Recente wijziging in Discourse spam flagging?

AI-generated summary

neounix inquired whether there had been a recent change to Discourse’s spam flagging system, noting a sudden surge in spam messages requiring manual review—something uncommon over the past 8 months. They speculated it might relate to a specific GitHub commit linked in their post, and asked if it could be reverted. Roman responded by implementing a fix: users will now only be queued for review if their read time is under one minute, aiming to reduce false positives. neounix later confirmed that the spam queue had returned to normal levels since their initial report, suggesting the issue may have been temporary or resolved by Roman’s adjustment. The thread reflects a brief but effective troubleshooting interaction between a community admin and a Discourse team member to restore optimal spam filtering performance.

Hello,

Was there a recent change to the way Discourse manages spam?

I did not see any changes in the commit history, but maybe I missed it?

For the first time in months, we are seeing 5 -15 new user spam messages in the review queue (per hour), and all are spam which must be deleted.

For the past 8 months, most of these spam messages were blocked without a need to review them.

Did I miss a change?

How can we get the “great Discourse spam filtering” working again?

Thanks!

Update:

Could it be related to this change:

https://github.com/discourse/discourse/pull/12145/commits/4369065b7343943b78e223d7ac3fde121f375505

If so, can we revert this change; because the antispam filtering was working great before and now our “review the junk” queue is filling up?

I just made a change to only queue users if their read time is lower than one minute. This will greatly reduce the number of false-positives.

Thanks for the feedback @Roman

On our end, we have not seen our “spam review queue full of spam” since I posted this “issue”, so I’m hoping it was just some anomaly, as everything is back to normal at this time.