Text Classifier node is outputting two branches when I have "Allow Multiple Classes To Be True" disabled

Questions
1

Describe the problem/error/question

Text Classifier node is outputting two branches when I have “Allow Multiple Classes To Be True” disabled.

Loom video of issue: Automating Client Responses with Text Classification 🤖 | Loom

Please share your workflow

Share the output returned by the last node

Information on your n8n setup

  • n8n version: Cloud
1 Like
2

Hey @Nick_Steeves Welcome to the community!

It’s likely the model which you are using which is the cause of the error. Under the hood, the LLM adheres to the generated prompt and returns the matched results which then n8n only displays as the result of the node. So the quick answer is to try to use a more powerful LLM.

Alternatively, if you do find this happening quite often it could be that you’re trying to match too many things in one text classifier. My suggestion is to try and break them down by context which could make it simpler to reason about. In the following example, I added a simple split between existing and new customers.

3

Thanks Jim!

  • Which model do you recommend? (I’m using gpt-4.1-mini)
  • Do you have a recommendation for maximum number of categories per classification node?
4

  1. Naturally, GPT-5 would be the obvious upgrade but Claude 4.5 and Gemini 3.0 would be powerful options as well.

  2. In theory, I don’t think there is a maximum as long as all your categories can remain wholly exclusive - which is very difficult if not near impossible. In any user (“human”) query, there’s bound to be some overlap to a degree which confuses the AI so you can never be sure.

5

Ok gotcha, thank you!