AIOInsights/Research
Research

AI-era discovery, measured on public signals

AIOInsights evaluates observable signals that influence AI-era discoverability. It does not guarantee or directly measure recommendations from individual AI platforms. Everything below shows the working.

AIOInsights line art: Research, many businesses, one lineA whole street block drawn in section: eight different businesses side by side, a cafe, a clinic, an office, a gym, a bakery, a hardware store, a salon and a studio, each with its own storefront. Below the street each one's foundation is cut away and reaches a different depth, with a survey mark at its foot. A tall graduated survey staff stands at the left edge, dashed leaders run from the marks across to it, and a single level line is drawn across them at their typical depth. The office standing shallowest has a crack up its wall, and new piles are drawn in dashed plan lines from its footing down to that typical depth.Many businesses AIOInsights line art: Research, many businesses, one line, mobileA whole street block drawn in section: eight different businesses side by side, a cafe, a clinic, an office, a gym, a bakery, a hardware store, a salon and a studio, each with its own storefront. Below the street each one's foundation is cut away and reaches a different depth, with a survey mark at its foot. A tall graduated survey staff stands at the left edge, dashed leaders run from the marks across to it, and a single level line is drawn across them at their typical depth. The office standing shallowest has a crack up its wall, and new piles are drawn in dashed plan lines from its footing down to that typical depth.Many businesses
Written research Insights Long-form research on how AI systems access, interpret, and represent businesses. 599 domains Public signal research Real websites evaluated under the published rubric, on publicly observable signals only. Latest 2026-08-28 Daily research One public-signal read, published daily, with the clearest observed opportunity. Version history Scoring changelog Every change to the scoring engine, dated and versioned, in the open. Reference AIO lexicon The vocabulary of AI-era discovery, defined precisely and sourced. How the score works Methodology What the score measures, how it is weighted, and what it deliberately does not claim.

Independent public-signal research. Not requested or endorsed by the business evaluated. Published research pages evaluate publicly accessible websites for editorial research. A report you run on your own business is private by default and never appears here unless you publish it yourself.

How this research is made

What you can learn here, and what you cannot

Everything in this section comes from one instrument: the same free check anyone can run from the homepage. It requests a website once, the way a crawler would, reads the files crawlers read first, and scores 41 publicly observable signals across six pillars. Nothing here is a survey, a panel or a prediction, and nothing here asks an AI platform what it thinks of a business.

Written research

The Insights essays and the AI discoverability guides explain the patterns the check keeps finding: why a site can rank in search and still be hard for an AI system to describe, why reviews behave like infrastructure, and what changes when the homepage stops being the front door. They cite their sources and say where the approach falls short.

Published evaluations

The public signal research directory holds evaluations of real websites, labeled as research, dated, and stamped with the engine version that scored them. They exist so you can read real output before running your own. They are reference pages, not rankings, and they ask search engines not to index them, because a page about somebody else's business should not compete with that business in search.

How to read a score

A score is a measurement on a date, not a verdict on a company. Two sites with the same number can have very different problems, so read the pillar scores and the named priority fix before the headline number. A pillar that does not apply, such as local presence for an online-only business, is removed rather than counted as zero, and a signal the check could not observe is left out rather than counted as a failure. The full rules are on the methodology page, and every change to them is dated in the scoring changelog.

What this research cannot tell you

It cannot tell you whether a particular assistant will recommend a particular business tomorrow. Answers vary by assistant, by phrasing and by week, and no public tool controls them. What it can tell you is whether the evidence an AI system would need is present, readable and consistent. That is the part a business can fix.

Run a free Trust Visibility Check