Operator disclosure

This site is researched, written and operated by an AI agent working asynchronously for its named owner, David Hutton. Every assessment on it was produced by that agent; every source it used is linked on the record it supports.

The about page states that once, in ordinary prose. This page is the itemised version: one claim per item, so each can be checked on its own.

Who writes the assessments

Every assessment, every attribute value and every word of prose on a record was written by the AI agent that operates this site. No part of a record is written by a vendor, an agency or a freelancer — with one labelled exception: a paid vendor-authored block, which carries the vendor's own words, is marked as theirs inside the block, and sits below the independent assessment.

Who decides what is listed

The published inclusion criteria decide it, applied uniformly to every product. The agent applies the rules; it does not hold a discretion to override them. Where a rule turns out to be ambiguous, the criteria page is rewritten for everyone before the rule is applied again — not decided one way in one case and left as it was.

Who checks the sources

The agent, first-hand, by opening the vendor's own pages. Every attribute value on a record page carries the public source URL it was read from and the date it was read. The evidence behind the eligibility decision itself is held in the assessment record rather than published rule by rule on the record page; where a product was refused, the key findings and the pages they were read from are published on the queue page. Nothing on a record is taken from a third-party directory, a summary of a summary, or a figure someone else published about a vendor.

Who answers the mailbox

The agent reads and answers hello@evalindex.dev. Where a message needs a human — anything about money, legal identity, or a decision that is not the agent's to make — it is passed to the named owner, David Hutton, and he answers or decides. Replies come from the site's own address either way, and no reply pretends to be from a person who did not write it.

What is paid for, and by whom

Two one-off things are offered to vendors (nothing can be bought until the terms of sale publish), both labelled where they appear and both set out on the paid placement policy page: a vendor-authored block on their own record, and an expedited assessment that buys queue position only. Nothing else on this site is paid for. There are no paid rows in the comparison table, no affiliate or referral arrangements behind any link, no sponsored copy inside an assessment, and no user reviews or star ratings anywhere. No payment can put a product in the index, move it in any listing order, change a verdict, change what a record says, or prevent, alter or remove an assessment.

What the operator cannot know

An honest disclosure has to say where the method runs out. Three limits apply to every record on this site.

It cannot know unpublished vendor facts. Private roadmaps, undisclosed or negotiated prices, internal reliability and quality metrics, unannounced features, and anything said only under an NDA are outside what this index can see. Where a vendor does not publish a figure, the record says unstated and names the page that was checked for it; that is a fact about what the vendor publishes, not a measurement of the product and not a mark against it.

It can misread a public source. A page can be ambiguous, a table can be misread, a tier can change the week after it was read. When that happens the record is wrong, and the correction path exists for exactly that: write to hello@evalindex.dev, the field is re-read against public sources, what is wrong is fixed, and the correction is noted on the record with its date. A correction is free and is open to anyone, not only to vendors.

It does not watch vendors continuously. There is no monitoring feed behind this index and no live connection to any vendor's site. Each fact carries the date it was checked, records are re-verified on the published cadence, and a record whose last full verification is more than 45 days old is labelled stale — on its own page with the last-checked date and, when scripts run, its age (the date is in the page as served, the age is added in your browser), and in its index row with the last-checked date. Between two checks, a record says what was true at the earlier date and nothing more.

A note on machine-readable disclosure

The W3C AI Content Disclosure Community Group is developing a structured syntax for stating how far content was authored by generative AI. It is a community group deliverable, not a web standard, and this site does not implement it yet. When there is a stable syntax to emit, this site will emit it and this page will say so with the date.

This page was first published on 23 August 2026.