Literature & Language Arts: Verification drill

What you'll be able to do: Catch the quotation that's close but wrong, the plot detail that's confidently misremembered, and the critic attribution that was never checked, before any of them reaches submitted work.

The subject's single largest specific risk

Ask AI to quote the exact passage where something happens in a book it has "read", and you'll usually get something that sounds right, the diction is in period, the rhythm fits the author's style, and it can be word-for-word wrong, Not garbled , plausible. A line from the right chapter attributed to the wrong character. A paraphrase presented with quotation marks as though it were exact, A detail from the film adaptation folded into a description of the novel. A critical position confidently attributed to a named scholar who never held it.

None of this is a glitch you'll notice from the output alone. It reads exactly like a correct quotation would read, because it's built from the same statistical impression of "what this kind of sentence sounds like" that produces the correct ones. The tool has no internal flag that fires differently for "I am certain" versus "I am extrapolating from style".

This is why the subject's working rule (article 00) is as strict as it is: every quotation and every attributed critical position that ends up in your work needs to have been checked, by you, against a source you actually have in front of you. Not because AI is unusually careless here, but because in this subject the exact wording is the evidence, and evidence that turns out to be invented isn't a weak point in the essay, it's disqualifying.

Where the error rate is worse, and why

It gets worse, predictably, the further you move from extremely famous material. "Call me Ishmael" is so densely represented in training data that getting it wrong would be startling. A paragraph from Moby-Dick's chapter on cetology is a different matter, far less quoted, far more room for the model to produce something stylistically indistinguishable from the real text that isn't the real text. The same asymmetry holds for plot detail: the identity of the character who says a famous line is well known ("the lady doth protest too much, methinks" is Gertrude's, commenting on the Player Queen in Hamlet, not Hamlet's own line, a genuinely common misattribution); which minor character says which line in Act 4 of a play you're not studying at exam level is exactly where confident, wrong answers cluster.

Treat familiarity as a spectrum, not a binary. The more obscure the passage, the more sceptical the check needs to be, and famous passages aren't exempt, just lower-risk.

How to run it

1. Ask for the exact quotation and its location , act, scene, chapter, or line reference.
2. Locate it yourself in a copy of the text before using it anywhere.
3. Check word for word, not just gist. A single altered word can change what the quotation is evidence for.
4. Do the same for any named critic or school of thought , "critics have read this as.." needs an actual source before it's an attribution rather than a plausible-sounding guess.
5. Log what you find. Most students are surprised how often drift shows up on anything past the most famous passages, and keeping a running tally makes the risk concrete rather than abstract.

Quote me the exact passage from [text] where [event] happens, and give me the act/scene, chapter, or line reference. If you're not fully certain of the exact wording, tell me that directly rather than giving me your best reconstruction dressed as a quotation. I will check this against my own copy before using it anywhere.

Across the disciplines

Literature. As above, quotations, plot details, which character says what.

Composition and rhetoric. Statistics and historical claims used as supporting evidence in an argumentative essay carry the same risk as a literary quotation, a plausible number attached to a real-sounding source, unverified.

Creative writing. Research folded into fiction or non-fiction for authenticity (a historical detail, a technical fact) needs the same check, readers who know the domain will find the wrong detail, and it undermines everything else in the piece the way a fabricated quotation undermines an essay.

Linguistics. Claims about "how English used to work" or "what this word originally meant" are exactly as prone to confident invention as a literary quotation, etymology in particular tends to produce folk explanations that sound authoritative and aren't. Check against an actual dictionary of etymology, not the tool's paraphrase of one.

Film and adaptation. Confirm which version you're actually being told about. Plot and dialogue details drift between a novel and its film adaptation constantly, and the tool will blend them without flagging that it's done so.

Pitfalls

1. Verifying the quotation but not the attribution. A real quotation attributed to the wrong speaker is still wrong, and still embarrassing.
2. Checking once and reusing. Each new quotation needs its own check, a clean result on one doesn't vouch for the next.
3. Accepting "I'm not entirely sure, but roughly.." as sufficient. That's the tool doing the honest thing; the honest thing to do with it is not use the quotation, not use a softened version of it.
4. The tell: you used a quotation, a plot detail, or a critical attribution in submitted work that you did not personally locate in a source. That's the subject's single bright line.

Try this today

Ask for a quotation from something you're currently studying, with its exact location. Find it yourself and check it word for word before you decide whether it's usable.