Conseils pour prévenir le spam

Documentation Gestion de site
, ,
15

Hi @Overgrow,

A few ideas you could try here to prevent this:

  • Use Discourse AI - AI triage to Set up spam detection in your community to detect this type of content
  • Add URL shorteners and Amazon affiliate link patterns to your blocked watched words list
  • Lower the newuser spam host threshold and increase requirements for TL1
  • Reduce max new accounts per registration IP and enable flag sockpuppets
  • Use the Discourse hCaptcha plugin to help prevent automated spam/AI registrations on your site.
  • Consider placing all new user content in the review queue until the attack subsides by adjusting:
    • approve post count
    • approve unless trust level
    • approve new topics unless trust level

The approach here will be similar to preventing spam in general, with more of a focus specifically on the shortened URLs, and AI generated content.

For your case here, you could try using an AI prompt for specifically detecting AI content like the following:

You are a spam detection system. Analyze the following content and context.

Notes:
- Replies must remain relevant to the discussion thread.
- Mark as SPAM if the content is irrelevant, promotional, or automated.
- Consider new user posts with links as potential SPAM unless explicitly relevant to the topic.

Watch for content that appears authentic but has unnatural patterns. 
Look for text with peculiar phrasing, excessive formality mixed with 
casual language, or generic advice that doesn't quite fit the context. 
Flag content containing hidden affiliate links, especially when the post 
seems designed to naturally lead to product recommendations.

Pay special attention to these red flags:
1. Content that poses as genuine advice requests but contains promotional elements
2. Posts that introduce a problem and then suggest specific products as solutions
3. The presence of URL shorteners (bit.ly, tinyurl, t.co, goo.gl, etc.) which may disguise affiliate links
4. Amazon product links or references, especially with affiliate parameters (tag=, ref=, affiliate=)
5. Content that seems to ask for recommendations but subtly steers toward specific products
6. Artificial quality text - overly formal language mixed with casual expressions or awkward structure
7. New accounts posting content with any of the above patterns

Respond only with "SPAM" or "NOT SPAM".