Editorial
Editorial policy
Every question on Target Dentistry is written by a dentist, reviewed by a second dentist, referenced, versioned and, when wrong, corrected in public.
Last updated 25 October 2026
What this policy covers
This policy applies to every multiple-choice item, clinical scenario question (CSQ), explanation, exam hub, notice transcription and guide published on Target Dentistry, in English and in Nepali. It is the standard our writers and reviewers sign up to, and the standard you can hold us to.
The two-signature workflow
- Topic brief. Each item starts from a topic brief derived from the exam's subject blueprint and, later, from the weak-topic heatmap of real attempts.
- Draft. A BDS writer drafts the item from a named reference, or edits a machine draft (see AI assistance below). We plan for a high rejection rate of drafts; a weak draft is discarded, not polished.
- Verification by the writer. The writer verifies the stem, four or five homogeneous options, a 120-250 word explanation, one reference with edition and page, tags and difficulty against an item-writing flaw checklist modelled on the NBME guide. This is the first signature.
- Independent review. A second dentist with an MDS reviews the item in full and signs it. The lead reviewer signs every CSQ and every image item. An editor runs a stem-plus-answer duplicate check. This is the second signature; nothing is published without it.
- Publication as a version. The signed item is published as an immutable version. The byline shows the writer, the reviewer, their NMC numbers and the review date.
What every item must contain
- A single best answer; options are homogeneous in form and length, with no “all of the above” or “none of the above”.
- An explanation of 120-250 words that explains the correct answer, addresses each distractor and ends with one exam tip.
- One reference with edition; page numbers are verified by the reviewer before the second signature.
- Exam tags, a subject and topic, and a difficulty rating that the quality loop later checks against real response data.
- Where an image is used, licence, creator and consent details stored with the file (see Images and licences).
AI assistance, disclosed
We use large language models to produce first drafts of some text items from a reference and a topic brief. A dentist edits or rejects every machine draft, and no item is published on a machine draft alone. Language models never write image-based stems. Where an item started from a machine draft, its item record says so and the disclosure is available on the question page.
Explanations are checked for factual accuracy by the reviewer against the cited reference, not against the model. We do not use models to generate references, and a reference that cannot be located is a reason to reject the item.
Recall-based items
Candidates sometimes report what an exam asked. We treat these reports as topic signals only. A recall-based item is rewritten into a new stem, new options and a new explanation; we never republish a reconstructed paper. Every such item carries this visible label:
Recall-based item, reconstructed and rewritten from candidate reports; not an official NMC/MEC/PSC paper
No past-paper page opens before a written intellectual-property opinion on Nepal's Copyright Act 2059 and on candidate confidentiality rules, and a takedown workflow is in place for any content a rights holder disputes.
Sample items before launch
Until two dentists have signed an item it is labelled “Sample item, pending dentist review” wherever it appears, is kept out of search engines and is not counted as part of the question bank. Sample items exist so you can see the format; treat their content as unverified.
Images and licences
Radiographs and clinical photographs come from an owned, consented library (written patient consent, de-identified) and from sources whose licences allow use in a paid educational product, checked per file. We do not use material under licences that forbid use in paywalled question banks. Illustrations generated with AI are labelled as synthetic and are never presented as patient images. Each image record stores its licence, creator, source link and any share-alike flag.
Corrections and errata
A published item is never silently edited. A correction creates a new item version and an entry in the public errata log with the item, the change, the new version number and the dentist who signed the correction. Attempts you made against an earlier version keep that version, so your past results remain comparable. Our quality target is fewer than 1% of published items corrected per quarter.
You can challenge any answer from the question page or from a mock review. Challenges are triaged within 72 hours and either answered with the reference or converted into a correction.
Quality loop after publication
Once an item has 100 responses we screen its difficulty and discrimination statistics. Items that are too easy, too hard or that fail to discriminate are returned to review within 14 days and either revised or withdrawn. New items ship with their pages excluded from search engines; a page becomes indexable only after length, two-signature, duplicate and tag checks pass.
Exam formats and attribution
Items, marks, timing, negative marking and pass rules for each exam are configuration, not claims. Each exam page names the source for every figure and its confidence, and says explicitly when a value is not confirmed by the Nepal Medical Council, the Medical Education Commission or the Public Service Commission. We never describe a format as “official” on the strength of a secondary portal. When a regulator corrects a rule, the configuration version changes and earlier attempts keep the version they were taken under.
Bylines and profiles
Every byline names a real person with a public profile on /authors, their credentials, NMC number and the count of items they have reviewed as read from the question bank. We do not publish “editorial team” or “admin” bylines. Until a reviewer agreement is signed, the profile is shown as a placeholder and is kept out of search engines.
