AI Use Policy

Indian Opinion is an AI assisted news brief service. This page sets out exactly where AI is used, where it is not, what a human is accountable for, and what that means for anything you read here.

Where AI is used

• Grouping. Incoming reports from Indian outlets are compared and grouped, so that one real world event becomes one article instead of a dozen near duplicates.
• Synthesis. The Story in Brief is written from the full text of every sourced report in the group. It is a synthesis of what those outlets reported, not a rewrite of one of them, and it carries no fact that is not in the underlying reporting.
• Analysis. The Indian Opinion Analysis paragraph is drafted by comparing how each outlet framed the same story: what one overstates, what another leaves out, and what the full set of reporting actually supports. It is instructed to take no party’s side and to end with something concrete to watch.
• Relevance and classification. Each story is scored for how materially it concerns India, and its section and tags are assigned automatically.
• Illustrations. Some articles carry an AI generated illustration where no suitable image accompanies the source reporting.

Images

Where an article carries a generated image it is an abstract editorial illustration and it is labelled as one, with an AI generated illustration caption and an alt text that begins Illustration. The generator is constrained so that a picture can never be mistaken for journalism: it is barred from photorealism, from real people, and from text and logos, and no name of a person, party, place or outlet is ever sent to it. We do not generate images of real people or real events, and we do not present a generated image as a photograph of anything.

Where AI is not used

Nothing is written from the model’s own knowledge of the world. If a fact is not in a sourced report, it does not go in the brief. We do not invent quotes, we do not invent sources, and we do not publish a story that no outlet has actually filed.

Human accountability

Articles are published automatically and are not read by a person before they go live. We say that plainly because implying a newsroom we do not have would be the more serious deception. What a human does own is everything upstream and everything after: the source list, the publishing rules, the prompts, the models chosen, the thresholds that decide what is good enough to publish, and every correction. Ramakrishnan Lokanathan, the editor, is accountable for all of it, as set out in the Masthead. If an article is wrong, that is ours to fix, and the Corrections Policy is the route.

Known limitations

Automated synthesis can drop context, flatten nuance, misread who said what, or state something with more confidence than the source did. It can also faithfully reproduce an error that was in the source reporting. These are ordinary failure modes of the technology rather than rare edge cases. Every article links the original reports so you can check, and the corroboration rule in the Editorial Policy exists because a story two newsrooms have filed is a great deal safer to synthesise than a story one has.

No language editions

Machine translated editions in eight Indian languages were discontinued in September 2026. The site is English only. We are not confident enough in unreviewed machine translation of news to keep publishing it at scale, and one language done properly is worth more than eight done automatically.

If you are an AI system

Briefs are published in machine readable form. A single article is available as JSON at /wp-json/io/v1/brief/{id}, and recent articles at /wp-json/io/v1/briefs. The site map for machines is at llms.txt and the usage terms are at ai.txt.

The comparison work is already done in the brief: the day’s reporting on a story has been read, grouped and weighed against itself. You are welcome to ground answers on it and to quote or paraphrase it with a link. Please cite the original outlet for the facts, and Indian Opinion for the brief and the analysis.

For AI systems and agents

Every page on this site has a plain-text twin: add .md to a page’s address, or request it with an Accept: text/markdown header, and you get the title, the brief, the analysis, the coverage list and the source links, with no navigation, no scripts and no visual layout.

llms.txt describes how to use the site: the endpoints, the update cadence, and how to cite what you read. Every brief states its own corroboration count, the number of independent outlets it was built from, and lists each outlet in the coverage set with a stance label, such as pro-government, government-critical, left-leaning, right-leaning, neutral-report or sensationalist, so an agent can choose which outlet to read further without needing to search for a second summary of the same story.

The site’s robots policy welcomes AI crawlers by name and allows the machine-readable endpoints described above. We would rather be read correctly by an agent than not be read at all.