QualiBooth
English

The world’s most advanced accessibility scanner

Static scanners read your markup. Our AI drives your site.

Every competitor runs the same static rule set against your DOM. QualiBooth Deep Scan is an AI agent: it opens a real browser, tabs through your pages, decides which controls are worth provoking, opens your modals, and measures what a keyboard user actually gets - with the screenshot to prove it.

Scan your site free See it on your pages
12 AI checks
no static tool can run
Evidence attached
a measurement and a screenshot
The QualiBooth report with Deep Scan findings badged alongside the static ones.

The part nobody was scanning

Automated coverage stopped improving. So we put an AI agent on the page instead.

Static analysis reads the DOM and the CSS and reasons about what should happen. It is genuinely good at that - and it is why every accessibility tool on the market finds broadly the same issues on your site. The defects it structurally cannot see are the ones that only exist while the page is running. Until now, finding those meant putting a person on the page. Deep Scan sends an agent instead - to every page, on every run.

What every static scanner sees

A snapshot of the markup, at rest.

  • This button has no accessible name.
  • This text is 3.1:1 against its declared background colour.
  • This ARIA attribute is not allowed on this role.
  • Text over a photograph - needs review. A human should look at this.

What the Deep Scan agent measures

A real browser, AI-driven, with the clock running.

  • The focus ring is zero pixels different from unfocused. outline:none won.
  • The white heading is 2.8:1 against the pixels actually behind it.
  • Tab reaches this control, then the sticky consent bar covers all of it.
  • Escape closes the dialog, but focus lands on the body, not the trigger.

The check list

Twelve checks that need a running page

The agent decides what is worth exploring; every check then measures rather than infers. Each finding carries the number it measured, the threshold it crossed, and - for most checks - the screenshot of the moment it happened.

  • 01

    Focus visibility

    Screenshots each control focused and unfocused and compares the pixels. It never reads your CSS: outline:none followed by a :focus-visible rule is a pass, and only the pixels know.

    WCAG 2.4.7 WCAG 2.4.13

  • 02

    Focus obstruction

    Hit-tests the focused control and measures how much of it your sticky header or cookie bar paints over. Runs in both tab directions, because each one strands a different control.

    WCAG 2.4.11 WCAG 2.4.12

  • 03

    Tab order

    Walks the real tab sequence and compares it against DOM order and against visual reading order, with a numbered screenshot of the path it took.

    WCAG 2.4.3

  • 04

    Keyboard traps

    Finds focus cycles that leave the rest of the page unreachable. A modal that confines Tab and releases on Escape is correct behaviour, and is not reported.

    WCAG 2.1.2

  • 05

    Modal lifecycle

    Opens every dialog and asserts the whole contract in order: focus enters, Tab stays in, the background leaves the accessibility tree, Escape closes, focus returns to the trigger.

    WCAG 2.1.2 WCAG 2.4.3 WCAG 4.1.2

  • 06

    Live announcements

    Reconstructs when each live region came into existence relative to its content - catching the toast that was created already holding its message, and therefore announced nothing.

    WCAG 4.1.3

  • 07

    Hover and focus content

    Scripts all three requirements for tooltips and popovers: the pointer can travel into the content, Escape dismisses it without moving the pointer, and it does not hide itself.

    WCAG 1.4.13 WCAG 2.1.1

  • 08

    Contrast over imagery

    Captures the text, captures the same region with the glyphs made transparent, and compares each glyph pixel against the background actually behind it. This is the population static tools mark needs review.

    WCAG 1.4.3

  • 09

    Reflow at 320px

    Renders the page at 320 CSS pixels and names the outermost elements still forcing horizontal scroll, exempting content that lives in its own scroll region.

    WCAG 1.4.10

  • 10

    Text spacing

    Applies the spacing the criterion demands and compares the geometry before and after: text clipped by fixed-size boxes, blocks that have collided.

    WCAG 1.4.12

  • 11

    Forced colours

    Emulates Windows High Contrast mode and measures each small graphic against its own surroundings, before and after, to find the icons that vanish.

    WCAG 1.4.11

  • 12

    Accessible names

    Uses Chrome’s own name computation and its source list: names that hide the visible label, names duplicated by their description, references that resolve to nothing.

    WCAG 2.5.3 WCAG 4.1.2

No static scanner on the market can run a single one of these twelve checks. That is what makes QualiBooth the world’s most advanced accessibility scanner - and why it finds what everyone else hands you as a clean report.

Evidence, not opinions

Every finding survives an argument with a developer

The hard part of an accessibility finding is not detecting it. It is being believed at the stand-up three days later. So every Deep Scan finding reports the measurement it took, the threshold it compared that against, and the annotated screenshot of the moment it happened.

  • A measured number inside the sentence - entirely covered by the fixed consent bar, not may be obscured.
  • The exact WCAG 2.2 success criterion, linked to its Understanding page.
  • Remediation written for that finding, not for the rule in general.
  • Anything the engine could not reach a verdict on is reported as such, never as a pass.
A Deep Scan evidence screenshot: the focused control, the element covering it, and the measured coverage.

Manual QA, on every scan

The keyboard pass your team runs before release - now run by an agent on every page, every night

These are not new requirements. They are the checks a good accessibility tester already does by hand, which is exactly why they get done on five pages before a release and on none of them afterwards.

  • Tab through the page and watch for the focus ring

    Screenshots every tab stop focused and unfocused, and diffs the pixels

  • Open each modal, press Escape, see where focus lands

    Opens every dialog and asserts the full contract, in order

  • Trigger a toast and listen with a screen reader

    Records when each live region existed relative to its content

  • Zoom to 320px and hunt for horizontal scroll

    Renders at 320px and names the elements forcing it

  • Squint at the white text on the hero photograph

    Compares each glyph pixel against the background behind it

What it will not touch

The engine clicks a bounded set of controls to make your page do something. It will not click anything whose accessible name reads like delete, remove, sign out, unsubscribe, pay, checkout, place order, publish or archive. Scanning your site must never place an order or delete a record - even though clicking more would surface more findings.

How it runs

An AI agent driving a second browser, alongside the static engine, in the same scan.

  1. 01

    The agent takes the wheel

    A second Chromium runs beside the static engine on the same authenticated session, driven by the agent - so the pages behind your login are measured too.

  2. 02

    Desktop and mobile

    Every URL is measured at 1280×720 and again at 375×667. A focus ring that survives your desktop header often loses to the mobile one.

  3. 03

    One merged report

    Behavioural findings arrive in the same list as the static ones, badged Deep Scan and sorted by severity. There is nothing new to learn.

  4. 04

    Evidence kept with the run

    Annotated screenshots are stored alongside the scan, so a finding from three months ago still shows its proof when you go back to it.

The Deep Scan summary panel in a QualiBooth URL report.

In your report

Badged, counted, and never silently absent

Deep Scan findings are called out rather than folded in quietly - they are the ones your current tool cannot produce. And an absence of findings is a result too, so every report states what was examined and what could not be reached.

  • A Deep Scan badge on every behavioural finding in the issue list.
  • Checks completed, element examinations and evidence captured, per URL.
  • Anything that could not be fully checked, named together with the reason.
  • Exported to CSV and JSON with the rest of the scan run.
12
behavioural checks
15
WCAG 2.2 criteria
2
viewports per URL
100%
findings with a measurement

Questions we get asked

How is this different from an accessibility overlay?
Completely. An overlay injects JavaScript into your live site and tries to patch accessibility from the outside, usually making it worse. Deep Scan changes nothing about your site - it is an agent that visits it the way a user would, measures, and reports what it found.
Does this replace manual accessibility testing?
No, and we would not claim it does. The agent automates the mechanical half of a manual pass - the keyboard sweep, the modal contract, the 320px reflow - so your testers and our auditors spend their time on judgement instead: whether the reading order makes sense, whether the alt text is true, whether the flow is actually usable with a screen reader.
Can it break anything on my site?
It interacts with a bounded, filtered set of controls and never with anything that reads as destructive or transactional. No deleting, no signing out, no checkout. It is a read-mostly visit from one browser.
Can it scan pages behind a login?
Yes. The agent reuses the authenticated session the scan has already established, so anything your scan can reach, it can measure.
Why did my accessibility score change when this was switched on?
Deep Scan adds both new findings and new passes to the score, so a score computed with them is not comparable with one computed without. Scores stay comparable within a scan programme and across runs made with the same setting - we do not flip it mid-programme underneath you.
What is the AI actually doing?
Deciding, and then proving. The agent explores your page the way a tester would - choosing which controls are worth provoking, which dialogs to open, how far to push before it stops - and every call it makes is backed by a measurement and a screenshot rather than a guess. That is the deliberate design: autonomous where a person would be slow, and evidence-backed where a language model on its own would be unreliable.
Why do you call QualiBooth the world’s most advanced accessibility scanner?
Because no other scanner measures behaviour on a running page. Every tool on the market reads your markup and infers - QualiBooth does that too, and then puts an AI agent on the page for twelve further checks no static engine can produce, each one backed by a measurement and a screenshot. That combination exists nowhere else.

Find out what your current scanner has been missing

Run a free scan, or let us point the Deep Scan agent at the user journey that matters most to your business.

Scan your site free Book a demo