The checks you can run right now, free.

With a free account, scan up to ten of your pages, each loaded in a real browser and checked against all 90 supported automated rules. Findings include rule evidence and a fix guide when a verified match is available. The report keeps 27 best-practice checks separate from the 63 WCAG-mapped checks, which touch only 20 of the 55 criteria at Level A and AA.

Every WCAG reference is organized by rule number, which is useful once you know the number. People arrive holding a symptom instead, so describe what is going wrong in the words you would say out loud and this searches the documented failures behind all 86 criteria to find where it belongs.

Two questions, where you sell and what kind of organization you are, narrow 38 accessibility laws down to the handful that can reach you. Treat the result as a filter rather than an answer, because each law page carries the scope, the size thresholds and the dates that actually decide it.

Paste a component and check it against the ARIA model. It finds invalid roles, the abstract ones no browser implements, roles missing a state they oblige you to supply, roles sitting outside the parent they need, and attributes that do not belong on the role you used. It reads markup, so it tells you your ARIA is valid rather than that it is true.

Pick a text color and a background and read the ratio against all four thresholds together. 4.5 to 1 for body text at AA, 3 to 1 for large text, 3 to 1 again for borders, icons and focus rings under the separate non-text rule, and 7 to 1 at AAA. A live preview shows you the pair at the sizes those thresholds are about.

One brand color in, a light-to-dark ladder out, every step measured against white text and against dark text. The tick marks the 4.5 to 1 bar that ordinary body copy has to clear. A shade that misses it is not ruled out, because large text and interface parts need only 3 to 1, and the table says which shades those are.

Enter a page and read its heading outline the way the source order hands it over, with a missing h1, skipped levels and empty headings all flagged. It also tells you which of those a conformance claim actually turns on, which is fewer than most heading tools let on.

Screen reader users often pull up a list of a page's links and read it with none of the surrounding text, which turns twelve links all saying "read more" into twelve identical choices. This builds you that list. Read it as the stricter AAA rule, because at AA the card or paragraph around a link is allowed to supply its purpose.

Paste your alt text and it finds the mechanical failures first. Filenames read out letter by letter, the "image of" prefix heard twice, the single word that names a format instead of a picture. It reads only the text you paste and never your markup, so whether those words serve the image is still yours to judge.

Decorative, functional, words inside a picture, or a chart. Up to four questions land you on the approach that fits, with the code to go with it, and with the cases where the right answer is an empty alt spelled out, because that answer comes up far more often than people expect.

Drop in a PDF and find out in about ten seconds whether it carries a tag tree, whether it holds real text or a photograph of text, and whether the document title and language are set. It runs in your browser and the file is never uploaded anywhere.

The US Disabled Access Credit covers half of qualifying accessibility spending between $250 and $10,250, which caps the credit at $5,000 a year. This checks your business against the size test and does the arithmetic. Whether website work is qualifying spending is genuinely unsettled, so what comes back is an estimate rather than tax advice.

Fill in a short form and get a plain-English accessibility statement to copy or download, free. It is a starting draft rather than a finished document, and it says so, because the conformance status, the known limitations and the enforcement body all have to come from you.

A brief that asks for a page count tells a supplier nothing, which is why two quotes for one site come back ten times apart. Name your views, your journeys and the states they run in, and this writes the scope instead: a structured sample, a random set at a tenth of it on top, and the sentence saying a sample cannot answer for pages nobody opened.

Audits go slowly for boring reasons: nobody could reach the signed-in pages, the build changed under the tester, the findings landed the week before launch. Ten questions about the arrangements, including the access ones, which take longest and get left last. No credential fields anywhere, because whether a test account exists is a yes or no question.

A WCAG conformance claim has five required components and most drafts carry three. This walks all five, then the scope conditions underneath them: whether part of a page was carved out, whether a checkout is only half covered, and what the two routes are for content you do not control. It checks the components, never whether the claim is true.

A supplier sent you a conformance report and almost every row says Supports. Nine questions about the document's own rules, answered with it open beside you: which edition and so which WCAG version, whether Not Evaluated appears where the template forbids it, and what is missing by design. Nothing is uploaded, and there is no score, because counting rows is not a comparison.