Is The Great Gatsby AI-generated? Only 1 of 7 AI detectors said yes
No. The Great Gatsby, published by F. Scott Fitzgerald in 1925, is the most recent of the 30 famous texts we tested. Only TMR flagged its opening as AI, with a 96.8% probability, and TMR flagged 83.3% of the human texts in our main test. The other detectors passed it, Desklib at 0.1% and Fakespot at 2.2%.
Tested
What each AI detector said
| Detector | Verdict | AI probability |
|---|---|---|
| Fast-DetectGPT (Qwen2.5-1.5B) | Human | raw score only |
| Desklib AI text detector | Human | 0.1% |
| Fakespot AI text detector | Human | 2.2% |
| MAGE | Human | raw score only |
| E5-small LoRA detector | Human | 21.2% |
| OpenAI RoBERTa detector | Human | 2.7% |
| TMR (Target Mining RoBERTa) | AI-generated | 96.8% |
The passage we tested
In my younger and more vulnerable years my father gave me some advice that I’ve been turning over in my mind ever since. “Whenever you feel like criticizing anyone,” he told me, “just remember that all the people in this world haven’t had the advantages that you’ve had.” He didn’t say any more, but we’ve always been unusually communicative in...
The verdicts
TMR called the passage AI-generated at 96.8%. Everyone else passed it: Desklib at 0.1%, Fakespot at 2.2%, the OpenAI RoBERTa detector at 2.7%, the E5-small LoRA detector at 21.2%, and neither MAGE nor Fast-DetectGPT flagged it. TMR flagged 28 of the 30 famous texts we ran and every human text with more than one paragraph in our main test, so a flag from it on a four-paragraph excerpt is expected and says nothing about Fitzgerald.
Why the newest classic passed almost every detector
Gatsby’s prose is polished and literary, the kind of writing that sometimes trips a classifier, as Fakespot showed with the opening of Jane Eyre. What sets this passage apart is Nick Carraway’s voice. He makes small, specific judgments that a writer would only make on purpose: he was “the victim of not a few veteran bores”, he has “feigned sleep, preoccupation, or a hostile levity”, and his father “snobbishly suggested, and I snobbishly repeat” a lesson about “the fundamental decencies”. Phrases like “unusually communicative in a reserved way” turn on a contradiction, and conduct “may be founded on the hard rock or the wet marshes”.
These are unexpected word pairings, unlike the most likely next word that a language model tends to choose. That probably explains why the other classifiers saw nothing machine-like, and why Fast-DetectGPT passed the passage despite its famous first line, “In my younger and more vulnerable years”. One well-known sentence is likely not enough to make a whole passage predictable.
Distinctive voices and conventional essays
Gatsby suggests that a distinctive voice with surprising word choices is hard for these detectors to mistake for AI. That is no defense for a student whose clear, conventional essay gets flagged, and conventional writing is what most people are asked to produce. A flag from TMR alone, as here, should be ignored. A Scandal in Bohemia and Frankenstein were also flagged by TMR and nothing else. The famous texts overview has every result, and the false positives page shows how detectors treat ordinary human writing.
Questions people ask
Do AI detectors flag The Great Gatsby?
Only one of the seven we tested did. TMR gave the opening a 96.8% AI probability. Desklib, Fakespot, E5-small LoRA, the OpenAI RoBERTa detector, MAGE and Fast-DetectGPT all called it human.
Is modern-sounding writing more likely to be flagged as AI?
Not by itself. The Gatsby opening is the most recent text in our famous-text set and reads much like present-day English, yet only TMR flagged it, most likely because of its paragraph breaks.