Image description: A blind tester wearing headphones runs JAWS on a laptop while a colleague using a screen magnifier reviews the same page on a second display — two testers with disabilities conducting a manual accessibility audit side by side.

Reading Time: 12 minutes

Automated accessibility scanners catch somewhere between 57% and 70% of WCAG violations on a typical page. The remaining 30 to 40% is where lawsuits, complaints, and real user friction live. Keyboard traps that only trigger on the third tab cycle. Focus order that runs visually top-down but jumps to the footer in DOM order. ARIA-live regions that fire on page load instead of when content changes. Alt text that passes the “alt attribute present” check but reads “Image 47.png” to a screen reader. None of these are detectable by static analysis.

A manual accessibility audit fills that gap. A manual audit performed by people with disabilities goes further — it tests whether the fixes you shipped actually work for the users they were supposed to help. For organisations in the scope of the European Accessibility Act or ADA Title III, it is the only way to verify lived usability. Before commissioning one, establish a quick baseline with our free accessibility scanner.

Why manual audits matter — the limits of automated scanning

Automated scanners are excellent at what they do, and what they do is bounded. They run rule engines — axe-core, Lighthouse, WAVE — against a parsed DOM, flagging missing alt attributes, contrast ratios, form labels, landmark errors, and broken ARIA references. The WebAIM Million 2024 analysis found scanners detect roughly 57% of axe-rule violations, a figure that tracks with what most accessibility engineers report from internal data.

Now consider what they cannot detect.

A page can have an alt="chart showing quarterly revenue" on every image and still be unusable for a screen-reader user, because the chart actually shows headcount. The scanner sees a non-empty alt string and moves on.

A form can have every label correctly associated, perfect contrast, and clean semantics, and still trap a keyboard user when the date-picker opens a modal that never receives focus. Static analysis sees a valid component tree; only a human pressing Tab discovers the trap.

A single-page app can announce nothing when content updates, because the aria-live region was added at build time and the framework re-renders it on every navigation. The scanner sees the attribute. The screen reader hears silence.

Focus order is a famous trouble area. CSS order, flex-direction: row-reverse, and absolutely-positioned elements decouple visual order from DOM order. A page can look correct and read like a jigsaw.

These categories — meaningful alt text, focus order, keyboard usability of complex widgets, screen-reader narrative quality, dynamic content announcements, error recovery flows — form the 30 to 40% gap that only a human reviewer can close. Among human reviewers, the ones who catch the most are the ones who use this technology every day.

Why testers with disabilities, specifically

It helps to distinguish three audit modes, because vendors blur the language.

Automated scan. What an axe-core-based scanner can detect — roughly 60 to 70% of WCAG-testable issues, at near-zero marginal cost. Your baseline, not your audit.

Manual audit by a sighted accessibility specialist. A trained reviewer walks through the site, exercises every interactive component with keyboard and screen reader, judges contrast in context, and reasons about ARIA semantics. Fills in most of the technical gap — focus order, keyboard traps, ARIA misuse, conformance-as-written against the WCAG 2.2 success criteria.

Manual audit by people with disabilities (PWD testing). The specialist coverage plus a dimension the first two modes structurally cannot produce: lived experience. Is the alt text actually useful? Does the screen reader read the form in an order that lets you complete it? Does the keyboard navigation feel usable after 50 tab presses? Is the magnified version legible at 400% zoom?

The third mode catches usability failures masked as conformance passes. A button labelled “Click here” passes 2.4.4 with sufficient context, but a screen-reader user navigating by link list will hear “Click here, click here, click here” and give up.

”Nothing about us without us” is not a slogan in accessibility work — it is a methodological position. The people who experience the barriers are the only ones who can definitively say whether a fix removed them.

What a manual audit actually delivers

A manual audit is a defined deliverable, not an open-ended consulting engagement. Before you sign a statement of work, you should expect each of the following.

A scoped page set. You pick 20 to 50 representative pages, journeys, or screens with the auditor’s input. The auditor does not crawl your entire site. If a firm offers to audit “your whole site” for a fixed price on a 10,000-page property, that is marketing, not methodology.

WCAG 2.2 AA conformance findings. A register mapping each issue to a specific success criterion. On a moderately mature site, expect 80 to 200 findings in a medium-scope audit.

Severity rating per finding. Blocker / major / minor. Severity is what turns a 200-row spreadsheet into an actionable backlog.

User-impact narrative. Not “1.4.3 contrast fail at #888 on #fff” but “screen-reader users cannot tell required form fields apart because the asterisk is low-contrast grey and is not announced.” User-impact phrasing is what gets tickets prioritised.

Recommended remediation. Specific enough to act on. “Add aria-required='true' and include ‘required’ in the visible label” beats “fix contrast issue.”

Executive summary. A two-to-three-page document a senior reader can absorb in fifteen minutes: conformance verdict, top five risks, remediation effort estimate.

Optional: published accessibility statement. If your audit feeds into a public-facing statement, see our accessibility statement audit for examples of what good statements look like.

Re-test cycle. A scheduled re-test 3 to 6 months after delivery to verify the fixes actually fix the issues. Without this, you’ve bought a snapshot, not a programme.

What manual audits cost — honest ranges

Vendors are coy with public pricing, because every engagement is scoped per-customer. The ranges below are realistic 2026 ballparks based on conversations with buyers and observed contract values.

Small audit — 5 to 10 pages, single-product SPA, 1 to 2 weeks turnaround, a single tester or a small team without PDF or mobile scope: $5,000 to $15,000.

Medium audit — 20 to 50 pages, multiple user journeys, 4 to 6 weeks, a multi-disability tester panel, written report, one round of remediation Q&A: $15,000 to $50,000.

Enterprise audit — 100+ pages, mobile native apps (iOS and Android), multiple authenticated flows, multi-language scope, 8 to 12 weeks, multiple-tester panel, formal report, executive briefing: $50,000 to $250,000+.

Re-test — after remediation, scoped to the same page set: typically 30 to 40% of the original cost.

Factors that push price up: PDF accessibility scope (labour-intensive, often quoted separately), mobile native apps, SPAs with authenticated flows, multi-language sites, urgent legal-defence framing, and any requirement for expert testimony.

Factors that pull price down: a stable static site, tight scope, prior automated-scan output the auditor can use, and a multi-engagement relationship.

Do not expect transparent menu pricing. Every vendor scopes per engagement. A flat per-page rate quoted without seeing your product is a yellow flag, not a feature.

How to choose an audit company — buying criteria

Use these as a shortlist filter when you put a request for proposals out to three or four firms.

Tester composition. Non-negotiable. Ask for the disability profile of the testers on your engagement — screen-reader, low-vision, motor-impairment, cognitive, and deaf-or-hard-of-hearing testers. If the firm cannot answer, this is the wrong firm.

WCAG version. WCAG 2.2 should be the floor in 2026. Some firms still default to 2.1 — push for 2.2 AA explicitly in the SOW.

Audit framework. Does the firm use an established methodology — WCAG-EM, Trusted Tester, or a published in-house equivalent — or are they making it up per engagement? A published methodology signals operational maturity.

Sample-size methodology. The SOW should specify how pages were chosen and what is in versus out of scope.

Re-test policy. Included as a fixed deliverable, or charged separately? Six months is reasonable; “indefinite” is marketing.

Mobile and PDF scope. Usually quoted as add-ons. Get them in scope from the start — bolting them on later costs more.

Reporting format. Developer-facing JIRA imports, executive PDF, legal-defence affidavit-ready — different audiences, different formats. Ask up front.

Manual-audit handoff. Does the firm integrate findings into a monitoring platform, or hand you a static PDF that ages in a SharePoint folder? Integration is worth real money.

Geographic and linguistic coverage. Don’t assume a US firm has EAA expertise, or that a European firm has worked on ADA Title III defence.

References from peer organisations. Ask for three references in your industry. Sector expertise takes years to accumulate.

A fair shortlist of four firms worth requesting proposals from, with one specialist worth knowing about.

Qualibooth combines automated scanning, continuous monitoring, and manual audits by testers with disabilities in a single integrated platform. The workflow runs from scan → triage → manual verification → published accessibility statement, with findings persisting in a dashboard rather than a static report. This is the clearest fit when you want one vendor for both ongoing monitoring and scheduled manual audits, and when you’d rather not stitch two tools together. Caveat: a younger platform than the long-established US specialists, so longer enterprise references are still building. Qualibooth.

Deque (axe DevTools and Deque Systems audits) is engineering-led, with the deepest accessibility-engineering bench in the market and the strongest tooling integration story — axe is the de facto rule engine across the industry. Best fit when you want auditors who can sit with your engineers and pair-program remediation. Caveat: pricing reflects the brand, and the tooling-first culture can feel less PWD-centric than some competitors.

Level Access is a long-running specialist firm with strong enterprise positioning, broad jurisdictional coverage, and well-established US legal-defence experience. Best fit when ADA Title III litigation risk is the primary driver and you need a firm that has been deposed before. Caveat: large-firm pricing and a process orientation that some smaller buyers find heavy.

TPGi (formerly The Paciello Group) has deep heritage in WCAG work going back to the standard’s earliest drafts, JAWS integration through parent company Vispero, and unusually strong assistive-technology expertise. Best fit when screen-reader behaviour is the central concern. Caveat: smaller scale than Deque or Level Access for global mobile-app engagements.

Fable specialises in remote user testing by people with disabilities — usability research, not WCAG-conformance audits. Best fit as a complement to a conformance audit, when you want lived-experience feedback on a flow but don’t need a 200-finding conformance register. Caveat: not a substitute for a WCAG audit; complementary to one.

In practice, most mature buyers shortlist three of these firms, scope the work tightly, and choose based on tester composition and fit rather than price.

Common pitfalls in buying an audit

Treating the audit as a compliance certificate. It is a snapshot of one defined scope at one moment. It reduces risk and produces an evidence trail. It is not a license.

Scoping too narrowly. A 10-page audit on a 10,000-page site is a small sample. Budget for a scope that represents your real product surface.

Not budgeting for remediation. The audit reveals the cost of your accessibility debt — usually five to twenty times the audit fee. Without remediation budget, you’ve bought a shelf document.

Skipping the re-test. Without verification, you don’t know whether your fixes worked. Some did, some didn’t, some introduced new issues.

Hiring a firm with no testers with disabilities. The most common failure mode. Ask for the tester composition for your specific engagement, not the firm’s marketing-page averages.

Confusing automated-scan-as-a-service with a manual audit. Several vendors sell scan output dressed up as an audit. If it arrived in 48 hours and costs under $3,000, it is a scan report.

Buying the audit but not changing the process. The findings will return on the next release if design, engineering, and QA don’t have accessibility built in. The audit is a diagnostic; the cure is process.

Frequently asked questions

How much does an accessibility audit cost?
Realistic 2026 ranges: a small audit covering 5–10 pages of a single SPA runs roughly $5,000–$15,000; a medium audit of 20–50 pages with multiple journeys is $15,000–$50,000; and an enterprise engagement covering 100+ pages, mobile apps, and several flows can land anywhere from $50,000 to $250,000 or more. Re-tests after remediation are usually 30–40% of the original cost.

How long does a manual accessibility audit take?
Most engagements run 2–6 weeks for small-to-medium scopes and 8–12 weeks for enterprise audits that include mobile apps, PDFs, or multiple language variants. The fieldwork itself is rarely the bottleneck — scheduling testers with the right assistive-technology profiles and writing the report take the bulk of the calendar time.

What’s the difference between automated and manual accessibility audits?
Automated scanners detect roughly 57–70% of programmatically testable WCAG violations — missing alt attributes, low colour contrast, form labels, landmark structure. Manual audits cover the remaining 30–40%: focus order, keyboard traps, meaningful alt text quality, screen-reader legibility, dynamic-content announcements, and other issues that require human judgement. Manual audits performed by testers with disabilities add a third layer — lived usability, not just technical conformance.

Do all accessibility audit firms employ testers with disabilities?
No. Many firms run audits exclusively with sighted accessibility specialists who emulate assistive technology. Reputable disability-led audits include testers who use screen readers, magnification, switch input, or voice control daily, across a range of disabilities. Ask any prospective vendor for their tester composition before signing.

How often should we re-audit?
A full manual audit is typically valid for 12 months on a stable product, or until a major release changes the audited flows. Most mature teams pair an annual manual audit with continuous automated monitoring between audits, and run a targeted re-test 3–6 months after remediation to verify fixes.

Can a manual audit guarantee ADA or EAA compliance?
No reputable audit firm offers a compliance guarantee. An audit is a snapshot of conformance against a standard like WCAG 2.2 AA on a defined scope at a moment in time. It reduces legal risk and produces an evidence trail, but neither the ADA nor the EAA recognises a private certification as binding.

What deliverables come with a manual audit?
A typical engagement delivers: a defined page set, a findings register against WCAG 2.2 AA with severity ratings, user-impact narratives per finding, recommended remediation, an executive summary, optional published accessibility statement, and a re-test cycle 3–6 months out. Some vendors also hand off findings into a monitoring platform; others deliver a static PDF.

Where to go from here

For a quick baseline before you brief any vendor, run our free accessibility scanner on your highest-traffic pages — the output will give you the obvious 60% before an auditor begins. For ongoing observation between audits, our monitoring buyer’s guide compares the platforms that handle the continuous half of the problem.

And before committing to a single firm, request scoped proposals from at least two or three of the audit providers on the list above — pricing varies more than the marketing pages suggest, and tester composition varies even more.