How to choose an accessibility agency: what to ask, and what the market actually looks like

Most advice about hiring an accessibility agency is written by accessibility agencies. That is not a conspiracy, it is a structural problem: a vendor cannot credibly tell you when to pick a competitor, so the genuinely useful questions — who is even eligible, how thin is the field, which claims mean anything — never get asked in public.

We maintain a catalogue of 574 agencies across 39 countries, every entry with a cited source. That lets us answer those questions with counts rather than opinion. The numbers below are measurements of the catalogue as of August 2026, not a market forecast.

The field is narrower than it looks

Nearly every agency offers an audit — 545 of 574. That is the commodity. The specialisms are where the field thins out fast:

Service	Agencies offering it	Share
Accessibility audit	545	95%
Consulting	378	66%
Training	287	50%
Remediation	270	47%
Monitoring	66	11%
VPAT / ACR	45	8%

Two of these rows should change how you shortlist.

VPAT work is rare. If you sell into US federal agencies, state governments or universities, procurement will ask for a VPAT or an Accessibility Conformance Report. Forty-five agencies list that capability. Lead with the requirement, not with the budget — starting from price and discovering at the end that your preferred vendor has never produced an ACR wastes a whole procurement cycle.

Almost nobody stays after the audit. Sixty-six agencies list monitoring. Accessibility is not a defect you fix once: every deploy, every CMS edit and every third-party embed can reintroduce failures. If the engagement ends when the PDF arrives, budget separately for who watches the site afterwards — whether that is a monitoring retainer, your own CI checks, or both.

Geography is a real constraint, not a formality

Of 574 agencies, 536 list a single country of operation. Thirty-eight list two or more, and the broadest coverage in the catalogue reaches four countries.

That matters because accessibility law is national even when the standard is shared. The European Accessibility Act is a directive that each member state transposed separately, with its own enforcement body, its own deadlines and its own language requirements for the accessibility statement. Germany's BFSG, France's RGAA and Italy's Legge Stanca are not interchangeable, and neither are the auditors who work under them.

So a vendor offering pan-European delivery is almost certainly coordinating partners. That is a normal and often sensible arrangement. Ask three things anyway: who performs the testing in each market, who signs the report, and who is liable if a regulator disputes the conclusion.

Standard claims are uneven — filter on the one that binds you

What agencies say they work against, across the catalogue:

Standard	Agencies claiming it
WCAG 2.2	423
EN 301 549	162
European Accessibility Act	129
ADA	79
Section 508	72
BITV	62

WCAG 2.2 is close to universal, which makes it useless as a filter. The others are not. If your obligation comes from Section 508, you are choosing among 72 rather than 574. If it comes from the German BFSG, 62. Work out which instrument actually binds you before you shortlist, because that single question removes most of the market for you.

The credential that means something, and the two that mean less

Credentials in this market are a mess of things that sound alike and prove very different amounts. Three markers appear in our catalogue, and they are not equivalent.

Named as the auditor in someone else's accessibility statement — 96 agencies. This is the strongest signal available, and it is not a certificate. Public bodies across the EU are obliged to publish an accessibility declaration, and many name the firm that carried out the assessment. Ninety-six agencies appear that way: 83 named by a public body and 13 by a private organisation, across Poland (26), France (25), Germany (18), the Netherlands (15), Belgium, Austria, Spain and Italy. Every one of the 96 carries a link to the declaration itself, and every link was opened and checked rather than taken on trust — a record pointing at the agency's own website is rejected at build time, because a claim about yourself is not evidence. It is a third party stating on the public record that this firm did the work.

BITV-Prüfstelle accreditation — 20 agencies. A German accreditation for BITV testing bodies, checkable against the public register. Meaningful within its scope, which is German BITV testing and nothing else.

IAAP organisational membership — 12 agencies. A company subscription. It signals engagement with the profession and says nothing about whether anyone staffed on your project has passed an exam.

Individual credentials — CPACC, WAS, CPWA — belong to people, not companies. An agency can hold none of the three markers above and still put a certified specialist on your project, or hold all of them and staff you with a junior. Ask who specifically is assigned and what they hold.

The general point survives beyond our catalogue: when a directory shows a credential, ask what was checked and by whom. A field in a database is not evidence; a link to a public declaration that names the auditor is.

The questions that actually separate vendors

Once the field is narrowed by service, geography and standard, most remaining differences come down to method. Five questions surface them:

1. How many hours of manual testing are in this quote? Automated tooling reliably detects roughly a third of WCAG failures. Keyboard traps, focus order, meaningful alt text, error recovery and screen reader announcement order require a person. A quote that does not separate automated from manual hours is hiding its own shape.
2. Which assistive technology, on which platform? "We test with screen readers" is not an answer. NVDA on Windows, JAWS on Windows and VoiceOver on macOS and iOS behave differently, and a bug reproduced in one may not exist in another.
3. Do disabled people test the product? Lived experience finds problems that conformance checklists pass. Ask whether testers are employed, contracted, or absent.
4. What does the deliverable look like? Ask for a redacted sample report before signing. You are looking for reproduction steps, affected user groups and suggested remediation — not a list of rule IDs exported from a scanner.
5. What happens on re-test? Confirm whether verification of fixes is included, time-boxed or billed again. This is the single most common source of budget surprise.

What to do before you contact anyone

Run a scan of your own site first. Not because an automated result decides anything — it does not — but because arriving with a concrete list of failures changes the conversation from a sales pitch into a scoping discussion, and it lets you compare quotes against the same known baseline.

Then shortlist on the constraints that actually eliminate: the standard that binds you, the country you operate in, whether you need a VPAT, and whether you need someone after the audit as well as during it.