About InsightArise

A small independent publisher. One person, a server, and a rule that every number on the page has to be traceable to a public record.

What this is

InsightArise publishes reference sites built from public datasets. The premise is simple: a great deal of genuinely useful public data is technically available and practically unusable, because reading it requires knowing the agency's own filing conventions, or writing code, or both. Turning it into pages that answer a plain question is unglamorous work that often goes undone.

Each site is a static site. There are no accounts, no logins, and nothing to buy. Pages are generated by a pipeline that pulls the source data, computes the figures, and writes out every page in nine languages on a fixed schedule.

Editorial and AI policy

This matters enough to state plainly, because "AI-generated" covers two very different things.

• Every statistic is computed by code. Counts, rates, indices, rankings, trends and comparisons come from tested, deterministic programs reading the source database. No site statistic is calculated or estimated by a language model, and any number appearing in an AI-drafted summary must already appear in the source passage it was drawn from.
• Explanatory text is templated. The sentences that describe a result are built from a fixed library of conditional phrasings, held as version-controlled source for each language and selected by the computed values. That is why the same finding reads the same way across all nine languages.
• Where a language model is used, it only rewrites. On some pages a short English summary is produced by a language model that is given the site's own already-written passage and asked to condense it. Automated checks then reject the summary if it introduces a number that is absent from the source, adds a negation or a trend word from a defined list that the source does not support, or departs too far from the source vocabulary. Anything rejected is discarded and the page falls back to the templated text. Passages produced this way are labelled on the page.
• Translations are not machine-generated at publication time. The interface strings, phrasing libraries, guides and long-form articles are stored as version-controlled source for each language; nothing is translated at request time.

Independence and funding

InsightArise is not affiliated with, endorsed by, or acting on behalf of any government agency, regulator, manufacturer, dealer or insurer. It does not sell reports, subscriptions, leads or introductions, and it takes no payment from any organisation named on its pages.

Our sites do not currently display advertising. We expect to introduce third-party advertising to support the work; when we do, this page and each site's privacy page will say so before any ad is served. Advertising will not influence any figure, ranking or judgement — those are computed from public records by code that runs before any page, or any ad slot, exists — and ads will be visually separated from content.

Limits we state up front

• Public complaint and incident records are self-reported. They record that someone told an agency something; they are not verified findings, and they are not proof that a product is defective.
• Counts are shaped by how many people know the reporting system exists, not only by how often problems occur. Comparisons are normalized where a fair denominator is available, and the method page says which denominator was used.
• Each site covers a specific jurisdiction's records. Coverage limits are stated on the site, not buried in a disclaimer.

Corrections

If something is wrong, write to us with the page and what you believe the correct value is. We check it against the source record and correct mistakes in our processing. If the source itself is wrong, we say so and leave the source value visible rather than silently substituting our own.

Technical