WCAG audit and remediation

WCAG audit, the fixes, and the re-test

We audit what you already run, write the fixes ourselves, re-test the same criteria and hand you a dated document naming the views tested. Nobody certifies any of this.

Start with the scope

Tell us what has to be assessed and by when. We come back with the scope and the shape of the work.

How we handle what you send is set out in our privacy notice.

Read what the quotes in front of you actually commit to

Of the sixteen pages we read, three say the provider writes the fix, four say a re-test happens, and one mentions an attestation of any kind. None say all three, and fifteen do not say what the evidence is worth after the next release. Sixteen do not say who signs the document you receive. Three questions settle most of it: does the work include writing the fix, does it include a re-test of the same criteria, and does it name who signs.

3 of 16say the provider writes the fix

Atyantik review of sixteen provider pages ranking for this work

4 of 16say a re-test follows the fixes

Atyantik review of sixteen provider pages ranking for this work

1 of 16mentions an attestation of any kind

Atyantik review of sixteen provider pages ranking for this work

0 of 16say who signs the document

Atyantik review of sixteen provider pages ranking for this work

Every one of these is signed by somebody

For most of them, the signer is the party being assessed. That changes what each one is worth to whoever asked you for it.

The eight documents people mean when they ask for an accessibility report, and what each one is signed by
DocumentWho signs itIs it a certificationWhat it cannot show
VPAT, the blank templateThe product owner, as a self-declarationNo. ITI, which publishes the template, states there is no certification for VPATA blank template shows nothing, and a completed one shows only what its author asserts
Accessibility conformance report, the completed documentThe vendor whose product it describesNo. No federal agency signs, approves or registers oneThat the product is accessible. Only that the vendor says so
WCAG 2.2 conformance claimWhoever publishes the content, and they must date itNo. W3C publishes the rule and reviews no claim made under itAnything outside the pages listed and the technologies named in the claim
Third-party audit report on W3C methodologyThe evaluating firm, in its own nameNo. W3C publishes the method and issues nothingConformance, because it samples. W3C: in the majority of situations, using this methodology alone does not result in being able to make WCAG 2 conformance claims
Remediation planThe vendor, as a commitment to the purchaserNo. There is no issuing body for itThat any defect has been fixed. Only a re-test shows that
Accessibility statement on the siteThe organisation, publicly, at a dated momentNo. W3C WAI publishes the guidance and approves nothingEvidence. It is a public commitment plus a disclosure of known gaps
Letter of validation after remediationThe auditing organisation, only after a re-test, and it cannot be self-issuedNo, and it does not claim to be. It states its own scope on its faceAnything outside the screens examined, or anything after the date of the re-test
Conformance badge in a footerNobody. It carries no author, no date and no scopeNo. W3C: these logos do not represent review or validation of conformance by W3C and/or WAIEverything it looks like it shows

What the UK Government Digital Service found when it re-audited 47 sites

These are public sector organisations under a legal duty to be accessible, so this is the optimistic end of the range. GDS audited 47 sites, then audited them again two to four years later. Conformance moved in both directions, and GDS gives a staffing explanation: the knowledge can disappear as team members leave or change roles.

Show data table
47 public sector websites audited twice, two to four years apart, by the UK Government Digital Service. Not Atyantik data.
Passed the re-audit Failed the re-audit
Passed the first audit 17 sites 6 sites
Failed the first audit 6 sites 18 sites

Six of the 23 sites that had passed lost it by the second audit, and six that had failed reached a pass.

Sites 47 public sector websites audited twice, two to four years apart, by the UK Government Digital Service. Not Atyantik data. UK Government Digital Service, re-audit of 47 public sector websites

The loop, and why it has these five parts

Across five agreements the US Department of Justice reached with companies over inaccessible websites, every one of the five names a WCAG version and level, carries a dated deadline, requires an outside consultant, and requires testing that carries on past the first audit.

What happens, in order

  1. Scope, agreed in writing

    We settle which views, templates, documents and apps are in, and name the WCAG version and level we test against, before anyone tests anything.

  2. We run the audit

    We test the agreed views against the named criteria, by hand and with tools, recording what was examined and what failed.

  3. We write the fixes

    We make the changes in your codebase. You do not get a list telling your team to rewrite code that works.

  4. Re-test of the same criteria

    We test the same criteria on the same views again, so what you hold is a result rather than an intention.

  5. The dated document

    On its face it names the views tested and the WCAG version and level, gives the date, lists the technologies relied upon and the re-test result.

What our re-test is not

One of those Department of Justice agreements requires the company to retain an independent web accessibility consultant approved by all parties. Independent there means someone who is not the party doing the remediation. We fix, and we re-test our own fixes. That is not independence, and we will not describe it as though it were. If your situation requires an independent verifier, that is a third party, and what we produce is built so that party can check it rather than start again.

A closed loop, or an independent verifier

Do you need a verifier who is not the party doing the remediation, or do you need the failures on what already ships cleared by one accountable firm?

  • One firm, closed loop

    When this is the one

    Nobody outside has named a verifier. You need the failures cleared on what already ships, and evidence of what was tested.

    What that means in practice

    The audit, the fixes and the re-test all come from us, and the re-test is ours. One accountable party, with the limit above stated on the document.

  • An independent verifier

    When this is the one

    Verification by an outside party has been required by a decree, a signed settlement, a regulator or a large customer.

    What that means in practice

    A third party runs the verification. We do the remediation and hand over the tested evidence in a form that party can check rather than repeat.

What you can send onward

Where a customer asks for an accessibility conformance report, that document is signed by the product owner, which is you. We hand over the tested evidence behind it. VPAT is a registered trademark of the Information Technology Industry Council, and nothing we produce is called one.

What is in the handover pack, what each item states, and what it does not
In the packWhat it statesWhat it does not state
Audit findings, view by viewEvery view we tested, the criteria applied, and each failure with the place in the product it livesAnything about views outside the agreed scope
The fixes, written in your codebaseThe change that was made and where it was madeThat nothing new arrives with your next release
Re-test resultThe same criteria re-tested on the same views, with the date it was doneA status that holds after that date
Scope and method statementThe tested views on the face of the document, the WCAG version and level, and the technologies relied uponA certificate. Nobody issues one for this
Evidence behind the report your customer asked forThe tested material a product owner needs to complete and sign that reportA legal opinion, and not a prediction that no letter arrives

Where the failures sit, by Deque's own rule set

Deque publishes the axe testing engine, and it has 100 non-deprecated rules. Sorted by which layer owns the markup that fails them, most of them land in one place. In the US Web Design System's own self-assessment, all ten disclosed WCAG A and AA exceptions name a component as the defect site rather than the markup of a team consuming it.

Show data table
Deque's 100 non-deprecated axe rules, split by the layer that owns the failing markup. The counts come from Deque Systems, which publishes the rule set, and they are not Atyantik measurements.
Segment Value (rules) Share
Markup an application team wrote 61 61%
ARIA emitted by a component library 33 33%
Third-party embeds 6 6%

Most of it is in markup your own team owns, which is where we write the change.

Deque's 100 non-deprecated axe rules by owning layer Deque's 100 non-deprecated axe rules, split by the layer that owns the failing markup. The counts come from Deque Systems, which publishes the rule set, and they are not Atyantik measurements. Deque Systems, axe-core non-deprecated rule set

What public agencies paid, from United States federal award records

None of these are our numbers. They are medians from 425 United States federal contract awards where accessibility is the stated purpose, grouped by what was bought, every one an award to a third-party vendor.

Show data table
Median award value by scope class across 425 United States federal contract awards where accessibility is the primary purpose. Public awards to third-party vendors, not an Atyantik figure.
Item Median award value
Documents only (n=41) 39,397 USD
One product or website (n=23) 66,515 USD
Scanning tool licence alone (n=65) 82,980 USD
General Section 508 services (n=199) 93,619 USD
Training content (n=10) 189,336 USD
Department-wide programme (n=87) 627,864 USD

Scope class moves the figure by more than an order of magnitude, which is why no single median across classes is shown.

Median award, USD Median award value by scope class across 425 United States federal contract awards where accessibility is the primary purpose. Public awards to third-party vendors, not an Atyantik figure. United States federal contract award records, 425 awards, n printed per class

What moves the number here

Five scope questions you can answer before speaking to anyone. How many distinct templates and screens are in scope, rather than how many pages. Whether PDFs and other documents are included. Whether native apps are included. Whether the report has to be produced in a format your customer names. And whether a re-test of the same criteria is part of the work.

Whether to call us

  • You have a live product and someone outside the company has asked you for evidence about it.

  • You want the failures cleared on what already ships, without changing how the team builds.

  • Your team will keep releasing while this runs, and the evidence has to survive that.

Do not buy this when

  • It is not built yet, and you want it designed and tested to the standard from the start.

    Build it in from the start

  • You want somebody to hold the level over the following year rather than reach it once.

    Ongoing support

  • What landed on your desk is a security questionnaire, and not an accessibility one.

    Security

  • You want the whole experience rethought, not an existing one brought to a stated level.

    UX design

The six questions we get asked

Nothing here is legal advice, and none of it is a statement about your own exposure.

We already have an accessibility widget running on the site. Isn't that covered?
No, and the reason is not our opinion.
  • The United States Federal Trade Commission's complaint in its case against accessiBe states, citing W3C and Utah State University's Institute for Disability Research, developer of the WAVE tool, that no automated testing tool alone can determine if a website meets accessibility standards, and that manual human testing is required.
  • About that one company: the FTC alleged its compliance claim was deceptive.
  • accessiBe agreed to an order, neither admitting nor denying the allegations, and agreed to pay USD 1,000,000 in monetary relief usable for consumer redress.
  • The final order was issued on 21 April 2025 and bars that company from representing that its automated technology can make any website WCAG compliant.
  • That order binds one company on its own marketing claims.
Is what you hand over a certification, or not?
It is not, and nobody can give you one.
  • The Information Technology Industry Council, which publishes the VPAT template, states plainly that there is no certification for VPAT, and that the product owner completes it as a self-declaration.
  • W3C says its own conformance logos do not represent review or validation of conformance by W3C or WAI.
  • So what you get here is a dated document naming the views we tested, the WCAG version and level, the technologies relied upon and the re-test result. It is evidence with its scope written on its face.
Will this stop the letters and the lawsuits?
No. We will not say otherwise to win the work. Nothing here is legal advice and nothing here is a statement about your own exposure. What we can do is clear the failures on what we tested, write the fixes into your codebase, re-test the same criteria, and give you the record of what was tested, when, and against which criteria.
What does it actually cost, and what makes it go up?
Scope decides it. For a sense of the range, from 425 United States federal awards where accessibility is the stated purpose, the median for one product or website was USD 66,515 across 23 awards, and for a department-wide programme USD 627,864 across 87. Those are public awards to other vendors, not our figures. What moves it for you is the five scope questions above.
Are you going to make my team rewrite code that works fine?
Mostly not, and the shape of the failures says why. Of Deque's 100 non-deprecated axe rules, 61 sit in markup an application team wrote, 33 come out of component libraries as emitted ARIA, and 6 belong to third-party embeds. We write the change ourselves rather than hand back instructions. Where the defect is inside a library you depend on, we work around it in your own code.
What is any of this worth after next month's release?
It moves, and we would rather say so now. The UK Government Digital Service re-audited 47 public sector websites two to four years apart. Six of the 23 that had passed no longer passed. So what you hold is a statement about a date, for the views listed on it. If holding the level over a year matters more than reaching it once, say that at the scoping stage, because it is a different arrangement.

Tell us what has to be assessed

Tell us what has to be assessed, by when, and who asked you for it. We come back with the scope, the shape of the work and whether it is ours to do.

Two good ways this ends. Either we agree a scope and get to work, or we tell you straight that this is not ours.

The founder who replies to enquiries
Tirth BodawalaFounder, Atyantik Technologies
  • What you send stays between us.
  • A senior software engineer reads it, not a sales desk.
  • Nothing above is held back behind an email address.

Tell us what has to be assessed

How we handle what you send is set out in our privacy notice.