Webscan Documentation

What Webscan Does

Webscan is an accessibility operations platform: it finds barriers, helps you fix them, proves they're fixed, and keeps them from coming back. Below is every core capability and the accessibility value it delivers.

For step-by-step instructions, see the User Guide.

Find barriers

Capability	What it does	Why it matters
Automated scanning	Crawls a site and runs axe-core on every page	Catches code-level WCAG failures at scale, fast
Behaviour engine	Instruments real page interaction to detect focus traps, focus stealing, invisible focus, and unannounced status changes	Catches barriers that only appear when a keyboard or screen-reader user actually interacts — invisible to static checkers (WCAG 2.1.2, 3.2.1, 2.4.3, 4.1.2, 4.1.3)
Recorded audit sessions	You browse the real site through a browser extension; webscan analyses the captured trace	Audits the site as used, including logged-in flows, without scripting
Agentic exploration	Webscan navigates the site on its own (read-only or gated full-action) to reach states a shallow crawl misses	Coverage of deep and interactive states
Journeys	Saved, repeatable multi-step flows (e.g. checkout) that can be re-run	Regression-checks the paths that matter most
Manual audits	Structured human checkpoints and results	Covers judgement-based criteria (alt-text quality, reading order, focus order) machines can't decide
Custom rules	Organization-defined DOM/selector checks	Encodes your house patterns and client-specific requirements
Authenticated scanning	Encrypted captured login state	Reaches barriers behind a sign-in

Judge and prioritise

Capability	What it does	Why it matters
Verdicts	Mark findings Confirmed or False positive; they carry forward across scans by fingerprint	Kills repeat noise; you never triage the same false positive twice
AI-suggested verdicts	Optional pre-triage with rationale	Speeds review; you stay the decision-maker
Deterministic prioritisation	Ranks findings by severity, WCAG level, and recurrence	Fix what matters first
Per-organization model	Learns from your verdicts to sharpen prioritisation, trainable from a dashboard	The tool adapts to your standards over time
Dataset export	Export the labelled verdict dataset	Transparency and offline analysis

Fix

Capability	What it does	Why it matters
Issues	Track remediation with comments, evidence, and source artifacts	A durable record from discovery to done
AI remediation assistance	Drafts a concrete fix (explanation + code) for review	Cuts time-to-fix; always human-reviewed
GitHub pull requests	Opens a PR from a reviewed fix draft	Fix lands in the codebase with a full audit trail

Prove and sustain

Capability	What it does	Why it matters
Retest loop	Re-checks whether a finding is still present	Objective confirmation a fix worked
Auto-verify	Re-renders the page and confirms the specific finding is gone, with evidence	Trustworthy, evidence-backed closure
Monitoring	Scheduled re-scans with regression alerts	One-off audit becomes ongoing assurance
Reports	Client-ready exports to HTML, PDF, and Word, with org branding	Deliverables that combine automated, behavioural, and manual evidence

How the pieces fit

        FIND                    JUDGE                 FIX                 PROVE
  ┌──────────────┐        ┌──────────────┐     ┌──────────────┐    ┌──────────────┐
  │ scans        │        │ verdicts     │     │ issues       │    │ retest       │
  │ behaviour    │  ───►  │ prioritise   │ ──► │ AI fixes     │──► │ auto-verify  │
  │ recordings   │        │ (model)      │     │ GitHub PRs   │    │ monitoring   │
  │ exploration  │        └──────────────┘     └──────────────┘    └──────────────┘
  │ journeys     │                                                        │
  │ manual audit │                                                        ▼
  │ custom rules │                                                   ┌──────────┐
  └──────────────┘                                                   │ reports  │
                                                                     └──────────┘

Every source of evidence lands in the same finding/verdict/issue pipeline, so a single report tells one coherent story — no matter whether a barrier was caught by a scanner, a real interaction, or a human's judgement.

What it deliberately does not do

It doesn't replace human judgement — machines can't decide whether alt text is meaningful. Webscan makes the human parts faster, not optional.
It doesn't act recklessly in production — consequential actions and full-action exploration are gated by the website's declared environment.
It doesn't capture your secrets — the recorder drops typed text and redacts sensitive fields in the browser before anything is transmitted.