MojQuiz publishes two kinds of content: short-form quiz questions played in timed rounds, and long-form editorial articles on learning, memory and trivia culture. Both sit on the same site, but they're produced and verified through different processes, and neither is meant to replace primary sources. This page documents exactly how that work happens — what we promise, what we don't, and where our limits are. It also exists so you can hold us accountable to it: if a question is wrong or a citation doesn't check out, the email at the bottom of this page is real and monitored.

Who writes the content

The site is run by a small in-house editorial and engineering team working under Metaflare Interactive LLP (Punjab, India). There is no parent media group, no syndication arrangement with a content mill, and no third-party publisher relationship. Editorial decisions sit with the team alone.

Editorial work is led by Anu Sawhney (M.A., B.Ed.), who is responsible for the editorial line, signs off on every quiz pool before it ships, and writes most of the long-form pieces. Research, drafting, fact-checking and revision are usually split between Anu and the rest of the team, and the default byline on articles reflects that — when a single author leads a piece end-to-end we credit them by name; otherwise the byline reads as Anu's because she signs off on what ships.

How quiz questions are written and verified

Each of the 13 quiz categories has three difficulty pools (easy, medium, hard) with up to 100 questions per pool. Questions go through a four-step process before they ship into a live pool:

  1. Drafting. An initial set of candidate questions is drafted from a category brief that specifies the topics covered and the rough split between subdomains (for example, "Geography — 40% physical geography, 30% political, 20% cultural, 10% cartographic"). We use AI tooling to accelerate the drafting stage, then treat every draft question as a candidate, not a finished question.
  2. Fact verification. Every question — including the AI-drafted ones — is checked against at least two independent reliable sources before it's added to the live pool. Sources include national statistics offices, encyclopedias, primary scientific publications, established reference works, and major news archives. Questions that can't be cleanly verified are cut.
  3. Phrasing review. A separate pass reviews the question for ambiguity, regional bias, dated phrasing ("the current Prime Minister of…"), and answer-choice plausibility. The goal is one unambiguously correct answer and three plausible distractors. Trick questions, double negatives, and answers that depend on date interpretation are revised or removed.
  4. Pool review. Before a pool ships, a final pass looks at the whole 100-question set for duplicate facts asked differently, overrepresentation of any one country or era, and difficulty calibration.

The questions are stored in plain JSON in our repository, which means corrections can be applied and redeployed in minutes once confirmed.

How long-form articles are researched

Editorial articles — the pieces in the notebook and the resources section — are written to a stricter bar. Each piece begins from a working bibliography (typically 5–15 sources for a 2,000-word article), and every empirical claim in the finished piece is tied to a named source either inline or in the supporting paragraph. Where a claim summarises a scientific finding, we name the researcher and the year so a reader can find the original paper: for example, Roediger and Karpicke's 2006 Science paper on retrieval practice, Cepeda et al.'s 2008 meta-analysis on spaced practice, Walker and Stickgold's work on sleep-dependent memory consolidation, or Maguire and colleagues' 2003 study of London taxi drivers and hippocampal volume.

We avoid the common shortcut of citing a secondary blog as if it were the primary source. Where a finding originates in a journal article, the journal article is the citation, not a press release or a magazine piece about it. Where statistics come from a public body (government, WHO, OECD, UNESCO, etc.), we name the body and the year of the report.

Citation policy

For long-form articles, our standing rules are:

  • Cite the researcher, not the explainer. If a piece references the testing effect, the citation is "Roediger and Karpicke (2006)", not a third-party blog post describing the study.
  • Cite the source of statistics. Numbers without attribution are removed in fact-checking. We don't fabricate statistics, and we don't repeat unsourced ones we've seen elsewhere.
  • Cite the year. Especially for population figures, current-affairs claims, and "best X of [year]" articles, the year of the figure or claim is part of the statement.
  • Distinguish replicated findings from speculative ones. The testing effect and the spacing effect are heavily replicated and stated as such; areas with mixed evidence (single-modality "brain training" transfer to general cognition, for example) are framed honestly rather than presented as settled science.

AI assistance disclosure

We use AI tooling (GPT-class large language models) in two specific stages of our work: drafting candidate quiz questions from a category brief, and as a research and outlining assistant for long-form pieces. We disclose this openly because the alternative — pretending the AI step doesn't exist — is dishonest and unhelpful for readers trying to judge what they're reading.

What we do not do is publish AI output unchanged. The AI step accelerates the drafting stage; the verification, citation and editorial revision stages are done by people, and content that fails verification doesn't ship. The fingerprint of an article you read here is the editorial choices, not the AI draft: which studies we chose to cite, which counterexamples we surfaced, which framing we picked, which paragraphs we cut. If you ever read a piece on this site that feels like undifferentiated AI output, please tell us — that's a failure of our process and we want to know.

Update cycle

Different content has different refresh cadences:

  • Reference question pools (13 categories × 3 levels). Reviewed annually. The bulk of questions stay stable year over year; date-sensitive questions (heads of state, populations, sporting records, Nobel laureates) are re-verified during the review pass.
  • "100 questions" trivia sets. Our reference list is reviewed annually. Our 2026 edition is refreshed quarterly, with a heavier weighting toward recent events. Our most-asked list is rebuilt yearly from current search-demand patterns.
  • Editorial articles. When new replicated research changes the picture on a topic we've written about, we revisit the piece and either revise it in place (with an "Updated" date in the header) or publish a follow-up.
  • App and tool reviews. Comparison pieces like "the best quiz apps of 2026" are refreshed annually. Reviews fix at the date of last verification.

Independence and conflicts of interest

On our app review pieces, we note in the article itself when one of the products being compared is our own — the "best quiz apps" guide carries an editorial disclosure at the top stating that MojQuiz is included for completeness and asking readers to judge our entry accordingly. We don't accept paid placements in editorial pieces; we don't accept sponsored content; affiliate links, where used, are nofollow-tagged and disclosed.

What we won't do

  • Fake testimonials or invented statistics. An earlier version of MojQuiz carried homepage testimonials and usage statistics that weren't tied to real data. Those were removed during the 2025 rebuild and won't return. If you see something on the site that looks like a fabricated quote or an unsupported number, please flag it.
  • Clickbait headlines or padded content. Articles are written to be as short as they can be while still being useful, not as long as the topic could possibly stretch.
  • Hidden ads or undisclosed sponsorship. When ads arrive, they will be clearly demarcated as advertising. Reviews and comparisons stay editorial.
  • "AI-detector" arms races. We don't try to make content "AI-undetectable." The right standard is whether the content is accurate, useful and properly sourced — not whether it triggers a particular classifier.

Corrections policy

Mistakes happen. When we publish something incorrect — a wrong answer, an outdated figure, a misattributed quote, a broken citation — we want to know and we want to fix it fast. Here's how:

  1. Email the issue to contact@mojquiz.com with a subject line that starts "Correction:" and the page URL.
  2. We aim to acknowledge corrections within two working days and to verify and ship the fix within five working days, depending on the complexity of the verification.
  3. Material corrections to long-form articles are noted with an "Updated" date in the article header. Quiz-pool corrections replace the affected question and don't carry visible change-logs (the question is updated in place and the new version goes live with the next deploy).

Contact

For corrections, factual questions, editorial concerns, or to flag a piece that doesn't live up to the standards above, write to contact@mojquiz.com. MojQuiz is published by Metaflare Interactive LLP, 729 Surya Enclave, Jalandhar 144009, India. A real human reads everything that comes in.