Inside NTA’s Exam Paper-Setting Process, From Question Creation To Final Selection

The National Testing Agency’s paper-setting system has come under scrutiny after it ordered fresh UGC-NET examinations in English, Commerce and Sociology following the discovery of extensive errors in the June 2026 papers.

An NTA-appointed committee found factual, typographical and translation mistakes, including misspelt names of scholars, garbled book titles, grammatical errors, incorrectly worded questions and non-standard terminology. It also found that a “significant number” of questions had appeared in previous examinations. NTA said the problems were too extensive to be addressed merely by dropping disputed questions.

English and Commerce will be re-conducted on September 9 and Sociology on September 10. Candidates will not have to pay an additional examination fee.

NTA follows several moves to set-up the examination papers.

It Begins With A Question Bank

The process starts with subject experts preparing questions and possible answers. NTA says it works with pools of subject-matter experts to write test items and uses software and artificial intelligence in test creation.

Rather than building one final paper from the outset, a much larger pool of questions is created. According to a government affidavit cited by reports on the NTA’s examination process, the question bank can contain at least five times the number eventually required.

NTA says its model involves creating large, properly indexed question banks for individual subjects. Questions can then be randomly selected using a “scientifically designed blueprint and algorithm”.

Questions Go Through Moderation And Vetting

After questions are prepared, another set of subject experts examines them during moderation.

The checks cover whether questions conform to the syllabus, whether the answers are correct and whether the difficulty level is broadly consistent across possible sets.

A separate vetting stage provides another check. Vetters independently solve questions selected by moderators and compare their answers with those already identified. Differences can then be examined before the paper proceeds further.

The system is structured to limit any single person’s access to the complete examination paper.

Translation Adds Another Layer

For examinations conducted in multiple languages, questions also go through translation and back-translation.

Subject experts translate the material, after which another group checks it against the English version to identify changes in meaning or context.

The government’s description of the process also refers to an air-gapped AI-assisted translation system, where machine-generated translations are reviewed by human experts. Access to the environment is restricted, with no internet or external storage connectivity, according to reports citing the government affidavit.

This stage is particularly relevant to the latest UGC-NET controversy because NTA’s committee specifically found translation, grammar, terminology and wording errors in the three papers being re-conducted.

How Is The Final Paper Created?

The expert panels do not necessarily prepare the exact paper eventually seen by candidates.

NTA says questions are drawn randomly from indexed banks according to an examination blueprint and algorithm.

People familiar with the agency’s paper-setting process told Hindustan Times that panels prepare large question pools from which final sets are generated through algorithm-based randomisation.

“Until the final generation stage, no one knows the exact paper composition,” a central university faculty member involved in NTA paper preparation told the Hindustan Times.

Multiple sets are prepared, providing the agency with alternatives if there are concerns that one set has been compromised.

How Could Old Questions Be Repeated?

The repetition of previously administered questions has raised questions about how material in NTA’s question banks is managed.

A university faculty member involved in UGC-NET paper-setting told Hindustan Times that repetition could occur when an older set is reused alongside newly prepared material and questions are subsequently mixed through randomisation.

“So, this time, they may have picked half the questions from a new set and half from a previous set. When the questions were mixed, the previously used questions could have been picked up randomly. That is how the repetition could have happened,” the faculty member told HT.

This is an explanation offered by a person familiar with the process, not NTA’s official finding on how the June 2026 repetitions occurred. NTA has confirmed the repetition but has not publicly identified its precise cause.

What About Spelling And Typographical Errors?

Typing, proofreading and translation are among the stages where textual errors can enter.

A state university faculty member involved in UGC-NET question-setting told Hindustan Times: “The errors in words are due to typos and oversight by persons responsible for proofreading and translation – mostly contractual staff.”

NTA’s findings now put the effectiveness of those checks under scrutiny. Its committee concluded that the English, Commerce and Sociology papers contained defects serious enough to fall short of the agency’s standards for a “fair and error-free examination”.