Website & Document Accessibility Review
Your website, page by page. Your documents, file by file.
Enter your address and we find your pages from your own sitemap, audit as many as your plan covers against the exact standard your law references, pin each finding on a screenshot, status-check your links and hand you a client-ready PDF. Then every PDF, deck, Word document, workbook, image and SVG you upload gets its own document check — with a plain list of what was tested, what was not, and what still needs a person.
Text must have sufficient color contrast
Measured from the rendered pixels: 2.1:1 against a required 4.5:1.
ARIA attributes must conform to valid values
4 elements affected · fix guidance linked.
Automated checks test part of WCAG — a clean result is a floor, not a statement that the standard is met, and anything needing human review says so.
Free accessibility check
Check your homepage for accessibility first.
One scan. The whole accessibility workflow.
Not another single-page checker — a full audit workflow, from discovery to the statement you publish.
Whole-site scans
We read your sitemap (or homepage links) and audit the pages we find, up to your plan per-scan limit — 25 on trial, 150 Starter, 500 Professional, 2,000 Agency — with axe-core running in a real browser. If the site is larger, the report says so.
Findings pinned on your pages
Every violating element is outlined and numbered on a screenshot of the actual page — you see which button, which image, which heading.
Menus, dialogs and the keyboard
Most scanners audit the page as it loads. Ours also opens the menus, dialogs and expandable panels it finds and re-audits each state, then walks every page with the Tab key — catching focus that vanishes into hidden elements, keyboard traps and missing focus indicators.
Linked PDFs checked too
Documents on your site are web content under the ADA Title II rule and EN 301 549. Every linked PDF is checked for structure tagging, a title, a language, alt text on figures and a readable text layer — the basics assistive technology depends on.
Broken-link report
Links across your scanned pages are status-checked, the most widespread first, up to 250 unique links per scan — and the report shows how many of the links found were checked. Dead ends frustrate everyone, and worst of all people using assistive technology.
Guided manual checks
Automation catches roughly a third of WCAG. Our 22 plain-language checks walk your team through the rest — keyboard, forms, media, behaviour — with who-checked-what recorded.
Documents checked too
Every PDF, PowerPoint, Word, Excel, JPEG, PNG and SVG uploaded for a brand review gets its own automated document check against EN 301 549 clause 10 and PDF/UA: tagging, titles, language, alt text, headings, tables, contrast, text baked into images. Each result opens with what was checked, what was not, and what still needs a person.
Tickets your developer can act on
One ticket per failing rule — pages, exact elements, WCAG criterion and a plain-language fix — exported as a CSV for Jira, Linear, Asana or a spreadsheet, or as ready-to-paste GitHub issues. Nobody translates the report.
Scheduled re-scans and alerts
Weekly or monthly automatic re-audits, with an email or Slack message showing exactly what changed: violations fixed, new issues appeared, rules resolved.
Evidence of effort, over time
A chart of every scan of the site, and an evidence pack: dated results, what changed between scans, the manual checks with who judged them, the statement and the report — the good-faith record accessibility law rewards.
PDF reports, a statement, your branding
A client-ready PDF of every scan, a read-only link to share with a client, and an honest accessibility statement generator. On the Agency plan, all three carry your logo, name and colour.
Accessibility laws are not generic. Neither are our checks.
Every accessibility law references a specific WCAG version — the US ADA rule points at WCAG 2.1 AA, Section 508 at 2.0 AA, the European Accessibility Act at EN 301 549. Pick your jurisdiction and the audit runs that exact rule set, and the report is stamped with the standard it tested.
| Jurisdiction | Law | Standard we test |
|---|---|---|
| International | Latest W3C recommendation | WCAG 2.2 AA |
| USA — state & local government, businesses | ADA (DOJ Title II rule; Title III case law) | WCAG 2.1 AA |
| USA — federal agencies & suppliers | Section 508 | WCAG 2.0 AA + Section 508 rules |
| European Union | European Accessibility Act / EN 301 549 | WCAG 2.1 AA + EN 301 549 rules |
| United Kingdom | Public Sector Bodies Accessibility Regulations · Equality Act | WCAG 2.2 AA |
| Canada | AODA (Ontario) · Accessible Canada Act | WCAG 2.0 AA |
| Australia | Disability Discrimination Act (AHRC guidance) | WCAG 2.2 AA |
Laws and deadlines evolve — the picker is kept in step with what each law currently references. Testing against a standard is what we do; declaring that a site satisfies one is a judgement only a full audit can make.
Document accessibility
The files nobody has measured. Checked file by file, on the same upload.
The brochure, the deck, the social graphic: the assets that drift from your brand are the same ones that fail assistive technology, and nobody in the organisation owns either question. Every file uploaded for a Brand Review now gets a second answer — a document accessibility result with a specific list of what was tested. Here is that list.
Read with a PDF parser: the structure tree, metadata, viewer preferences, outline, page operators, form fields and annotations are inspected directly. 22 criteria.
A machine tests
- Tagged — has a structure tree at all (a PDF without one is capped near zero, and the report says why)
- Title set in the metadata, and shown in the window instead of the file name
- Primary language declared
- Figures carry alternate descriptions; decorative images are marked as artifacts
- Heading levels present and not skipped
- Tables have header cells
- Reading order defined in the structure tree
- Bookmarks present for documents over 20 pages
- Real text, not a scanned image — an image-only PDF returns “could not be read”, not a number
- Form fields have labels and tooltips
- Links have descriptive text
- Security settings allow assistive technology to read the file
A person judges — listed on every result, never scored
- Alternate descriptions are meaningful
- Heading structure is logical
- Reading order is sensible
- No information conveyed by colour alone
- Changes of language are marked
PowerPoint (.pptx)
The package is opened and every slide, placeholder, shape, picture, table, link, media object and note is read from its XML, including the theme colours text actually inherits. 17 criteria.
A machine tests
- Every slide has a populated title, and titles are unique across the deck
- Shape reading order follows the visual order
- Images, charts, SmartArt and groups have alt text — and it is not a file name or “Picture 4”
- Tables have a header row
- Text contrast against the slide background, using the real theme colours
- Body text is not below the minimum size
- No text baked into images (OCR on every picture)
- Hyperlinks have descriptive text
- Document language set
A person judges — listed on every result, never scored
- Reading order makes sense
- Alt text conveys the purpose of the image
- Text placed over an image is readable
- Embedded video and audio have captions
- Notes do not carry meaning missing from the slide
- No information conveyed by colour alone
Word (.docx)
The document package is opened and styles, paragraphs, images, tables, lists, links and language settings are read from its XML. 16 criteria.
A machine tests
- Real heading styles used, not bold text pretending to be headings — and levels are sequential
- Images and charts have alt text, and it is not a file name
- Tables have a designated header row, with no merged cells in it
- Document language set
- Text contrast
- Lists built with list styles, not typed dashes and numbers
- Hyperlinks have descriptive text
- No layout built with tabs, spaces or empty paragraphs
- Document title set
A person judges — listed on every result, never scored
- Heading structure is logical
- Alt text conveys the purpose of the image
- Foreign-language passages marked
- No information conveyed by colour alone
Excel (.xlsx)
The workbook package is opened and sheets, data regions, cell styles, drawings and properties are read from its XML. 8 criteria.
A machine tests
- Sheet tabs named meaningfully, not “Sheet1”
- Data ranges have a header row
- Images and charts have alt text
- No blank rows or columns splitting a data region
- Cell text contrast, from the actual cell styles
- No merged cells inside data regions
- Workbook title set
A person judges — listed on every result, never scored
- No information conveyed by colour alone
Images (JPEG, PNG)
Pixels are analysed and the metadata is read. 7 criteria.
A machine tests
- No text rendered as pixels — and if there is, what it says
- Contrast of text in the image against its own background
- Dominant colour pairs clear the contrast thresholds, matched to your brand palette
- Embedded accessibility description (IPTC/XMP) present
A person judges — listed on every result, never scored
- Embedded description is meaningful
- No information conveyed by colour alone
- Alt text where the image is used — stated, not scored
SVG
The vector source is parsed; outlined text is caught by rasterising the graphic and reading it back. 6 criteria.
A machine tests
- Has a <title> element
- role="img" where the SVG is a single graphic
- Text is real <text>, not outlined paths
- Text contrast
A person judges — listed on every result, never scored
- Complex graphics have a <desc>
- No information conveyed by colour alone
Every criterion, its clause and who tests it is on the methodology page, rendered from the same file the engine scores against.
Image metadata check
The description that travels with the image. Or does not.
Alt text lives in the page or the document that uses an image. The image file itself can carry its own accessibility description in its metadata — the IPTC and XMP fields that Photoshop, Bridge and most digital asset managers expose — and when it does, every downstream use can inherit it. Most brand asset libraries have never had this measured, because nobody had a tool that read the field. We read it, report it as a first-class finding on every JPEG and PNG, and flag the images where it is missing or meaningless.
Alongside it: OCR on every image and every picture inside a deck, so text baked into pixels — the headline on the social tile, the phone number in the banner — is surfaced with what it says, and its contrast measured against its own background.
Brand palette contrast
We test the palette itself, not just its use.
Most contrast failures in documents are not a designer’s mistake. They are two approved brand colours that were never meant to carry text on each other. The palette in your Brand Blueprint is measured once, pair by pair, and sorted into the pairs that can carry body text, the pairs that can carry large text only, and the pairs that cannot carry text at all — with the colours that fail against both black and white called out.
When a document fails contrast on one of those approved pairs, the finding says so and links back to the palette result: the fix is in the brand guidelines, not in that file. That is the difference between correcting one deck and correcting the next hundred.
Where the line is
What automated checking cannot establish.
We can tell you exactly where the line is because we work on both sides of it every day. A machine can establish that things exist and that measurable values clear a threshold. It cannot establish meaning. Anyone who tells you otherwise is selling the number, not the measurement.
Read the full methodology →It finds the alt text.
It cannot tell you whether the alt text describes what the image is for.
It finds the reading order.
It cannot tell you whether that order makes sense to a listener.
It finds the headings.
It cannot tell you whether the structure reflects the content.
It measures the colours.
It cannot tell you whether colour is the only thing carrying the meaning.
It finds the video.
It cannot tell you whether the captions are accurate.
It counts the criteria.
It cannot tell you the file meets a standard — and neither do we.
The standards we test documents against.
WCAG is written for web content. Documents have their own standards that map WCAG onto them, and every document criterion in our catalogue cites its clause.
EN 301 549, clause 10
The European ICT accessibility standard; clause 10 is the part written for non-web documents — PDFs, presentations, spreadsheets, images. Referenced by the European Accessibility Act and by UK public-sector regulations.
PDF/UA-1 (ISO 14289-1)
The ISO standard for accessible PDF. Tagging, title, language, figure descriptions, artifacts and fonts each cite their PDF/UA clause alongside the clause 10 reference.
WCAG 2.2 AA criteria, as mapped
The underlying success criterion is cited on every check for reference, always through the clause 10 mapping — because that is the standard that actually speaks to documents.
We test against these standards. We do not declare that a file satisfies any of them — that is a judgement a full audit makes once the human criteria are answered, and it is not something an automated tool can sell.
If a regulation names your documents, you already know it does.
A few kinds of organisation have a dated obligation that explicitly covers the files they publish. For them, a document result that states what was tested and what was not is the record their reviewers will ask for.
Suppliers to US state and local government
The DOJ Title II rule treats PDFs, Word, PowerPoint and Excel files as web content, and it reaches content a public entity provides through third-party arrangements — the agency that builds the county parks brochure included. The dates are April 2027 for larger entities and April 2028 for smaller ones.
Healthcare organisations funded by HHS
The HHS Section 504 rule brings the same web and document expectations to organisations that receive HHS funding, on a similar timetable. Patient leaflets, forms and portals are documents first.
Federal contractors under Section 508
Deliverables to federal agencies are expected to follow Section 508, which references WCAG 2.0 AA and has its own document rules. Knowing what a machine has and has not tested is exactly what a contracting officer will ask.
Universities and colleges
Public institutions fall under Title II; most others hold federal funding. Course packs, slide decks and PDFs are the bulk of what students actually receive.
Everyone else: this is governance, not a deadline.
If none of the above is you, no regulation is naming your brochures — and we will not pretend one is. The argument is the same one that applies to brand drift: the files you ship are unmeasured and unowned, and unmeasured things get worse quietly. A number with its coverage beside it, kept file by file and rolled up by brand, turns an invisible risk into something a team can own and improve. That is worth having whether or not a lawyer ever asks.
Pixel-measured contrast
Text over photos and gradients is measured from the rendered page pixels — the cases source-code scanners cannot see at all.
Progress you can prove
Every scan is kept, and each report shows what changed since the last one — rules fixed, new issues, counts moving. Fixes become a story you can show clients.
Honest about its limits
Automated testing catches roughly a third of WCAG. Anything it cannot decide is listed as needing human review — never guessed, and never dressed up as more than it is. The full catalogue and formula are public.
Website & Document Accessibility, answered.
Does a clean result mean my site or my document meets the standard?
No — and anyone selling that from automated checks alone is overclaiming. Automated tools can test only part of any accessibility standard. Every violation we report is a measured fact (a contrast ratio, a missing attribute, a missing tag), and anything automation cannot decide is listed as needing human review, with its count shown next to every number. It is an early-warning system that catches real, fixable problems before they become complaints. The full catalogue and the formula are on the methodology page.
Which files get a document accessibility check?
Every PDF, PowerPoint (.pptx), Word (.docx), Excel (.xlsx), JPEG, PNG and SVG uploaded for a Brand Review, automatically — the same upload produces the brand result and the document result, and the document check always reads the original file, not a conversion. Each result opens with a “What we checked” panel: which criteria a machine tested, which need a person, which do not apply to that file and why.
What is the number on a document result?
The share of automated checks that applied to the file and found no problem, always shown with how many of the catalogue criteria were machine-tested and how many still need a person. Checks that do not apply are left out of the denominator and never counted as passed. A PDF with no structure tree is capped near zero with the uncapped figure beside it, because nothing a screen reader needs can exist in an untagged file. A scanned PDF gets no number at all, just the sentence that explains why.
Which standard should I pick for a website scan?
Pick the jurisdiction whose law applies to you — for example the ADA option for a US business, or the European Accessibility Act for anyone selling into the EU. If you are unsure, the default International WCAG 2.2 AA is the strictest superset: clearing its checks covers the checks of every older version the other laws reference.
How many pages can I scan?
Accessibility scans have their own generous monthly page allowance, separate from your review quota, because no AI is involved: 25 pages on trial, 300 on Starter, 1,500 on Professional and 6,000 on Agency — enough for roughly three full scans of your site every month. A single scan covers up to 25 pages on trial, 150 on Starter, 500 on Professional and 2,000 on Agency. A scheduled re-scan uses the same allowance.
What are the guided manual checks?
The WCAG checks no automated tool can judge — can everything be operated by keyboard alone, do error messages actually explain the fix, do videos have accurate captions. We turn them into 22 plain-language tasks your team works through with the site open, each recorded with who checked it and when, and the results appear in the PDF report. That combination — automated plus documented human checks — is what a real audit looks like.
Can it generate an accessibility statement for my site?
Yes — from any completed scan you can generate a publishable accessibility statement built from your real results: the standard tested, what was found, known limitations and your contact route. It deliberately says “tested against”, never more — an honest statement protects you, an overclaiming one is evidence against you.
What about checking my website against my brand?
That is its own product — Website Brand Review — which screenshots your key pages and reviews them against your Brand Blueprint using the review pipeline. The two run separately, so accessibility scans never spend your AI review allowance.
The rest of the platform
Asset Review →
Check a single asset before it ships — five checks, one score, findings drawn on the artwork.
Brand Audit →
One score across your whole organization — sampled from every touchpoint, exported as a client-ready PDF.
7-day free trial · Files never used to train AI