The AI Fermi Paradox: If "humans can write like this", then why haven't they?

08 Jul, 2026

Whenever I hear someone argue that it's impossible to detect AI-generated writing because "AI is just copying human writing" -- a disproven myth, by the way -- or "humans write like this all the time", I often think of the Fermi paradox. The Fermi paradox, named for physicist Enrico Fermi, is an observation about aliens: the universe is unfathomably vast, and has existed for an equally unfathomable amount of time, far beyond the conception of puny human lifespans. Yet despite this unimaginable amount of time and space for lifeforms to evolve, no one has ever found credible evidence of intelligent extraterrestrial life. Why? The blog Wait But Why did some back-of-the-envelope math on the odds of this:

When confronted with the topic of stars and galaxies, a question that tantalizes most humans is, “Is there other intelligent life out there?” Let’s put some numbers to it. [math omitted for brevity, but you can follow along on the site] We’d estimate that there are 1 billion Earth-like planets and 100,000 intelligent civilizations in our galaxy. If we’re right that there are 100,000 or more intelligent civilizations in our galaxy, and even a fraction of them are sending out radio waves or laser beams or other modes of attempting to contact others, shouldn’t SETI’s satellite dish array pick up all kinds of signals? But it hasn’t. Not one. Ever.

As Fermi (supposedly) remarked: "Where is everybody?"

And so it goes with LLMs. AI-generated writing has identifiable and idiosyncratic phrases that it will apply to virtually any topic. And you have to -- absolutely must -- think of these as specific phrases: verbatim strings of words. These words are often banal, and they often resemble the cliches of human writing readers have encountered thousands of times. There is nothing preventing a person from writing them. But if you look at the things that people have actually written over the years, in writing both good and phoned-in, these strings of words are rare. Shockingly rare. Like, single digits.

Variations of this can be seen in AI-generated writing on basically any subject matter; every topic has its little idiosyncracies. For this post, we'll focus on writing about entertainment.

AI-generated text about music, seemingly wherever possible, will say that an album or song “blends” or is “blending” multiple things, usually multiple genres. Often it will specifically blend "elements" of things, e.g., "The album blends [elements of] R&B and pop." At the time of writing, the phrase "The album blends" appears 164 times on Wikipedia, and in most cases, it appeared after 2022. Most of these are by different editors, in articles on albums from different genres.

This sounds unremarkable, right? It's just a normal cliche. Maybe you remember reading it in a review somewhere, or a YouTube video. It certainly sounds like something a music writer would write, or at least could write.

But as it turns out...

In the entire history of Pitchfork, spanning over two decades, dozens of editors, hundreds of writers, and an evolution from amateur blog about indie bands to Conde Nast property covering pop music to subsidiary of fashion magazine GQ, the phrase "the album blends" has appeared a total of five times. Of those five times:

• Two of those five times are the same content pasted into different articles, so it's really more like four times.
• One of them, a 2016 Kid Cudi review, uses the phrase completely differently, where "the album blends together" means "it's all a blur."
• All of them have a specificity -- e.g., "the album blends her sugar-rush raps with a mosher's fury for maximum thrash" -- that is usually absent from the more generic AI writing. The sentences are more complicated. The wording is more colorful.

Maybe this is just a coincidence, though. Maybe Pitchfork has unusually idiosyncratic writing, or some weird longstanding ban on that phrase. How about other publications?

• NME has only used it once, and it's part of a direct quotation of a sentence structured differently.
• Stereogum has only used it once, and it's another long, specific comment: "the album blends those impulses with much of the spectral pop music that tends to dominate festival stages outside the DJ tent."
• Spin has only used it twice. Both articles were published after 2022 (the release year of ChatGPT).
• Billboard has used it a little more often, nine times. About half of these are after 2022.
• Rolling Stone is the only publication so far with a whole two pages of Google results.. The majority are after 2022, and many of them are specifically from the subsidiary site Rolling Stone Australia, in articles that are often dubious sponcon. This review, for instance, turns the word "synths" into an ad for a "24/7 AI-assisted SOC" enterprise software company.
• Allmusic seems like the jackpot so far: three pages of Google results! But yet again, these usages either have specificity ("The album blends the headbanger-worthy riffs and hooks of stoner metal with the spacey nostalgia of psychedelic prog rock and the loud, raw energy of alternative rock.") or were published after 2023, often in unbylined reviews or low-quality clickbait. Sorry, but this is AI slop.

Similar results can be found for variations like "the song blends" (125 Wikipedia results, 2 Pitchfork results, 1 NME result, 1 Spin result, two Stereogum results, 5 Allmusic results, 20 Billboard results often after 2023, and 18 Rolling Stone results, mostly Australian sponcon). As someone who has read more AI-generated writing than I ever wanted to, and has seen the same shit over and over, I could have chosen dozens of other phrases.

So if anyone could write these phrases, why have so few people actually written them?¹ And why, given the decades-long existences of these publications, do so many of these shockingly few hits come from questionable articles published in the past two or three years?

Another common phrase in AI articles about entertainment is "explores themes of". This can be seen in any medium, but let's choose movies this time. The phrase "the film explores themes of" returns 187 hits on Wikipedia, usually from text added after 2022. Again, that seems normal. Films obviously explore themes, right? Why wouldn't someone talk about themes? This time, instead of professional film critics, let's look at user-submitted reviews. Wikipedia is written by amateurs, after all. Maybe AI isn't copying professional writers, but the vernacular of the average human normie.²

For this, let's go right to the motherlode, RottenTomatoes.com. The phrase returns roughly 32 hits, which fall into two distinct buckets: the canned "synopsis" Rotten Tomatoes includes on their landing pages, and user reviews that seem specifically curated to prove the dead internet theory. For example, on the page for the 2011 Disney film Lemonade Mouth, the phrase appears in a review by the user "Timmaryturner". The full review is unusually stilted and redundant, and trails off in an odd way:

Lemonade Mouth (2011) is a Disney Channel Original Movie about five high school students who meet in detention and form a band to stand up for their beliefs and overcome personal struggles, becoming a voice for their peers against the school's elite. Directed by Patricia Riggen, it stars Bridgit Mendler, Adam Hicks, Hayley Kiyoko, Naomi Scott, and Blake Michael, and is based on the novel by Mark Peter Hughes. The film explores themes of friendship, self-expression, and empowerment through music, with a soundtrack that gained popularity. Key aspects Plot: Five disparate students—Olivia, Stella, Wen, Mo, and Charlie—form a band called Lemonade Mouth after meeting in detention, using music to challenge the school's status quo and their own personal issues. Cast: The main cast includes Bridgit Mendler (Olivia), Adam Hicks (Wen), Hayley Kiyoko (Stella), Naomi Scott (Mo), and Blake Michael (Charlie). Themes: The movie focuses on themes like standing up for what you believe in, overcoming

But maybe this is just Timmaryturner's natural writing style. His profile is certainly full of reviews just like these. It's also full of stuff like this:

"Here is a summary of the plots for both Ben 10: Alien Swarm and A Fairly Odd Movie: Grow Up, Timmy Turner!"

"Based on the image provided, you're looking at characters from the animated series The Fairly OddParents [...]"

"If you are looking for more details about this movie, I can provide:The most famous quotes from this sceneInformation on where you can stream the movie right nowDetails about the recent musical remakeLet me know what you would like to explore next!"

If you look at other reviews, you'll find similar results:

• The review for Black Mirror: White Christmas is by user "matthias S". I won't comment on the writing style, although if you know you know. I will comment, though, on the less ambiguous AI indicators, like stilted, only occasionally applied raw-markdown headers like **Review – Peaky Blinders: Mortal Man** -- some of which leave out the first few asterisks, as if they screwed up a copy-paste -- and this kicker: "Want me to punch it up with a snappier “closing line” for extra bite?"
• "Hadi K", the reviewer of Blood for Dust, has pumped out many similar canned reviews, all of which sound quite similar to the oeuvres of matthias S and Timmaryturner, and -- ah fuck: "ChatGPT said: Flow (2024), directed by Gints Zilbalodis, is a dialogue-free animated journey following a black cat[...]"
• The reviewer of "Sleep Stalker" doesn't have any obvious ChatGPT comments, but I am also expected to believe that the writer of "We explore what it means to thrive in a world that seeks to exploit the vulnerability of women—her evolution from object to subject is pivotal, if problematic" also wrote this: "It's ironic and disturbing to know Immigrants that are attempting to move here legally are put through this interrogation. What about the 8 million illegals that have come over the border already?" Quite the progressive awakening, and evolution writing style, for just one year!

Dig into the profiles of pretty much any of these reviewers and you will find elements that are at least sus -- writing styles that change dramatically after a certain date, stylistic or factual inconsistencies, reviews that suddenly cut off or include copy-pasted quotation marks -- and often things that are unambiguous, like the above "ChatGPT said". Again, we're searching for a simple five-word phrase: "the film explores themes of". Nothing about it is flowery or idiosyncratic; it could be applied to any film. It's easy to imagine the average non-AI writer putting such a normal phrase onto the page. So where are they?

This is why, whenever someone on social media glibly claims that humans have been writing like AI for years³, or that they personally write like AI and refuse to change⁴ -- I often feel gaslit. I don't like throwing that phrase around lightly, but I've been reading human writing both in print and online for the majority of my life now. I know what I've read, and I know what I haven't. And despite the enshittification of search engines, they do still allow you to search large portions of the Internet, providing strong evidence that it wasn't even possible to read this stuff, because for the most part it did not exist.

These aren't even AI boosters either. These are people who are nominally anti-AI, and yet they are actively carrying water for AI companies by suggesting that the products they release are so good and so effective that they are undetectable. There's a frustrating petulance to this commentary that reminds me of online conservatives who whine about how the libs are somehow "taking words away from them", except in this case the commenters were never even saying these things, and it's unclear why they'd want to. The semantic hills they are choosing to die on are the stuff of deeply sublime humanity, like "Independent reviews list availability in twelve states".⁵ It really, to borrow the voice of ChatGPT, highlights the crucial role of actually knowing what you're talking about.