Lately I’ve been getting comments on my posts that seem to be generated by AI. One made it past Akismet, and I had to reject it manually. It’s worth looking at where they are and where they’re headed. Here’s a comment which I just rejected for one of my recent posts:

Garys abolitionism tour in Boston sounds truly inspiring! The way he describes the historic sites like Faneuil Hall and Park Street Church really brings the fight against slavery to life. Its amazing to see how these places played such a crucial role in the movement. However, I was a bit disappointed that he couldnt visit Harriet Tubman Park. The dedication to exploring these historical sites is admirable, though I can understand why he had to prioritize. The detailed descriptions make me want to visit these places myself and learn more about this important chapter in history. Garys passion for the subject is clear, and its a reminder of how far weve come but also how vital it is to remember this past.

I don’t mind including it here, because the whole purpose of spam comments is to publish a link to a website, and I’ve left that out. The author’s supposed email address was obviously bogus and the link pointed to a Chinese business, confirming my suspicions. Let’s look at the content and what makes it stand out as not by a human.

The most obvious thing is the omission of all apostrophes. That might be sloppy coding, but perhaps the coder had heard that the wrong kind of apostrophe (ASCII vs. typographical) is a giveaway and decided it was safer to omit them altogether.

The second clue is the third-person writing. People commenting on blogs usually address the author rather than talking about them.

The flattery is another indicator. I didn’t say anything descriptive about Faneuil Hall and Park Street Church, let alone “detailed descriptions.” Human commenters may say excessively nice things, of course, but the combination of the pseudo-objective third person and the unwarranted praise comes across as weird.

The whole comment feels impersonal. If it isn’t AI-generated, it’s by someone paid to write comments quickly. People comment on blog posts because something grabs their interest, positively or negatively. There’s no sense of personal involvement; it’s more like an essay written as a school assignment.

The traditional spam comment is generic, saying something like “Great insight!” AI-generated comments will have a better chance of passing filters. As a result, bloggers will become more wary and will reject some authentic comments.

Here’s an article discussing AI-generated spam comments, which seem to be an especially big problem on LinkedIn. Considering that LinkedIn is constantly pressuring its users to let its AI write posts for them, it seems only fair.

I hope no spam developers use this article for ideas on improving their tools.

Update: Two more comments in my spam folder this morning, more or less similar to the one I quoted. One is partly in Chinese.