Editorial standards
1. Sourcing
Every aggregated or drafted story retains one or more source records: the canonical link, the publisher, when it was published and whether it is a primary source (official releases, filings, court registers, company statements, regulator data) or secondary reporting. We prefer primary sources for rules, statistics, corporate announcements and government actions. We do not publish factual claims whose support cannot be traced. We do not fabricate citations.
2. Aggregation and copyright
When several outlets report the same event we cluster them rather than repeat them. Our briefs are original, concise summaries that identify and link the source; we do not copy articles, long passages or photographs, and we do not bypass paywalls or technical restrictions. Rights holders can request removal or correction via the sources page.
3. Political and contested content
We report political actors, elections, legislation and contested public issues neutrally and factually. We separate verified facts from claims, attribute disputed interpretations, include relevant context and never use automated systems to endorse candidates, rank political choices or infer motives.
4. Risk review
Stories are classified by risk before review. Politics, elections, allegations, crime, court and legal matters, deaths, serious health claims, communal or religious issues, investigations and potentially defamatory material require review by an experienced editor and can never be published autonomously. Business, consumer, health, education and international reporting requires an editor's approval. Even routine official releases are approved by a person at launch.
5. Use of AI
AI tools assist our editors with collecting sources, clustering related reports, extracting facts, drafting summaries, classification, tagging and Bangla-English language assistance. AI is a drafting assistant, not the publisher of record: every story is reviewed by a human before publication; the model is never allowed to invent missing details; drafts carry claim-to-source references; and we log the model, prompt version and source set for every generated draft. Stories where AI assisted the draft are marked "AI-assisted draft, human-edited". Our reader assistant, "Ask Nirikh", answers only from our published stories and shows its sources.
6. Corrections
When facts change or we get something wrong, we correct the story, keep a revision history and publish the correction on the story and on the corrections page. Contact: corrections address to be configured.
7. Commercial separation
Advertising, sponsored content and house promotions for Melay are always labelled. Editorial ranking and coverage are never altered because a company advertises, sells on Melay or pays a commission. Sponsored content is produced through a separate workflow and marked on the page.
8. Opinion
Opinion pieces are clearly labelled and carry the author's name.
9. Privacy of people in the news
We include personal information only where it is relevant to a person's public role and already in the public record.