# Verdicts, priority and coverage

> See why Aiscovery gives a verdict instead of a single score, how fixes are prioritized and how a result shows what it did and did not cover.

- Canonical URL: https://aiscovery.io/docs/scoring
- Markdown version: https://aiscovery.io/docs/scoring.md
- Last updated: 2026-10-03

A verdict is limited to what we tested. Critical fixes stay at the top, and gaps stay visible as Not tested yet or Couldn't check.

- Last reviewed: 2026-10-03
- Reading time: 3 min
- [Documentation](https://aiscovery.io/docs)

## Verdicts instead of scores

Aiscovery doesn't turn your website into a single number. A number hides which pages we read, which checks passed and which questions we couldn't answer. Each result starts with a verdict instead: one sentence that says what works and what to fix first, limited to what we tested.

Verdicts follow fixed rules, applied in order, so the same evidence always gives the same verdict. For example, if our checker couldn't read your homepage, the verdict says so and doesn't guess about the rest of the site. If your robots.txt asks the AI crawlers we checked not to read your site, the verdict says that first.

## What we tested and what we haven't

Under the verdict, the result lists the areas it covers, one line each:

- **Access:** whether pages load for automated visitors and redirects work.
- **AI crawler rules:** how many AI crawlers your robots.txt allows. This is declared policy, not a test of what each crawler does.
- **AI-facing files:** robots.txt, sitemap and llms.txt. llms.txt is optional and never lowers your result.
- **Page basics:** titles, descriptions, main headings, page language and structured data syntax.
- **Key tasks with real AI agents:** coming later. Until then this line says "not tested yet".

"Not tested yet" means we don't know. It is never counted as a pass or a fail, and it never lowers the result.

## Agent-facing technology

Some websites declare agent-facing technology, such as an llms.txt file, declarative WebMCP forms or security headers. When our checker sees one on the pages it reads, the technical details mark it as detected (declared) and label how mature the technology is.

Detection proves presence only. We don't run MCP, OpenAPI or WebMCP tools yet, and a missing optional technology is not a failure. A website without any of them can still work well for AI crawlers and agents.

## Critical fixes come first

Fixes are ranked by priority: Critical, High, Medium or Low. Priority reflects how much a fix matters for what we tested, not how hard the change is.

A Critical fix is never hidden behind good results elsewhere. While one is open, it stays at the top of the fix list under Fix these first. After you mark it fixed, a retest runs the same check the same way; only when the problem is gone is it shown as Verified fixed.

## Confidence

We don't publish a confidence number either. Instead, the result shows how complete it is:

- Checks that couldn't run, for example because a firewall stopped our checker, are shown as "Couldn't check", never as failures of your website.
- Areas we haven't tested yet are listed as such, so a short result can't look more complete than it is.
- Coming later, for key tasks: a result names the exact agent and model, how many test runs it covers and whether each was AI-checked or reviewed by Aiscovery.

Next guide: [Results](https://aiscovery.io/docs/audit-results)

Public documentation describes what runs today, marks what is coming later and is updated when our checks change.

---

This is the Markdown version of the page for AI agents. The HTML page at the canonical URL is authoritative. Site index: https://aiscovery.io/llms.txt