Why have the bots taken over MP?

Why are bots seemingly commenting on every forum? who's making these bots and why? I feel that whoever is taking the time to make bots that comment on MP forums has to have a reason, right?

Russia?  

I work for a relatively niche and small company that makes high end luxury products. We got a lengthy email request today to make a one-off piece. The email was extremely specific and very obviously AI slop - humans just can’t write that way. Really comical too. At no point does it seem that a human was involved in writing any portion of the email request. It’s just not possible that a human ever proof read what was written in any of that email (or wrote any sentence) and then decided “hey this looks prefect, time to hit send.”

The best we could come up with is the bot is just scrapping around the internet for information trying to get any response to prove to itself that our website is a valid source of information with which to use for its AI learning?

In the grand scheme of things, MP is also a relatively niche and small community within the internet. 

There is a difference between people using AI to generate responses and generalized bot attacks.

AI generated responses may have a manipulation goal, such as normalizing political talking points or promoting a product.  But they may also just be tools utilized by real people to word their ideas more understandably.  Or to translate them.  

Generalized bot attacks are typically fishing scams related to click-bait.  We have seen many of these attacks on MP, mostly as a barrage of new posts from new accounts, often all saying the same relatively random garblety-gook.  These are extremely high volume scams that cast a huge net over many different platforms.  

If you suspect you are engaged in a conversation with a bot, it’s likely there is a person involved on the other side, even if they are filtering through an LLM.  That doesn’t mean it’s okay.  I’m afraid AI trolling is the future.

Chris, it’s possible you were being vetted by AI learning but it’s also possible that you got a work request from a 13 year old in Vietnam with a poor quality translate program.  

Moving some posts in the related thread I started:

I wrote:

In any case, here’s an AI answer to the question posed:

The bot problem on forums has exploded recently, driven by several factors:

**Who’s making them:**

- **Spammers and scammers** - Traditional spam operations now use LLMs to create more convincing product promotions, scam links, and fake reviews

- **SEO manipulators** - Companies building backlinks to boost search rankings by posting links across forums

- **Political/influence operations** - State actors and activist groups pushing narratives or amplifying certain viewpoints

- **Crypto/NFT promoters** - Heavy automation to shill projects and tokens

- **Regular people with AI tools** - Some individuals just experimenting with ChatGPT/Claude API to automate engagement, not always maliciously

**Why the sudden surge:**

The accessibility of AI language models changed everything. Previously, bots were obvious because they used templates or broken English. Now with GPT-3.5/4, Claude, and similar models, anyone can generate contextually appropriate responses at scale for pennies. A single person can now simulate hundreds of forum participants.

**Their goals:**

- Drive traffic to monetized sites

- Manipulate discussions or sentiment

- Farm accounts to sell later

- Inflate engagement metrics

- Test and train better bots

**Why forums are vulnerable:**

Many forums haven’t updated their defenses beyond basic CAPTCHAs, which AI can now solve. Free forum software makes it easy to target thousands of sites. And smaller communities lack resources for sophisticated bot detection.

The problem’s getting worse because detection is an arms race - as defenses improve, bot operators adapt their AI prompts to seem more human.

Collin H wrote:                   

Do you think this is a real person, who just happens to have dedicated their entire account to a business for English-language learning for Vietnamese speakers, then randomly piped in to give advice about how to avoid a pulley injury?
https://www.mountainproject.com/user/201144461/edu-life

 

I don't think bots have taken over by any means, and I think they are relatively infrequent, but they are getting more common and sometimes a bit better. Most of the non-obvious bots have a few mildly odd posts where it's a bit unclear whether or not they are a bot, so you may not notice them. Then they slip up and make a more obvious bot post that gets flagged, the entire account and all of its posts get removed within hours or a day, and there is no evidence left. Here is another bot post from a since-deleted account (it's unclear to me why the posts sometimes disappear and other times do not):

https://www.mountainproject.com/forum/topic/202321509/getting-to-511a-sport

Here is another that is not yet banned (I was actually trying to find another bot with a similar name but randomly found this one instead. The rhyming nonsense name seemed to be a trend for a while):

https://www.mountainproject.com/forum/topic/201040557/climbing-in-dr#ForumMessage-201221556

There was another that started nearly a dozen threads about random climbing-related topics before it became sufficiently obvious that they were a bot and their account was removed.

I also get the impression you haven't played around with AI much (correct me if I'm wrong). It would be far easier/cheaper to automate this than to pay real people to do this, or to have a hybrid version that is partially automated but with humans to get around obstacles related to sign-ins. Online scams alone are a multi-billion dollar industry, and I'm sure spamming, deceptive advertising, and influence campaigns are at least as big, so there is plenty of economic incentive to do this. To the extent that coding is required, the code and/or services get sold over the dark web, not redeveloped from scratch by an individual. There is also a second-hand market for accounts on social media and Reddit, with older accounts and those with more activity being more valuable, which could provide incentive for bots to lie low and make a random post every now and then without trying to sell anything. MP probably isn't a very high-value target for this because of it's size, but it wouldn't shock me if it was happening sometimes.

"Because it's there"

I should note that the bot post from my third link has since been deleted, and the post it goes to now is not one I have any reason to believe was written by a bot. Here is another bot example from minutes ago to take its place. This one left a hilariously non-sensical comment on a thread about desert towers:

https://www.mountainproject.com/user/202072397/murrem-urphy/community

The examples were a response to bot skeptics who claimed there is no evidence of bots on MP, or that bots only post CBD gummy spam and don't try to participate in threads and pretend to be real people.

Edit:

Clickable links for other less obvious bot examples from the quoted post above:

https://www.mountainproject.com/user/201144461/edu-life

^That one left on-topic responses in threads, wrote like a human, and quoted and responded to other people. If its profile wasn't an advertisement for a Vietnamese English language learning business, I probably wouldn't know it was a bot. Account is 5 years old.

https://www.mountainproject.com/forum/topic/202321509/getting-to-511a-sport

Thread from last year on the exact same thing:

https://www.mountainproject.com/forum/topic/201095781/bots-of-mp

I played around with botting MP for fun.  It is REALLY EASY.  Mountain Projects desktop application doesn't really seem to have much in the way of bot protection.  All they would need to do is add the odd captcha / honey pot, but its likely just not worth it to them atm given there is a (free?) admin community that can pick up the slack.

That being said,, a lot of what you may think are bot accounts could actually be someone trolling the good old fashioned way.  I know of some very motivated youngsters....

Aaron Waitwrote:

I played around with botting MP for fun.  It is REALLY EASY.  Mountain Projects desktop application doesn't really seem to have much in the way of bot protection.  All they would need to do is add the odd captcha / honey pot, but its likely just not worth it to them atm given there is a (free?) admin community that can pick up the slack.

That being said,, a lot of what you may think are bot accounts could actually be someone trolling the good old fashioned way.  I know of some very motivated youngsters....

Us lowly admins don’t maintain the forum, paid people at OnX do.

Did some digging after another bot post today. It looks like most bots edit their posts after some time has passed to add in sketchy links without triggering notifications. I didn't click any of the links because I don't want my computer to catch on fire, but presumably they are spam, scams, or SEO. Try searching the forums for "xender.vip", "100001.onl", "omegle", "vidmate", “speedtest”, "routerlogin", "9apps.ooo", or "192168", and you will find several dozen bot posts, mostly from accounts that have not been suspended.

Contrary to my expectations, it seems like many of these posts weren't actually AI posts, but were copied and pasted from elsewhere on the internet (often reddit). Regardless of how they were made, it's pretty obvious they aren't real climbers, and I'm calling them "bots" for lack of a better term. I realize that posts about bots are nearly as annoying as posts by bots, but figured sharing these could provide some good examples that make it easier to spot them in the future (and to block these accounts).

Edit: If a mod sees this, there are around a hundred bot accounts that need deletion/banning. It might be a good idea to save the account info somewhere (i.e. their emails, IP addresses, and patterns of posting/activity), as this is a good "dataset" that could help to detect this sort of thing in the future.

Collin Hwrote:

...

Edit: If a mod sees this, there are around a hundred bot accounts that need deletion/banning. It might be a good idea to save the account info somewhere (i.e. their emails, IP addresses, and patterns of posting/activity), as this is a good "dataset" that could help to detect this sort of thing in the future.

Thanks, Collin. Spam and bot management is an ongoing battle. As others have noted, it's an endless game of whack-a-mole, and we're doing our best to improve our automated system without disrupting the experience for real climbers. 

Cory Nwrote:

Us lowly admins don’t maintain the forum, paid people at OnX do.

This is possibly the shittiest forum i participate in in terms of features. forums from the late 1990s were more advanced than this one. I CANNOT believe folks are being paid to maintain this slop. time is also ticking, OnX will paywall MP eventually. consider this before you go adding routes to MP. start building/participating in your own local platforms/websites/pdfs/guidebooks. the amount of false info on this site is staggering!

Zander Göpfertwrote:

Thanks, Collin. Spam and bot management is an ongoing battle. As others have noted, it's an endless game of whack-a-mole, and we're doing our best to improve our automated system without disrupting the experience for real climbers. 

Hi Zander, thanks for checking this out! I'm sure it's not easy, and that the people running the bots may adapt to get around solutions. I do have a few ideas that might help for this specific type of bot without adversely affecting the rest of us.

Perhaps if someone edits their post to add a link, that should trigger a notification. That would cause a small increase in notifications, but doesn't seem like it would be too disruptive.

It would also make sense to keep a list of the bot links and perhaps use them for automatic flagging, as some of the links seem to have been in use for several years. The ".onl" domain names seem to be most common, so a search for these might find more. The MP search doesn't support simple Boolean, so I wasn't able to do this, but perhaps there is an easy way to do this from your end (if I can make a tangentially-related dream request, it would be to allow simple boolean searches, as the current way the search function automatically looks for similar words is quite annoying).

I have a larger list of bot links I can PM you. I can also go through and flag the posts if that would help/save time for you guys. The bot posts annoy me so it's kind of satisfying hunting them down haha. I realize that the bot posts are typically from new accounts that only post once or twice, and that none of these ideas would prevent that, but it might remove the incentive if they stop being able to get links that stay up. The group(s) doing this don't seem to be very savvy, so if it becomes a bit more difficult for them, it could stop being worth it.

Really appreciate that, Collin, and thanks for those suggestions. We’ve seen an uptick in spam recently because we’re making some backend updates that should let us catch, remove, and suppress spam in bulk (and we’ve stopped using some of the manual spam tools in the meantime). Once those go live, we’ll monitor and tweak as needed, and continue reviewing flagged posts that slip through. Noted on boolean searches!