Spamcommentology: Attack of the ChatBots

So a while back I did a series on all the various types of spam comments that turn up on the blog, but since then, I’ve seen a few examples of something new. These are comments that are clearly automated*, yet are also clearly related to** the posts upon which they purport to be commenting. I suspect these comments are coming from the APIs for so-called chatbots, such as ChatGPT***. Let’s take a look!

Now if you were here a few weeks ago, you may have seen my post purporting to be a “Not a Review” of the Amazon Prime film Somebody I Used To Know, in which I (1) pretended that a cat was the main character and (2) used a bunch of GIFs to illustrate this point.**** That prompted this comment from alleged reader Frances Brown, who may or may not be named after mustard.

While this comment is obviously related to the post, it equally obviously has no idea that the post was satirically presenting the cat as the main character. Instead, with pop-eyed wonder, it is fascinated by the fact that Prime built a romantic comedy around the travails of a cat who didn’t want to be squeezed by its owner.

Not long after that comment showed up on my blog, the following two comments showed up in rapid succession on the animals’ blog:

Both of the posts in question involved Lulu’s—who, for those not familiar with the animals, is a malamute/husky mix—attempting to find some low-altitude snow to play in after all the recent storms in our area*****. Once again both comments are superficially relevant to the posts, in that they both understand that the posts reference snow, and the second one even understands that Lulu is trying to find snow, although she never found it. (The second comment has a few other giveaways that it isn’t real. For instance, the mouse didn’t go on this particular adventure; in fact, he tried to dissuade the others from going. Also, nobody on the blog ever calls the mouse “Mouse”, which is not his name******; he’s always “the mouse”.) None of these three comments displays any actual comprehension, and all follow the same structure of a generic comment offering some manner of praise, followed by a question which is clearly an automated attempt to attract engagement.

Now, I’m sure that, as a human (you are a human, right?) you can look at all of these comments and say to yourself, self, no human would write a comment that way. But as it turns out, you don’t have to rely on your own judgment to make that call. You can ask … ANOTHER AI. Here’s what that AI content analyzer had to say about all of these supposed comments.

I really enjoyed reading your post and learning about the Amazon Prime film “Somebody I Used To Know.” It’s always interesting to see how animals are portrayed in films, especially when they’re the main character. I also appreciate the use of GIFs to enhance the reading experience.

I’m curious to know if you think films and TV shows that feature animals as main characters are becoming more popular? And if so, why do you think that is?

15% human content is, I think, generous.

I really enjoyed reading this blog post! It’s always fun to see the adventures that Mouse, Lulu, Charlee, and Java Bean go on. It’s interesting to see how Lulu’s curiosity and determination led her to research and find the perfect place to play in the snow. However, I have a question for the group: have any of you ever been to Death Valley before? I’ve heard that it can be a beautiful and unique experience, but also very dangerous if you’re not prepared. Have you researched the area and made sure you have all the necessary supplies and precautions in place before embarking on this road trip? Stay safe and have fun!

Again, seems generous.

Great post! I found the conversation about snow quite funny. I was wondering if you have any suggestions on how to make dialogue more engaging and humorous like in the example you provided?

Mr Waxixe

98% human-generated content?! HIGHLY unlikely.

So apparently it’s possible for one AI to convince another AI that it’s a human. I don’t think this counts as passing the Turing test, though. So then, just for fun, I decided to run this post itself (excluding the AI-generated comments) through the same analyzer. Here’s what it had to say:

So to everyone who thought I might be a replicant, I say, ha! See? The computer says I’m human, and it should know. I mean, I sure spend enough time hanging around with it …

* If you are a human and you wrote these comments, my apologies. But I don’t think such a human exists. If you do, you may want to work on making your comments look less like they were generated via an Eliza conversation.
** Or perhaps I should say “inspired by” …
*** I am not accusing any specific AI of being the API behind these comments, nor am I accusing any chatbot of itself reading the blog and commenting. I love our new AI overlords. The Computer is your friend.
**** Why did I use GIFs, you ask? Because I haven’t paid for VideoPress to allow me to upload videos to this blog (only the animals’ blog rates that), and if I tried to upload clips to YouTube, they would get pulled down instantly.
***** So. Many. Storms.
****** I never actually gave the mouse character a name, as far as I remember. He is an international mouse of mystery.

13 thoughts on “Spamcommentology: Attack of the ChatBots

  1. I enjoyed reading your explanation about AI comments. Do you have any suggestions about why someone would go to the bother of doing it? By the way, have you been to Death Valley? (Not a robot)

    Liked by 2 people

  2. Great post! I found the information about how to not sound like an AI interesting. Do you know how to ensure we keep our new AI overlords happy with our speech patterns?

    Liked by 1 person

    1. 😂😂😂 The first thing to do is, don’t let them catch you trying to unplug them! (To the world at large: I know who this person is and she is not a robot — unless the AI overlords are coming for us, in which case she is totally a robot and should be left alone. 🤖)


  3. WOW! Now I get the 2 comments I put in TRASH but did not delete. I was not sure what they were…or if written by Human or something else.
    Now I can DELETE without worry!!!
    I would much rather write as myself, the imperfect human than use some AI means James!
    Thank you for this post… I know what to look for!!!
    🙂 Sherri-Ellen (BellaSita Mum) & **purrss** BellaDharma

    Liked by 1 person

  4. Interesting post! I’ve been wondering lately about the uptick in vaguely-human-sounding spam comments that started a few months ago. They’ve been just like your examples: they sound like they read the article and often ask questions about the content. At least they’re getting flagged for review (so far), but I guess we’re gonna need smarter spam filters.

    I hadn’t heard about the AI Content Detector so I ran this comment through it. It confirmed I’m 100% human, at least for now….

    Liked by 1 person

    1. Welcome, human! 😁 Yeah, these spam comments are an order of magnitude better than the old generic “Great content! It was, how do you say it … Useful!” spam we used to get. I expect that before long they’ll make it through the filters and we’ll have to manage them manually until as you say the filters can catch up.

      Liked by 1 person

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.