How we write and check our NCERT content

Finds any run of twelve or more consecutive words taken verbatim from the textbook. Twelve words is the threshold because a shorter one flags ordinary phrasing and a longer one misses real copying.

Scores each passage as a percentage of its own words — which catches a copied definition, where the copy is most of the text.

Reports near-matches that fall under the verbatim threshold but still track the book too closely.

Compares every chapter against every other, so the same question cannot quietly appear in two places.

Flags questions where more than one option is actually correct — including numerically. One Maths question keyed 1/6 while offering 5/30, which is the same number.

Flags a key that the chapter's own explanation contradicts.

Re-solves the arithmetic in Maths question stems rather than trusting it.

Re-checks stated mathematical facts, such as prime factorisations and HCFs.

Caps sentence length in model answers, so a written answer is one a student could actually give.

Balances which position holds the answer, so the key is never guessable from where it sits.

Catches questions where the correct option is the longest — "pick the long one" is a strategy, not learning.

Catches the mirror of that: a bare correct option against three padded wrong ones.

Balances assertion–reason questions, which have their own answer pattern.

Also catches an option whose surface — the only capital letter, the only question mark — gives it away.

Applies a readability floor to the Class 5 and 6 material specifically.

Checks the structure of every chapter, with a stricter mode for a finished track.

Refuses a chapter that is too thin on any one surface to be worth opening.

Catches sentences broken by a generator rather than by a writer.

Catches a line left stranded by itself.

Catches gaps in a table or a list.

Checks every illustration renders at the size the page gives it, on a phone.

Checks the mathematical notation is valid.

Renders every formula the way a browser will, and fails on the ones that would not.

Runs each game block against the code that plays it, so a game cannot ship unplayable.

Checks the structure of every test.

1,560 copied passages that five earlier checks had passed

The verbatim-run check was written because the percentage-based one could not see a copied paragraph inside a long original passage: fifty lifted words in a six-hundred-word note score eight per cent and pass in silence. On its first run across thirteen chapters it found 1,560 runs of twelve or more consecutive words that every existing check had cleared. All were rewritten.

A wrong answer key

A question about tone was keyed to the one option its own chapter text ruled out — and the same chapter's test keyed the right one. No automated check found it; a person reading the chapter end to end did. It is the reason human review passes exist on top of the gates, and the reason this item is on this page.

A gap against the textbook

Checking a Class 8 Maths chapter heading by heading against the book found a section where the arithmetic of powers of ten was covered and the Indian number names were not. The chapter had passed every automated gate, because no gate knows what the book contains.

44 words quoted straight from the book, in a track already signed off

A Class 7 Hindi chapter was marked finished before anyone noticed its textbook had been filed in a different folder, so the copy check had never actually run on it. When it did, it found four verbatim quotations, the longest forty-four words. A check that examined nothing had been read as a check that found nothing.