PDF/UA-1 · Matterhorn Protocol · WCAG 2.1 AA · IS 5568
Cut a 40-minute remediation to 8 minutes — and produce the audit trail.
Taggart is a power tool for a specialist, not a one-click compliance button. It finds every machine-checkable Matterhorn failure condition, fixes the mechanical majority without touching a pixel of your page, and puts everything that needs judgement in front of you with the evidence attached.
- Check free, no account
- See the repair run before you pay
- Hebrew and Arabic
- Runs in the browser
The order the tags are in reads across the columns
1Revenue grew 14% in the third quarter,
3driven by three new contracts in the
5Northern district.
2Costs rose 4% over the same period,
4largely from the additional field
6staff those contracts required.
What a screen reader reads two articles, interleaved
“Revenue grew 14% in the third quarter, Costs rose 4% over the same period, driven by three new contracts in the largely from the additional field Northern district. staff those contracts required.”
After Taggart column order repaired
“Revenue grew 14% in the third quarter, driven by three new contracts in the Northern district. Costs rose 4% over the same period, largely from the additional field staff those contracts required.”
Reordering edits the structure tree, never the content streams that paint the page. Structural similarity against the original: 1.000000.
Three things no other tool does the same way
Everything here exists because a specialist's time is the expensive part, and because a tool that quietly damages a document is worse than no tool at all.
The reading-order editor
Numbered badges on the rendered page show the order a screen reader will actually follow. The tree and the page stay selected together in both directions, Alt plus the arrows moves an element, and undo spans the whole batch — none of which is true of Acrobat's Reading Order tool.
Hebrew and Arabic, properly
Many Hebrew PDFs store text pre-reversed. The page looks perfect and a screen reader reads it backwards. Taggart detects it structurally — by comparing paint order against page geometry — and repairs it without touching a painted glyph.
The report is the product
Issues found, issues fixed, issues accepted with a justification, issues remaining, every human decision with a timestamp and an actor, the tool version, and the SHA-256 of both files. That is the artifact a compliance buyer is actually paying for.
It cannot change how your document looks
This is the guarantee everything else rests on. Taggart's fixes edit the
PDF object model — structure elements, attributes, catalog entries — and
never the content streams that paint the page. The two content-stream
rewriters it does have insert only BMC, BDC and
EMC operators, which paint nothing.
Every remediation is re-rendered next to the original and compared page by page at structural similarity ≥ 0.999. A result that fails the gate is thrown away rather than handed back. The same test runs over the whole test corpus on every commit, and it measures 1.000000.
A tool that visibly damages documents is dead on arrival, and in this market you only get to do it once.
Start with the condition that is costing you time
Each of these explains what is in the file, what a screen reader does with it, and how to fix it free in Acrobat Pro before you spend anything.
- 09-001 Fix PDF reading order: why tags read out of sequence The structure tree, not the page layout, decides what a screen reader reads first. How to see the real order, and why Acrobat's two panels disagree.
- TG-RTL-001 Hebrew PDF reads backwards: visual order vs logical order Many Hebrew and Arabic PDFs store text pre-reversed. The page looks right, the extracted text is mirrored, and a screen reader reads it backwards.
- 15-003 TH cells need a Scope attribute — or Headers and IDs A header cell that does not say whether it heads a row or a column leaves the screen reader guessing. How to set Scope in Acrobat, and when Scope is not enough.
- 31-027 Missing ToUnicode CMap: looks fine, extracts nothing Text renders perfectly, copy-paste produces garbage, the screen reader is silent. Why ToUnicode is the number-one cause, and what can and cannot be repaired.
- 14-003 Fix skipped heading levels in a PDF (H1 to H3) Heading levels may go back up by any amount but may only go down one at a time. Why the rule exists, and how to repair it without flattening the outline.
- 28-005 PDF form field with no accessible name (missing TU) The label printed next to a field is not connected to it. The tooltip is the accessible name, and without it a screen reader announces 'edit text, blank'.
The deadlines
These are the dates driving the work. Each is stated as the instrument states it.
Directive (EU) 2019/882 applies to covered products and services placed on the EU market. It is already live.
State and local government entities serving a population of 50,000 or more must meet WCAG 2.1 Level AA. The date reflects the Department of Justice interim final rule of 20 April 2026, which extended the original deadline by one year.
Smaller public entities and special district governments have a further year. Same technical standard: WCAG 2.1 Level AA.
Summarised from the instruments named. Not legal advice — confirm your own obligations with counsel, and in Israel with a licensed מורשה נגישות.
What Taggart will not tell you
That your document is compliant. No tool can determine that, and the regulator agrees: when the US Department of Justice extended the ADA Title II deadlines by a year, its stated reason was that it had overestimated the capabilities — staffing and technology — of the entities it covers.
So Taggart draws the line explicitly. It reports what its engine verified. It says which conditions require human judgement and whether you have decided them. It refuses to write a PDF/UA conformance claim into a file its own checker cannot confirm. And every report carries the sentence: this is not a legal determination of compliance.
Every service vendor in this market charges per page because a person has to be in the loop. Taggart makes that person faster. It does not replace them.
Start here
If you are new to this
Four pages that cover the ground before any tool comes into it — what the work actually is, what the words mean, which standard applies to you, and what it costs.
- How to make a PDF accessible — the whole job in order, what is mechanical, where a person has to decide, with real timings.
- What is a tagged PDF? — and why tagged and accessible are not the same thing.
- PDF/UA vs WCAG — which one your obligation cites, and which one a validator can test.
- What remediation actually costs — vendor rates, what per-page pricing hides, and how to price your own backlog.
Questions people actually ask
Is the validator really free, or is it a trial?
Free, unlimited, and not a trial. Every machine-checkable Matterhorn condition, no page cap, no watermark on anything, no expiry, and no account required. Checking is commoditised — PAC, veraPDF and PDFix Lite all do it for nothing — so gating it would only send you to one of them. Remediation is what a plan pays for.
Will remediation change how my document looks?
No, and that is enforced rather than promised. Fixes edit the PDF object model, never the content streams that paint the page. Every remediated file is re-rendered alongside the original and compared page by page at structural similarity ≥ 0.999 before you get it back; a result that fails the gate is discarded rather than returned. The same test runs across the whole corpus on every commit.
Does Taggart make my documents legally compliant?
No tool can do that, and any that says otherwise is selling you a liability. The US Department of Justice extended the ADA Title II deadlines specifically because it had overestimated what technology could do here. Taggart reports what its check engine can verify, puts everything requiring judgement in front of a human, and records those decisions in the report. Whether that adds up to compliance is a determination a qualified person makes.
What makes the Hebrew and Arabic support different?
Every other tool assumes text is stored in logical order. Many Hebrew PDFs — particularly from older Israeli government and legal systems — store it pre-reversed, so the page looks correct and the extracted text is mirrored. Taggart detects that by comparing the order glyphs are painted in against their positions on the page, and repairs it by attaching logical-order ActualText without touching a single painted glyph.
Do you upload my documents anywhere?
A signed-out check is processed and discarded — nothing is stored. Signed in, a check is only kept if you tick the box, and uploads expire after seven days. There is no alt-text generation and no model of any kind, so no image is ever sent anywhere. This audience reads the privacy policy and checks the claims with a proxy, which is the right instinct.
New checks, as they are written
A short note when a batch of /checks/ pages goes up, when the rule set changes, or when something ships. No more than monthly, and nothing else — no drip sequence, no webinar invitations.
One click to unsubscribe, in every message. We do not sell or share the list.
Run your worst file through it. Not a sample — the one that has been sitting in your queue. Checking is free and unlimited, and the mechanical repair runs free too: you will see what Taggart changes, and that the page is pixel-identical afterwards, before there is any question of paying for it.