AI-generated reviews and recommendations for niche and prestige films are besieging Facebook, the latest in the canon of widely beloved slop hits like “Shrimp Jesus” and deepfaked dead people.
These accounts, usually masquerading as movie and pop culture enthusiast pages, are filled with dozens of film recommendations every day, all with the hopes of bubbling up to Facebook timelines and getting that sweet, sweet engagement. Most, if not all, of the posts focus on prestige-y films: foreign documentaries, critically acclaimed cult classics, and historically significant works are well-represented. It almost feels like someone dumped every movie with a Rotten Tomatoes rating higher than 80% into ChatGPT and asked it to produce crappy mini-essays on each movie.
I sent a handful of blurbs from three different accounts to Max Spero, the co-founder and CEO of AI detection firm Pangram Labs, with a hunch that they were produced by generative AI. All of them were most likely written by a chatbot, he confirmed.
“People posting full AI writeups and nobody else seems to notice or call them out on it,” Spero wrote over email.
Once you notice it, it’s really inescapable. These write-ups almost always have three to five paragraphs. They almost always start with a topic sentence that looks cribbed from Wikipedia. A good chunk of the time, they’re factually accurate, if very blandly written — unlike the best film criticism, there is no zeal or life here. Reading an AI-generated writeup about a movie is like wearing a poncho in the rain: It gets the job done, but you’re not really happy about it.
That in itself is well and truly upsetting. But, even worse is that, like plenty of other AI-generated text, it tends to be wrong, often egregiously so.
In one instance, the account Film Fanatics (20,000 followers) wrote about a 1943 movie called Angel of Sin. They proceed to fabricate an entire movie: From the director to the actors to the plot, the movie they describe simply does not exist. The screenshots that accompany the post are from a real movie that came out that year: A French film called Angels of Sin (note the plural). It’s a glaringly bad post, and it has a mere 17 likes.
But for a more instructive look into how grim these errors are, take this Film Fanatics write-up from earlier this month by. Of the prosaic, quietly melancholy Perfect Days — highly recommend if you haven’t seen it — the account writes:
“Perfect Days (2023) is a Japanese drama film directed by Hirokazu Kore-eda. The movie follows the life of Kaito (played by Koji Yakusho), a quiet and seemingly content man who works as a toilet cleaner for a company that offers a specialized service in cleaning public restrooms.”
Two things are wrong: It was directed by German auteur Wim Wenders, not Kore-eda; the main character’s name is Hirayama, not Kaito.
Some commenters rightly pointed both errors out. “How can you get the director wrong…????? and even talk about his signature style…????” one wrote. Many other commenters raved about the film, all without acknowledging the factual errors. It has more than 3,000 reactions and hundreds of comments. None of the top comments pointed out the possibility that this could be an AI fabrication.
These Facebook users will earnestly engage with these accounts — not knowing that instead of a person behind the screen, there is simply a machine cribbing words from better, smarter critics and winnowing them down into Engagement Bait. (Another seemingly AI-generated account, attributed to “Giuliana Hernandez,” also wrote about Perfect Days and also went viral — but did not have any glaring inaccuracies, which makes this fiasco even stranger.)
Film Fanatics did not respond to multiple DMs from Gazetteer SF asking if their content is AI generated.
“I would like to see more genuine content creators getting rewarded for putting out high quality stuff, but now it seems to be even harder to compete with the deluge of low-effort AI slop that people are putting out,” Spero told Gazetteer.
This all amounts to a strange, shitty ecosystem — fake content, real engagement. Meta head honcho Mark Zuckerberg wants more of this. Last October, in an investor call, the Zuck promised more “feed experiences” with even more AI-generated dreck coming to Meta's various social media properties. And despite users vociferously voicing their opposition to other Meta experiments into AI, people still continue to engage with posts like this made by enterprising hucksters looking to make side hustle money.
“It's insane to me how many comments and reactions these slop posts are getting, but I think this says the most about our media literacy and algorithmic diet,” Spero said.
Generative AI will rightfully get the most heat for disseminating misinformation, or for spreading illicit deepfakes, or for murdering multiple creative industries in cold blood. But this is the stuff that really underscores Meta’s bottom line — all the more reason to disengage the minute you notice any post that feels off, whether that’s film reviews or anything else you care about.
