each generated piece to a dedicated scoring sub-workflow. This external scoring mechanism analyzes the text based on metrics favored by animal advocates (impact, persuasiveness) and often leverages fine-tuned Hugging Face models for real-world performance prediction. The final output of this n8n automation is a list of all 10 content pieces, each tagged with its relevant AI score, allowing users to prioritize the most impactful messages for deployment. This comprehensive approach transforms content creation from guesswork to data science.

This powerful n8n workflow operates in several distinct stages:

1. Input Trigger: The process starts with the 'When Executed by Another Workflow' n8n trigger node. This allows the n8n workflow to be called by a parent flow, receiving structured inputs such as contentType, Topic or Instructions, Tone, and Style.


2. Content Generation: The 'Create Content' n8n node executes a linked 'General Research Agent' sub-workflow. This agent performs research and is instructed to generate exactly 10 full-length, distinct content versions (e.g., 10 blog posts) and return them as a raw text string.


3. Extraction and Structuring: The output is passed to the 'Extract Content' n8n node, which leverages the OpenRouter Chat Model (using the Claude-Sonnet-4 model) to parse the raw text output into a clean JSON array of 10 objects, ensuring all formatting constraints (like BB-code) are met.


4. Parallel Processing: The 'Split Out Content' n8n node splits the resulting array into 10 individual items, preparing each item to be scored independently. This is a crucial step for efficient n8n node processing.


5. AI Scoring: The 'Score Text' n8n node executes another sub-workflow ('Score Text') for each of the 10 items. This scoring sub-workflow evaluates the text's advocacy impact, returning a performance score for that specific content variation.


6. Final Aggregation: The 'Aggregate' n8n node collects all 10 individual items, now enriched with their performance scores, into a final, ranked list. This makes the output easily consumable by the next n8n workflow steps, such as automated posting or internal reporting.

To deploy this advanced n8n workflow, follow these steps:

1. Import the n8n workflow: Copy the provided JSON data and paste it directly into your n8n instance via the 'New' menu -> 'Import from JSON'.


2. Set Up Sub-Workflows: This n8n workflow relies on two external sub-workflows: the 'General Research Agent' (ID: k053fXGjIF7dUIQZ) and the 'Score Text' evaluator (ID: RyxYPLmF5sWDhC2Z). You must ensure these two linked n8n templates are also imported and active in your n8n environment.


3. Configure Credentials: The 'OpenRouter Chat Model2' n8n node requires an OpenRouter API credential. Set up a new OpenRouter connection in your n8n credentials panel and link it to the node.


4. Testing: Since the primary n8n trigger is 'When Executed by Another Workflow', you can test it by manually executing it and providing mock data in the 'Pin Data' section, or by calling this n8n workflow from a primary n8n template.

When Executed by Another Workflow (n8n trigger node):
Function: Serves as the entry point, receiving pre-defined user configuration (content type, topic, tone).
Key Configuration: Defines five expected input fields to guide the content generation process.
Create Content (executeWorkflow n8n node):
Function: Delegates the heavy lifting of research and initial content generation to a dedicated sub-workflow ('General Research Agent'). It instructs the sub-workflow to generate 10 variations.
Key Configuration: Passes all user inputs and a complex instruction set via the chatInput field.
OpenRouter Chat Model2 (n8n node):
Function: Provides the LLM backbone for structuring and extracting data.
Key Configuration: Uses the anthropic/claude-sonnet-4 model for high-quality, reliable JSON parsing.
Extract Content (informationExtractor n8n node):
Function: Takes the raw text output from the content generation step and forces it into a strict JSON array format of 10 objects, essential for subsequent processing in this n8n workflow.
Key Configuration: Uses a critical system prompt to ensure the output is pure JSON, without markdown wrappers.
Split Out Content (n8n node):
Function: Deconstructs the single JSON array into 10 separate items, allowing the subsequent 'Score Text' n8n node to process them in parallel.
Score Text (executeWorkflow n8n node):
Function: Executes a specialized sub-workflow to evaluate the advocacy performance of each content piece individually.
Key Configuration: Set to run in each mode, processing one content item at a time.
Aggregate (n8n node):
* Function: Reassembles the 10 scored content items back into a single, cohesive payload for final output of the n8n workflow.