Posts: 353
Threads: 5
Joined: Sep 2019
Reputation:
23
The spammers now use ChatGPT to generate more-or-less sensible replies to forum threads and sneak a link into them for spam or SEO. While the industry is exploring ways to counter this (including using ChatGPT itself as a spam detector, ha), I'm pondering about a different approach based on early trapping:
The spammers seem to generate only one post per user account, so how about configuring the system so that a new user, for the first post, has access to posting to only one special forum folder called The Passage or something, to say hello. The bots will post whatever they want in there and stop short.
Posts: 3,173
Threads: 37
Joined: Aug 2013
Reputation:
205
putting barrier to new user is a sure way not to get new user i would rather have to click a button that say i am not a robot or if that not enough a small puzzle everytime i want to come on than make it harder for people to join...
My Sketchbook
Perfection is unmeasurable therefor it impossible to reach it.
Posts: 353
Threads: 5
Joined: Sep 2019
Reputation:
23
Darktiste, the "I am a human" system has never really worked, the spammers hire people just to complete the registration process on thousands of forums that use it before their bots take over the account.
On the other hand, well intentioned people will understand the small inconvenience of posting a hello message. I hope.
Posts: 353
Threads: 5
Joined: Sep 2019
Reputation:
23
You may be right. I have bad memories of hunting spammers when I was a mod on other forums in the past, this has probably made me oversensitive to the infestation.
Some of the latest spammers have a much smarter strategy though, they post long sensible (but bland) answers written by ChatGPT to stay under the radar - and add their links in the user profile or inconspicuously inserted in the message body.
Posts: 432
Threads: 70
Joined: Mar 2020
Reputation:
69
I think that's a good idea, Leo. I've used other forums that had a similar requirement for new users, and I never personally had an issue with it.
The only issue I can see arising is we start getting bots that make multiple posts, but I have no idea if their sophistication extends beyond using ChatGPT to generate ALMOST sensible-sounding replies.
Posts: 623
Threads: 36
Joined: Jan 2012
Reputation:
50
You're right Leo, sometimes I'm not even sure myself if some posters are bots or not, their posts are definitely harder to distinguish than before. I'm trying to figure something out, maybe even something similar to what you've suggested or having to go through me to register. For now, I'll pay close attention to new registered users.
Posts: 802
Threads: 6
Joined: May 2018
Reputation:
109
I feel like I remember seeing bots that would kind of regurgitate other people's posts as well, even before there was AI like chatgpt available. I honestly don't even see why they exist in the first place, but it's definitely annoying lol
Posts: 353
Threads: 5
Joined: Sep 2019
Reputation:
23
@P.E. - They sure will get more sophisticated as counter-measures are put in place, it's an endless battle that might end up with sticks and stones.
@Dennis - I remember one that started with a flood of general advice, then, out of the blue, a spam link to whatever. Let us know if we can help in any way so you don't bear all the burden.
@Joseph - One of these recently landed onto an old CHOW thread. As for why they do this: Apart from trying to get visitors, it's also a way to boost their page ranking in search engines by having links pointing to them.
Posts: 105
Threads: 1
Joined: Apr 2021
Reputation:
18
I don't think it's as much of a issue of spotting bots as it's barring them from registering.
They tend to follow patterns, like the latest ones: Few minutes after registering they post something vaguely related to a thread, usually an inactive one. They're not the kind of active bots which keep posting after the initial spammy one, and they all have discrete spam links in their signatures instead of placing it on the body of the post. Less conspicuous than the average bot but not that much.
We're not big enough spammers would find it worth employing actual humans to analyse and work around bot restrictions. Probably a captcha upon registering is enough to block them. A custom captcha of the "write the name of the objects in this image" kind using a random pick out of 5 images not found anywhere on the internet is probably even better, as automated networks may have invested into breaking widely used captchas but would be near powerless against something unexpected without using humans, which I don't think they'd.
Posts: 353
Threads: 5
Joined: Sep 2019
Reputation:
23
A custom captcha might do the job, Knight.
Or, maybe, a non-aligned AI that lures the spammers into its dark den then exposes their remains at the entrance as a warning.
Posts: 353
Threads: 5
Joined: Sep 2019
Reputation:
23
Just saw a new strategy. There was a chat-gpt-like reply with no spam link, but I just looked at it again later, it has been edited with a "hello i'm new, and here is a link to a definitely not illegal site."