Is A Tale of Two Cities AI-generated? Only 1 of 7 AI detectors said yes
No, and nearly every detector we tested agreed. Charles Dickens published A Tale of Two Cities in 1859, and 6 of 7 AI detectors called its opening human, including Desklib (0.2%), Fakespot (0.3%), the OpenAI RoBERTa detector (2.6%) and E5-small LoRA (15.1%). Only TMR flagged it as AI-generated, at 80.7%.
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 | 0.2% |
| Fakespot AI text detector | Human | 0.3% |
| MAGE | Human | raw score only |
| E5-small LoRA detector | Human | 15.1% |
| OpenAI RoBERTa detector | Human | 2.6% |
| TMR (Target Mining RoBERTa) | AI-generated | 80.7% |
The passage we tested
It was the best of times, it was the worst of times, it was the age of wisdom, it was the age of foolishness, it was the epoch of belief, it was the epoch of incredulity, it was the season of Light, it was the season of Darkness, it was the spring of hope, it was the winter of despair,...
Six detectors passed it
None of the 30 famous texts we tested passed all seven detectors, and this one came as close as any. Desklib, Fakespot and the OpenAI RoBERTa detector all scored it at 2.6% or lower. E5-small LoRA, a detector that flagged 50.0% of human texts in our main benchmark, gave it 15.1%. MAGE and Fast-DetectGPT, the two detectors that report raw scores, both called it human.
Fakespot’s 0.3% is low for a detector that gave Moby-Dick 90.7% and A Modest Proposal 98.7%. It has no trouble calling old books AI, and it still read this one as human.
What the detectors made of the repetition
“It was the best of times, it was the worst of times” runs through fourteen matched clauses before it breaks off. People who look for AI writing by eye often treat that kind of parallel structure as a warning sign, and in a modern essay it might well draw suspicion. The detectors did not treat it that way.
A likely reason is what comes after the famous sentence. Within a few lines Dickens moves to “a king with a large jaw and a queen with a plain face”, to Mrs. Southcott’s “five-and-twentieth blessed birthday”, the Cock-lane ghost and “the chickens of the Cock-lane brood”. The irony is dry and the details belong to the England of 1775. A language model would struggle to predict that, which matters to Fast-DetectGPT, and it is probably unlike most of the AI text the classifiers were trained on.
Fame alone was not enough to make Fast-DetectGPT flag it. That detector flagged Hamlet’s soliloquy, which is just as well known, but passed this. Familiarity may make a text more predictable to a model, and here it did not push the score over the threshold.
Why the one flag means little
TMR’s 80.7% is the only AI verdict, and it comes from the detector with the highest false positive rate in our test, 83.3%. The passage has three paragraphs, and TMR flagged every human text with more than one paragraph in our benchmark. The paragraph break guide shows the effect.
Set this beside Moby-Dick, whose odd, digressive prose drew three flags. Prose a reader might find artificial passed here, and prose that sounds unmistakably human was flagged there. When you judge real writing, neither your own impression nor a detector’s verdict settles who wrote it. Walden got the same single flag from TMR. Every result is on the famous texts page.
Questions people ask
Did any detector call A Tale of Two Cities AI-generated?
One did. TMR flagged it, and TMR flagged 83.3% of all human texts in our main benchmark. Dickens published the novel in 1859.
Doesn't the repetition in "It was the best of times" look like AI?
To a human reader it might, since repeated parallel phrasing is a habit people associate with chatbots. The detectors did not react to it. Fakespot gave it 0.3% and Desklib 0.2%.
Why did TMR flag it?
Most likely because the passage has more than one paragraph. TMR flagged every multi-paragraph human text in our main test, and with the line breaks removed it flagged 25.7% of them.