AI transparency

• Detection — scanning sources continuously to spot a topic gaining momentum.
• Scoring — computing heat, velocity and status from interest and source signals.
• Extraction — pulling facts out of source reporting and rewriting them in our own words.
• Summarising — the “why it’s trending” paragraph and the short answers.
• Headlines — the title on a trend page is written from the headlines several outlets published, synthesised across them rather than copied from any one of them. A candidate is rejected outright if it reads too closely to a single source’s wording, if it characterises a named person rather than stating what they did, or if its subject does not appear in the underlying coverage. When nothing safe can be written, we fall back to plainly naming the entities the coverage is about — a duller title, never a riskier one.

Every draft is re-checked against the original source by a separate verification pass that scores factual accuracy, hallucination risk, legal risk and how faithfully it reflects the source. Failing any hard threshold blocks publication outright — the item is dropped, not softened.

If the source cannot be re-fetched at verification time, the item does not publish. We never fall back to what a model “remembers” about a topic.

Being specific here is the point of this page:

• Entity confusion — mistaking one similarly-named person, product or company for another.
• False precision — turning “about ₹1 lakh” into an exact figure the source never gave.
• Stale framing — treating a recirculated old story as new. Our timeline exists partly to catch this.
• Momentum ≠ importance — the ranking reflects attention, not significance, and can elevate trivia over substance.

• Invent a quote, a statistic or a source.
• Generate a photorealistic image presented as real.
• Write an allegation about a named person or company on its own.
• Alter a historical snapshot once recorded.

If you spot something wrong, tell us via contact. Corrections are logged publicly at corrections, including ones caused by automation. We do not hide behind the machine.