This page explains how Badiyaa produces its journalism, including our use of artificial intelligence.
How we report
Our newsroom continuously monitors real-time news and audience signals across multiple independent sources. Every article is then written in our own words, with its own structure and angle, and enriched with context from open knowledge sources such as Wikipedia, Wikimedia Commons, peer-reviewed research and official records. We publish only facts, figures and quotes we have verified, and stories touching politics, health, money, crime or allegations about a named person require corroboration from at least three independent sources. Each article closes with open reference material for further reading.
Our editorial core: original by design
Every Badiyaa article is built on four originality principles. They are not guidelines an editor may weigh; each is enforced by an automated gate that a draft must pass before it can be published, and each gate’s result is recorded with the article.
| Principle | What it means | How it is enforced |
|---|---|---|
| 1 · Our words | Every sentence is written by us. Exact wording from anyone else appears only as a quotation, in quotation marks, with the speaker named. | No more than 3% of an article’s eight-word phrases may match any single source, and no run of 12 or more consecutive words may match a source outside quotation marks. A draft over either limit is rewritten. |
| 2 · Our structure | A story is assembled from several independent sources, never adapted from one. We choose the opening, the order of points and the angle. | At least three independent sources (four for politics, health, money and crime). A draft that follows one source point by point, in that source’s order, is rejected even if every word has changed. |
| 3 · Our value | Every article adds something the reader would not get from the sources: context, a timeline, a practical guide, a comparison, a calculation, a plain-language explanation, or what happens next. | Each draft must name its added element and quote the passage that delivers it. The gate checks that the passage is in the article and is not itself taken from the research. |
| 4 · No repeats | We do not publish the same story twice. A follow-up takes a clearly different angle and links to our earlier piece. | Every draft is compared with our own articles from the last 30 days and with stories in production. A substantial repeat that does not link to the earlier article is held for an editor. |
These checks run alongside fact-checking and headline review (Figure 3), and every published article is also open to our Reader Verifiers. The limits are deliberately far stricter than the point at which wording would be considered copied, so that originality is never a judgment call.
Our workflow
Every Badiyaa story moves through a three-phase, human-in-the-loop editorial pipeline. People set direction and hold final authority; AI systems carry out the high-throughput work in between, inside a governance layer drawn from Google’s publisher policies and our own standards.
-
20%
Plan
Human-led · editorial strategy- Story ideas from our content strategy team and from readers; reader tips get the same review as our own picks
- Audience-interest and trend-signal analysis across global markets
- Daily editorial agenda and angle selection
- Exclusion screening: what we choose not to cover
Planning teamLed by Swati J, Story Editor (formerly Times Internet) · Trends Editor · Section Editor -
60%
Produce
AI-assisted · synthesis- Research: real-time signal monitoring and multi-source verification, enriched with open reference data
- Original-language synthesis in calibrated house voices
- Rights-cleared imagery with full attribution
- Automated integrity gates: the four editorial-core originality gates, claim-level fact consistency and headline integrity
-
20%
Review & approve
Human-led · editorial standards- Fact check: our verification team checks every story against its sources, usually within hours of publishing, and corrects or removes anything that doesn’t hold up
- Line editing and headline refinement
- Standards and quality assurance
- Approval authority: pre-publication sign-off for politics, health, finance and crime; post-publication review for every other story
Standards teamLed by Prasanna Iyer, Final Approval Editor (formerly Google and Times Internet) · Copy Editor
Reader Verifier Network
AI lets a small newsroom publish quickly. Human verification at scale is what keeps it accurate. Alongside our editors, Badiyaa is building a large, distributed network of human verifiers: readers across our markets who check stories against what they know and flag errors from the Spot an error? panel at the end of every article.
| Phase | Status | What it includes |
|---|---|---|
| 1 · Recognition | Live | Error reports from every article, reviewed by an editor; dated, public correction notes; optional credit by first name and city; a public roll of Badiyaa Verifiers; corrections declared in structured data. |
| 2 · Rewards | Planned | Limited-edition Badiyaa designs and other thank-yous for the most active verifiers, based on verified corrections. |
| Vision | Ongoing | A large, distributed network of human verifiers across every market and section we cover, so each story is read by many informed people, not only our own editors. |
Editorial authority does not move to the crowd. Reports are evidence for our editors to weigh; the Editorial Desk decides every correction and remains accountable for it.
Get involved
Send us a story you think more people should know about, or join the people who check our stories against their sources. Spotted a mistake in a story? Use the Spot an error? panel at the end of that article.
Suggest a story
Tell us a story you think more people should know about. Links to sources help.
Join our verification network
Help check stories against their sources. Tell us a little about yourself and the topics you know well.
Governance and safeguards
This section sets out, in detail, the control framework that governs every Badiyaa article: which external standards we align to, how each one is put into practice, and where human judgment sits in the process. It is written for readers, partners, compliance reviewers and automated quality systems alike. We draw first and most heavily on Google’s published guidance for publishers, and then on recognised academic and international frameworks for the responsible use of AI.
1. Alignment with Google’s publisher guidance
| Google standard | What it asks of publishers | How Badiyaa implements it |
|---|---|---|
| Helpful, reliable, people-first content | Content made primarily for people, with clear answers to “Who, How and Why” it was created, including disclosure of AI or automation where readers would reasonably expect it. | Who: every article carries a byline linked to a profile page. How: every article carries a visible AI-assistance disclosure linked to this policy. Why: stories are chosen for reader interest and must add context or explanation beyond the headline facts. |
| Spam policies, including scaled content abuse | No mass production of pages, by AI or any other method, whose primary purpose is manipulating rankings rather than helping people; no stitching or rewording of other content without adding value. | Daily volume is capped and increased gradually. Every draft passes the four gates of our editorial core: a 3% cap on shared phrasing with any source, no copied run of 12 or more words, at least three independent sources with no source followed point by point, a proven Badiyaa value-add, and no repeats of our own recent articles. Stories without added value are not published. We do not create doorway pages, keyword pages or networks of sites. |
| E-E-A-T and “Your Money or Your Life” topics | Trustworthiness above all, with the highest bar for topics affecting health, financial stability, safety or civic life. | Politics, health, finance, crime and allegations about named people are classified as high-risk. They require corroboration from at least three independent sources and pre-publication approval by an editor. We do not give medical, legal or financial advice. |
| Google News policies: transparency | Clear dates and bylines; information about authors, publication and publisher; ownership; contact information. | Published and updated timestamps on every article; bylines; About, Editorial Policy, Corrections and Contact pages linked from every page; publisher identity declared in structured data. |
| Google News and Discover content policies | No deceptive practices, manipulated media, misleading or sensational headlines; content that meets general news-quality standards. | Automated headline-integrity gate (no clickbait patterns, no capitals, no teaser constructions) followed by human headline review. Images are licensed photographs credited on the page; we do not publish AI-generated photographs presented as real events. |
| Search Quality Rater Guidelines | Effort, originality, accuracy and a clear purpose; low ratings for content with little effort or value. | House-style rules, section-specific editorial voices, a banned list of formulaic phrasing, and claim-level fact checking against our research before any story can publish. |
| Google News Initiative good practice | Responsible, transparent newsroom adoption of AI, with editorial standards kept in human hands. | AI is confined to research synthesis, drafting support and automated checks. Editorial decisions, standards and approval authority remain with people (see Figures 2 and 3). |
2. Academic and international frameworks
- Human oversight taxonomy. The European Commission’s High-Level Expert Group on AI distinguishes human-in-the-loop (a person approves each output), human-on-the-loop (a person monitors and can intervene) and human-in-command (people govern the overall system). Badiyaa applies human-in-the-loop control to every high-risk story, human-on-the-loop review to all other stories, and human-in-command governance over the pipeline as a whole.
- Levels of automation. Following Parasuraman, Sheridan and Wickens’ model of the types and levels of human interaction with automation (IEEE, 2000), automation at Badiyaa is concentrated in information acquisition and analysis. Decision selection (what to cover, what to publish) and action implementation (publication of high-risk stories) remain human functions.
- Human-centred AI. In line with Shneiderman’s human-centred AI framework (2020), the pipeline is designed for high automation and high human control at the same time, rather than trading one for the other.
- NIST AI Risk Management Framework 1.0 (2023). Govern: this Editorial Policy and named editorial roles. Map: risk classification of every story by topic. Measure: automated originality, fact-consistency and headline gates with recorded results. Manage: holds, revisions, corrections and continuous refinement of rules.
- Paris Charter on AI and Journalism (2023). We follow its core principles: human agency stays central to editorial decisions, the publisher is accountable for everything it publishes, and the use of AI is disclosed to readers.
- EU AI Act, Article 50. From 2 August 2026, AI-generated text published to inform the public on matters of public interest must be disclosed, unless it has undergone human review or editorial control and a person holds editorial responsibility. Badiyaa meets both conditions: our Editorial Desk holds editorial responsibility for every article, and we disclose AI assistance regardless.
3. Human checkpoints
| Checkpoint | Owner | Scope |
|---|---|---|
| H1 · Editorial agenda | Planning team, led by Swati J | Approves the daily story list, angles and exclusions before production begins. |
| H2 · Pre-publication approval | Prasanna Iyer, Final Approval Editor | Mandatory sign-off for politics, health, finance, crime and allegations about named people, and for any draft that fails an automated gate twice. |
| H3 · Post-publication review | Standards team, led by Prasanna Iyer | Every other story is reviewed by an editor after publication, and corrected or withdrawn where needed. |
| H4 · Corrections and feedback | Standards team | Reports from the Reader Verifier Network and other readers are triaged by an editor, fixed and disclosed on the article with credit, and recurring issues feed back into the rules used in planning and production. |
4. Signals for automated quality systems
For search engines and other automated reviewers, the same commitments are expressed in machine-readable form: every article is marked up as a NewsArticle with its publisher, publication and modification dates; the publisher record declares this page as its publishing principles and links its corrections and feedback policies; the AI-assistance disclosure appears in the visible text of every article; and a plain-language guide for AI systems is published at /llms.txt.
This framework is reviewed regularly and updated as Google’s guidance and the wider regulatory landscape evolve.
Use of AI
We use AI tools to help draft articles from the sources we have gathered and to check every claim and quote against those sources. Drafts that copy wording from a source are rejected and rewritten. Each article produced this way carries a note saying so.
Human review
Every article is reviewed by our Editorial Desk. Stories about politics, elections, health, money, crime, deaths and disasters, or the conduct of named people, are reviewed and approved by an editor before publication; other stories are reviewed after publication and corrected or removed if needed. Any story that fails our automated checks is held for an editor.
Ownership and editorial responsibility
Badiyaa is owned and published by Rezilient Digital, Mumbai. Our daily story list is set by Story Editor Swati J, and final editorial approval rests with Final Approval Editor Prasanna Iyer. See About us for ownership details.
Independence
We do not accept payment to cover or change a story. Advertising, if shown, is clearly separate from editorial content.
Corrections
When we get something wrong, we fix it and say so. See our Corrections page.