he Internet is full of articles giving you advice about how
to spot AI-generated text. A chatbot, they say, is always bland
and unimaginative. Often the sources and links that it uses are
incorrect, the facts are wrong, and the sentence structure is dull
and repetitive. Chatbots over-use passive voice and have a limited
vocabulary.
Other articles warn that chatbots could someday turn the Internet into a pile of regurgitated, useless shit.
Okay, stop laughing, we're being serious here.
Here at randombio dot com, we've been studying this phenomenon for over twenty minutes and we have discovered some infallible ways of telling if something is fake.
Chatbots often spell words correctly and use correct punctuation, and even, on occasion, use proper English grammar—something that obviously no human on the Internet can ever do.
Unless it's been hacked, a chatbot will never say anything politically incorrect. For example, here's something a chatbot is unlikely ever to say: “A chatbot is not going to cite another chatbot. Chatbots are like women: they all hate each other.”
The easiest trick, which most other sites recommend, is to post the suspect text to ChatGPT and ask it if it's AI-generated. If it is, the chatbot will likely say, “Hey! That's one of mine! You bastard, you're plagiarizing my stuff!” It's a sign of the end times that people actually would think that one chatbot is going to rat out another chatbot—or that it could know whether something has been plagiarized, or tell you if it did.
Unlike a real person, a chatbot usually won't pop up an annoying paywall halfway through the article demanding money to read the rest.
A dead giveaway is the content. Chatbots can't think. They have no imagination and they steal all their content from Wikipedia. So the result will end up sounding like a mishmash of politically-motivated falsehoods, incorrect attributions, and fake neutral tone to trick the reader into hating President Trump.
Chatbots also have no sense of humor. Unlike humans, chatbots don't sit in front of their computer writing stupid articles day after miserable day screaming into the howling void of the Internet in a desperate bid for attention. So if the writing contains a joke, includes a sexual innuendo, or sounds as if the writer has a mouth full of Cheez Puffs, there's a good chance a human wrote it. Chatbots dislike sexual innuendos and they never eat Cheez Puffs—only pretzels.
Some of the claims, like the idea that AI-generated text never puts spaces around an em-dash, are clearly false. Em-dashes, which people sometimes use instead of commas to insert a totally irrelevant idea into a sentence, are a big problem on the Internet. But the problem isn't the em-dash per se. It's that HTML is total shit at kerning—even worse than MS-Word—and it suffers from the same problem that parentheses have: you can't nest them.
First they came for our semicolons; then they—and we all know **who** they are—came for the em-dash. And don't get me started on en-dashes, which chatbots always substitute with hyphens (for example: “pages 25-50”, which looks like a screw size instead of a range). Oh, and minus signs. Anyone who uses a hyphen instead of a minus sign deserves to be driven out of polite society and forced to read a chemistry book that uses them, instead of something like this: H2O ⇆ H+ + OH-.
A bigger problem is HTML's inability to handle math. Just try, I dare
you, to get a summation sign to come out the correct size. A Σ is
too small. A ∑ doesn't let you put indices under it. Writing a
fraction in HTML is a nightmare. We had to create a whole new span
class just for the case when there's a square root in the denominator.
And it still looks weird unless we put it in a <code> tag.
Thanks to those ubiquitous sans-serif fonts,
it's so bad we even have to put AI in a
code tag because otherwise AI (the language models), Al (the guy),
and A1 (the sauce)
look identical. And for those of you who flunked chemistry, it's CO2.
There is no such thing as C02!
And another thing. We promise never to use that abomination, beloved of writers in the UK, of **emphasizing** a word with asterisks. It looks too much to us like a swear word, as in the sentence “First they came for the em-dash; now they're coming for the ******* semicolon; ******* *** ** * *****!”
We refuse to put spaces around em-dashes. The general rule we follow here is: if the AP style guide says to do something, we do the opposite.
In fact, the answer to the question is simple: there is no such thing as AI-generated text because there's no such thing as AI. There are only chatbots. And consider: if Elon Musk is right and the world is merely a computer simulation, then everything and everyone is a chatbot. In which case, no matter where you get it, the answer is yes, it comes from a chatbot, and the chatbot is you.
jun 14 2025, 5:44 am
