Update archive

What's new

Every update announced on the home page, newest first. The banner there shows only the most recent one and can be dismissed, so this is the full list.

ScoringOctober 6, 2026· current
PowerPoint text is now checked for contrast in the colour PowerPoint shows it in: most text takes its colour and size from the deck’s design rather than setting its own, and that text was never checked before. Across the real decks in our test set, about half of all text is now checked, up from about 7%; links are judged in the colour PowerPoint draws links in, and highlighted text against its highlight. A colour shaded lighter or darker, or made see-through, is left unchecked rather than guessed at. Separately, a picture “described” only by a label such as “Table (page 30)”, which automatic tagging tools write, now counts as undescribed, so some grades will go down, correctly. How scoring works
ScoringOctober 6, 2026
A picture whose only “description” is its file name or a placeholder word now counts as undescribed. Some programs write these automatically when the author leaves the description blank: Google Slides and some scripting tools use the file name (“chart.png”), and others write “Picture” or “image 1”. PowerPoint slides are now read the way PowerPoint shows them. Text on a slide that takes its background from the slide master is checked for contrast, and bullets typed by hand where the layout turns bullets off are caught. These came from building test documents the way real programs write them, so some grades will go down, correctly. How scoring works
ScoringOctober 6, 2026
Word, PowerPoint and Excel files are now read the same way whichever program wrote them. A new check writes one document 81 different but equally valid ways and requires the same grade every time. Its first run found seven misreadings, all now fixed. Among them: line breaks in a picture’s description were kept as codes; files saved in the UTF-16 text format were turned away; text marked “not bold” was read as bold, which could invent a typed heading in a Word file or pass faint grey text in a workbook; and a slide hidden with “false” was judged anyway. No grade on any file in the test set changed. How scoring works
ScoringOctober 6, 2026
PDF tables are now judged by what they show on the page, as Word tables already are. A table with no header cells and nothing drawn — no lines, no shading — is a grid used only to line things up, and is no longer accused of a missing header row; a table drawn with lines still needs its header cells. Reports now also point out table headers that say nothing: a header row that is marked but empty, or an Excel table still headed “Column1, Column2”. Those notes do not change the grade. How scoring works
ScoringOctober 6, 2026
Word tables are now judged by what they show on the page. A grid with its borders switched off — how Google Docs and LibreOffice save an invisible table — or one carrying a table style that draws nothing is a layout grid, not a data table, so it is no longer accused of a missing header row. A data table drawn with lines around each cell, instead of around the whole table, is now checked like any other and needs its header row marked. These came from a new round of test documents for tables, written before any fix. How scoring works
ScoringOctober 6, 2026
PDF section titles that look like headings but are tagged as plain text now cost what they always cost in a Word file — two annual reports that tag only three of their headings had been told “No issues found”. The check ignores cover pages, letterheads, pull quotes and captions, which look like headings and are not. A language marked on the text itself now counts in Word as it already did in PowerPoint, so a file fixed exactly as this report advises — select all the text, then set its proofing language — is no longer told it has no language; in both, the language must cover most of the text. PowerPoint also gets the best-practice check Word has for tables used only to line things up. How scoring works
ScoringOctober 5, 2026
The last small grading differences between file formats are closed. Light-colored text typed onto an Excel spreadsheet’s plain grid is now checked for contrast, as the same text in a Word document always was, and a Word or PowerPoint file whose declared language does not match its text — a Spanish notice labeled English, for example — is now flagged, as PDF files already were. A table used only to line things up on a slide is no longer graded as a data table, a damaged file is no longer reported as having no title, and the same share of described images now earns the same score whatever program made the file. A logo in a Word document’s page header still needs a description or a decorative mark, and the report now says where it is and how to mark it. How scoring works
ScoringOctober 5, 2026
Three more problems now cost the same whichever program made the file. A list typed by hand instead of made with the Bullets or Numbering buttons no longer drops a Word or PowerPoint file to D — every word is still there and in order — so it now costs what an unmarked table does (a C at most). A PowerPoint deck whose section headings are all typed into text boxes, with no slide titles anywhere, now costs what the same headings cost in Word or a PDF. And Word, PowerPoint and Excel titles now get the check PDF titles always had: a title that is a file name or a program’s default — including PowerPoint’s own “PowerPoint Presentation” — does not identify the document and costs half the title points. How scoring works
ScoringOctober 5, 2026
A Word memo or notice with one bold title line is no longer graded down for missing headings. One line that looks like a heading is the document’s title; two or more are sections that need real heading styles — the rule PDF reports have followed since September. Word was taking 70 points for that single title line, so a one-page memo could grade D in Word and A as a PDF, and PowerPoint took a smaller deduction for a lone typed title. All three formats now agree, and a document with two or more unmarked section headings is still flagged. How scoring works
ScoringOctober 5, 2026
A Word table with its Header Row box ticked is no longer reported as having no header row. That box — Table Design → Header Row, ticked by default — is how Microsoft tells authors to mark a table’s header row, and Microsoft’s own accessibility checker accepts it; this tool accepted only Repeat Header Rows, so a correctly built meeting agenda graded D. A table whose header row really is unmarked now costs the same in every format: it used to cap a Word or PowerPoint file at D, a PDF at C, and an Excel workbook at B, and now caps all four at C. Some Word and PowerPoint grades move up, and an Excel workbook with an unmarked table can move from B to C. The advice for PDFs exported from Word is corrected too: Word’s built-in Save as PDF does not write the header directions (Scope) that two-way tables need, so the report no longer suggests that re-exporting will add them. How scoring works
NewSeptember 8, 2026
The service status page now shows how hard the server is working. If a document is taking longer than you expect, you can see whether the machine is simply busy — a large report uses every processor it can get for a minute or so, and the page says plainly that busy is not the same as broken. It shows processor use, memory, and free disk space together, and it publishes numbers only: nothing about the hardware, nothing about any file, and nothing about anyone using the tool. Service status
ScoringSeptember 2, 2026
Two accusations are now made only on evidence — some PDF scores will move up. A PDF with no heading tags is failed for it only when the checker can see section headings painted on the page (larger or bold lines sitting over body text), never on page count alone. And a title is failed as a bare file name only when it really is one: a real title with a download timestamp or underscores glued on names the document and is reported for review instead. A text box exported as a picture is no longer counted as a missing description. How scoring works
NewSeptember 1, 2026
Your action plan now says how long each fix should take. Most document fixes are minutes of clicking, not an afternoon — a missing title is about two minutes in Word, nine hand-typed bullets about three — so each step now carries an estimate computed from your own document's counts, and the plan totals them when every step has one (for a typical short document: under 15 minutes). The numbers cover only the mechanical part; judgment work like writing good alt text is labelled, never timed, and a step with no defensible number simply shows none. The printable plan carries the same estimates, so the sheet beside your keyboard answers “can I do this now?” Release notes
ScoringSeptember 1, 2026
A title nobody hears now counts as half a title — some PDF scores will move. A PDF can have a perfectly good title and still tell every PDF viewer to show the filename instead; that is what an unset “show document title” preference does, and for someone using a screen reader the result is the same as having no title at all. The W3C’s own accessibility technique for PDF titles is to switch that preference on, and it is a two-click fix in Acrobat (Document Properties → Initial View → Show: Document Title). Until now the report mentioned it as a best practice and charged nothing; a title that is set but hidden now earns half of the title’s points instead of all of them, and the finding tells you exactly where the switch is. Most well-made documents already have it right — in our reference corpus, 43 of the 47 titled documents did. The same review also fixed accusations in the other direction: a white slide title sitting on a dark banner is no longer misread as white-on-white, a lone typed “1.” is no longer treated as a broken list, and titles stored only in a PDF’s newer metadata block are finally read at all. How scoring works
Fixed
Light mode had never been tested, and it was hard to read. This app opens in dark mode, so every automatic check it runs on itself had only ever looked at the dark version — including the checks behind the claim that both were fine. Switching the default and re-testing found 136 places where text was too faint against its background, most of them on the data-retention page, and one on this page at roughly a quarter of the required contrast. All are fixed, and every page now passes in both appearances. Publishing the number rather than quietly correcting it seemed the only consistent thing to do, given that faint text is one of the things this tool checks other people's documents for. How this app is tested
FixedAugust 31, 2026
The best-practices list no longer says “not applicable” about something that already cost you points. A reader spotted an annual report with no headings at all showing five “not applicable” labels about headings — while the same report scored that document’s heading structure zero and named the rule it broke. That label was the wrong fix, and it was withdrawn the same afternoon: rows about something already graded are now simply gone from the list. Best practices shows only what is genuinely extra credit — work worth doing that no rule requires — so nothing in it repeats a point already made in your action plan. How this app is tested
ImprovedAugust 31, 2026
Your audit no longer disappears if you click away. Opening the status page, reloading, or following a link while a document was being checked used to cancel it and make you upload the file all over again. It does not any more: the check runs on our server rather than in your browser, and this page now remembers which one is yours, so coming back picks it up where it left off — and if it finished while you were gone, the report is simply waiting. A report you have already read is kept the same way for the rest of your visit. Because of that, the warning about leaving mid-audit now appears only when leaving really would lose something, such as a batch of several files. Everything remembered is kept in your own browser, cleared when you close the tab, and never sent anywhere; the data-retention page lists exactly what it is. What your browser keeps (data retention, § 8)
ChangedAugust 31, 2026
Three checks stopped taking points for things this tool could not tie to the law. The biggest affects Word, PowerPoint and Excel files with vague link text: a document with two “click here” links used to lose its entire Link Quality score and be capped at a D, even though the rule behind it, WCAG 2.4.4, lets the sentence around a link supply its purpose — something no automated check can judge. Vague link text is still reported in full, with the links listed, but it no longer changes your grade. A link with no text at all does still count, because a link with no text has no name for a screen reader to announce, and WCAG 4.1.2 requires one. A PowerPoint slide whose title is not the first thing read on the slide is likewise now reported rather than counted. If one of these applies to your document, it may score higher than it did before. How this app is tested WCAG 2.4.4: Link Purpose (In Context) WCAG 4.1.2: Name, Role, Value
ChangedAugust 31, 2026
Reports now name WCAG 2.1 everywhere. That is the version ADA Title II and the Illinois IITAA both require, and the only version that has moved a grade here since the legal-only scoring change — but some labels and links still said 2.1's successor, so a report could state one standard and link to another. WCAG 2.2 still exists and What's new in WCAG 2.2 still explains what it adds; nothing it adds is counted. Two smaller additions: the font check now lists every font in the document and whether it is embedded, rather than only saying whether they all are; and every practice the checker could not examine now carries an icon explaining why it was not checked and what would settle it, beginning with the fact that nothing is wrong with your document. How this app is tested
ChangedAugust 31, 2026
Word documents now lose points for heading styles applied to blank lines. A heading style on an empty line puts a section in the outline that has no content in it, so someone moving through the document by heading lands on silence — it is counted as a WCAG 1.3.1 failure, with its own fix step. The same defect in a PDF is still reported without being counted, because there the evidence is an estimate rather than a certainty, and the report says so on the row. Every practice that could not be checked now carries an information icon explaining why, and each explanation begins by saying that nothing is wrong with your document. Alongside these, an audit of every claim the reports make about the standards corrected the ADA Title II compliance dates everywhere they appeared: the Department of Justice extended them in April 2026, to April 26, 2027 for entities serving 50,000 people or more and April 26, 2028 for smaller ones and special districts. How this app is tested WCAG 1.3.1: Info and Relationships
NewAugust 30, 2026
Reports now include a Best practices section between the action plan and Above and beyond, in every view: the visual report, the detailed report, a shared report, and the printout. It lists every non-scored best practice for your file's format — 38 in all, across PDF, Word, PowerPoint, and Excel — and for each one shows whether your document meets it, the evidence found in your document itself, both ways to fix it, and a link to learn more. None of it touches your score: a row that says work is worth doing is never a failure. Three new test documents prove the point rather than just claim it — a PDF, a Word document, and an Excel workbook that each pass WCAG 2.1 completely, score 100/A, and still carry real best-practice work for the new section to report. Above and beyond now shows only what veraPDF itself found, and six advisory findings for Word and Excel documents that were rendering under the wrong heading now render under the right one. How this app is tested
ChangedAugust 29, 2026
Scores now measure WCAG 2.1 — and nothing else. Every check in the scoring engine was audited against one rule: a deduction must correspond to a failing WCAG 2.1 A/AA criterion the report can name, or it may not touch your grade. Fifteen checks that measured best practices rather than the law — bookmarks, reading-order heuristics, heading-level conventions, link wording, slide titles, sheet names, and more — now appear as clearly labelled advisories that never affect the score, while confirmed legal failures gained explicit citations they were missing (untagged content, missing headings in long documents, wrong or unusable language declarations, filename-as-title). Most documents will score the same or higher on re-analysis; none score lower. A new automated gate re-verifies the rule on every change: every point lost must name the law it broke. How this app is tested
ChangedAugust 29, 2026
The action plan's labels now match the verdict's arithmetic. A report could say '3 criteria failing' while showing five fix steps, every one stamped 'Required by WCAG 2.1' — confusing, and for some steps wrong: bookmarks on a long document and the reading-order signal raise your score and genuinely help readers, but they are not WCAG 2.1 criterion failures. Now the REQUIRED label is earned — it appears only on steps whose category actually produced a failing criterion — the rest say RECOMMENDED, and the plan's opening line does the math for you: '3 of the 5 clear WCAG 2.1 criterion failures; the other 2 are recommended.' The summary also now guarantees in writing that nothing beyond WCAG 2.1 A/AA is counted — not the criteria WCAG 2.2 added, not PDF/UA — and tests enforce it. How this app is tested
NewAugust 29, 2026
Every PDF/UA rule veraPDF flags on your document now carries its own 'How to fix' — collapsed until you want it, and written both ways: how to prevent the issue in the source file (Word or InDesign) before you export, and how to repair it in the finished PDF with Acrobat. So if you choose to go beyond what WCAG 2.1 requires, each item gives you a place to start on whichever side of the export you work. Advice appears only for rules we can speak to confidently — an unfamiliar rule shows veraPDF's words alone, because wrong advice under the referee's verdict would be worse than none. How this app is tested
ChangedAugust 29, 2026
Reports now name the standard precisely: every required item says 'Required by WCAG 2.1' — the standard Illinois (IITAA) and federal law (ADA Title II) actually name — instead of the vaguer 'required by law', and everything optional is labelled a PDF/UA best practice. The action plan's 'Above and beyond' section now carries veraPDF's complete verdict for your document: every failing PDF/UA rule with its ISO clause and occurrence count, plainly marked as not affecting your grade. A same-day review of the new scoring also fixed one genuine blind spot before any file hit it: a one-row table with a leading header cell was being read as a two-axis crosstab and docked for a missing direction — it now correctly passes, and a trap document pins it forever. How this app is tested
ChangedAugust 29, 2026
Every report now opens by answering the compliance question directly: what the law requires, and whether this document meets it. WCAG 2.1 AA, ADA Title II and the Illinois IITAA get the headline; PDF/UA — the PDF industry's own standard, which is good practice and not law — sits underneath as a footnote, clearly marked as not counted in your score. We also audited every deduction the score is made of and removed two that no law requires: font embedding and nested tables. Both are still reported; neither affects your grade any more. Seven documents in our own test set scored slightly higher as a result, and none changed letter grade. How this app is tested
ChangedAugust 29, 2026
Your grade now measures only what the law requires. Reports are split into two clearly labelled parts: what Illinois and federal law require (WCAG 2.1 AA, ADA Title II, IITAA) — which is what the score is made of — and, separately, what would additionally meet PDF/UA, the PDF industry's own standard, marked as not counted in your score. The clearest example is table header directions: where a table's headers run along a single edge, the relationship is already determinable and the law is satisfied, so a missing direction no longer costs you points; where headers run along two edges, it genuinely cannot be worked out, and that stays a real WCAG failure. Some documents will score slightly higher than before, for a defensible reason. How this app is tested
FixedAugust 29, 2026
Two fixes and one new check, all found in documents published as accessible examples. First: a table's header directions and cell spans can be stored as references to shared values rather than written out in place — legal, and how Word often writes them. This checker read the reference instead of the value, which made correctly built tables look like they were missing header directions and had ragged columns. Both are fixed. Second, and new: a document can declare the wrong language. One reference syllabus written in English declared itself French, because a single French sentence in the Word source became the language of the whole PDF on export. Every conformance checker passes that file — the tag is present and correctly formed — but a screen reader reads the entire English document with French pronunciation. Reports now say so when the declared language plainly contradicts the text. How this app is tested
NewAugust 29, 2026
Reports now show you when the finding is not ours. veraPDF — the PDF industry's own validator, published by the PDF Association and not written here — runs beside this checker on every PDF report. Where it independently failed the same point, its exact words, its ISO clause, and its failed-check count now appear right beside the fix, in both the visual plan and the detailed report. And where veraPDF is stricter than this checker — it flags something we do not score — the report says that too, plainly marked as not counted in your score. Silence is never dressed up as agreement: when veraPDF did not run, or did not flag a point, nothing is shown. How this app is tested
FixedAugust 29, 2026
A real bug, found in reference documents published by the state's own IT agency — and their files were right. When a table has labels across the top AND down the side, the corner cell that labels both directions is marked with a third value the standard provides, "Both". This checker only recognised the other two, so a correctly built table was reported as missing header directions and its document was marked down. Fixed: all three values are now accepted, the fix-it advice now teaches the corner-cell case instead of steering around it, and a new trap document locks the behavior so it cannot come back. One document re-graded 89 to 100; nothing else moved. How this app is tested
VerifiedAugust 29, 2026
The trap battery now covers all four formats this checker grades. Fifteen new trap documents were hand-built in Word, PowerPoint, and Excel's own file formats — bold text posing as headings, pictures whose alt panel was never opened, header rows never marked, slides with no titles, sheets still named 'Sheet1' — beside matching corrected copies of the same files that must pass clean. All 115 traps then in the battery held, and a new rule is enforced on every flawed/corrected pair: the broken version may never outscore the corrected one. Every control's exact score is also now pinned in the public ledger — 151 documents whose grades cannot move without a person approving the change. See the trap documents
VerifiedAugust 29, 2026
Three new gates now stand between every code change and this site, built to guarantee the accuracy of your report. The golden score ledger: every control document's exact score and per-category verdict is pinned, and no grade can move unless a person approves the movement in review. The re-save test: every trap document rewritten by a different tool must grade identically, digit for digit — byte layout can never change a grade. And live-site sentinels: after each deploy, known trap documents are uploaded to this very site, where the disguised scan must still score 0 and the perfect document must still score 100. The Can I trust this? page now lists every way this app is tested before anything ships. How this app is tested
VerifiedAugust 28, 2026
The battery of trap documents built to fool this checker has doubled again: one hundred of them, every designed answer returned correctly. The second fifty are modeled on what Canva, InDesign, and Word actually export — flat untagged posters, decorative-shape swarms, custom naming soup, titles that are pictures of words, 'Print to PDF'. The Can I trust this? page (now the first link in the menu) lists all one hundred with what each one tries and how it was judged, and the whole battery re-runs before any change ships — a broken promise stops the release. See the trap documents
New pageAugust 28, 2026
A fair question deserves its own page: Can I trust this? The new link at the top and bottom of every screen answers it in plain English — the laws this checker serves (Title II of the ADA, the Illinois IITAA, and the WCAG rulebook both point to), the independent referee program that co-signs every report, the battery of trap documents built specifically to fool the checker (now 50 of them — all judged correctly), the working sessions with internal and external accessibility specialists, and the grade disputes the checker lost in public and fixed the same day. Every number on the page is pulled from the live system, under a big dated stamp so you can see exactly how fresh it is. Can I trust this?
VerifiedAugust 28, 2026
How do we know this checker actually works? Fair question — running it on real documents only proves it agrees with itself. So we built eighteen small trap documents designed to fool it, each constructed so the correct answer is known in advance: a document with a perfect title that is secretly one big photograph, a picture description made only of blank spaces, internal structure that loops back on itself, headings that are really whole paragraphs, a form with no labels beside an identical one with proper labels, the leftovers design tools like Canva and InDesign leave behind, and text carrying hostile computer code to prove it stays harmless. The checker passed all eighteen then in the battery — after it caught the first of its own bugs, which is exactly what it is for: a description of only blank spaces was counted as a real description. That is fixed; a check now asks what a person would hear, not whether the field has something in it. No real document in our test set had the defect, so no existing score changes. The battery is rebuilt from scratch on every run and anyone can re-run it. How scoring works
FixedAugust 28, 2026
If a report marked images as missing a description and you could not find those images, this was why — and your score may now be higher. The remediator of a document told us that two pictures we flagged simply were not in it. That was correct: the document has four pictures and every one already had a description. The two we complained about were leftovers from a page that had been rebuilt — a scrap of an older version left inside the file, on no page, connected to nothing, and unreachable by a screen reader. We had been counting them as real content. That report scores 100 out of 100; we had given it 79. Leftovers like this are completely ordinary — design and layout programs leave them behind whenever a page is remade after being prepared for accessibility — so nothing about the submitted document needed fixing. We now follow the path back to the top of a document to decide whether a piece is really part of it, rather than trusting a single step. How scoring works
ScoringAugust 28, 2026
Headings are now judged on what they say, not just how they are numbered — some reports will score lower. Someone asked us to prove that a report's findings were right, so we read the document itself instead of trusting our own tool. They were right, but one finding was far too kind: it noted that the headings “skipped levels” and marked the file down gently, when in truth 19 of its 96 headings contained no words at all, 14 were whole paragraphs marked as headings, and 29 stopped in the middle of a word — things like “Population d”. Headings are how most screen-reader users move around a long document, so an outline like that leaves them landing on silence and half-sentences. We were already showing those broken headings on screen and giving them no weight. Now they count. We also retired a piece of advice that pointed the wrong way: a document was told its lists were missing the part that holds each item's text and to go and add them, when all 43 were already there, spelled with one letter in the wrong case. That is a one-line fix, not hours of rebuilding, and the report now says so. How scoring works
FixedAugust 28, 2026
Long documents are no longer refused for being “too complex”. If you audit big reports — annual reports, budget books, anything over a hundred pages — you may have seen the tool give up and tell you the file was too complex to check in the time allowed. That was our fault, not your document's. The tool was starting several checking programs at the same moment on one small server, and they crowded each other out until one was cut off while it was waiting rather than working; a 246-page annual report that our own reader gets through in under two seconds was being turned away. The checks now run in order, the heaviest one runs a single pass at a time, and big documents get up to two minutes instead of one. The waiting screen says so, and if an audit does run out of time the message now tells you the truth — that it is usually about timing, and that trying again is normally enough. The same crowding was behind the “server offline” and “degraded” notices some of you saw at the top of the page: the server was fine, and that warning now clears itself. Service status
ClearerAugust 27, 2026
Table findings now tell you which way your headers point. A header cell can be marked correctly and still not say whether it labels the column beneath it or the row beside it — and the report used to ask you to set that without saying which value to use where. It now gives the rule: cells along the top label what is beneath them, cells down the left label what is across from them, and the empty corner cell needs nothing. In Word you rarely need to do it by hand — tick the Header Row box and Word writes it for you. We have also explained something that confuses people regularly: your accessibility expert may say a file is fully compliant while this report says something is missing, and both can be right, because the PDF standard and the web guidelines set different bars here. Doing it satisfies both. How scoring works
FixedAugust 27, 2026
If you audit forms, your score may go up. Someone told us a score was wrong, and it was: a form built by the State's own accessibility team was marked down for reading order, and its author said the document was fine. They were right. The check compared the order a document is tagged in against the order it is painted in — useful for ordinary documents, but meaningless for a form, because forms paint their field labels last no matter how carefully they are tagged, so the better the form was built the worse it scored. Reports now say plainly that reading order cannot be judged automatically in a form, and show you how to check it yourself. Low-scoring cards also stop putting a red cross beside plain facts and beside the report's own notes. How scoring works
New checkAugust 27, 2026
Reports now tell you when words in your document may not really be words. Lettering can end up baked into artwork — as part of a pasted logo or letterhead, or because text carried an effect like a drop shadow or a glow that gets flattened when the PDF is made. It still looks perfect on screen, but a screen reader has nothing to read, nobody can search it, and it can blur when zoomed. We found this on a real board agenda whose letterhead had quietly stopped being text, so the agency's own name was invisible to a screen reader. Both the plain-language step and the detailed report now point it out, explain which of the two causes you are looking at (only one of them is repairable), and show you how to check in ten seconds: try to select the words — if they will not highlight, they are artwork. What gets checked
ImportantAugust 27, 2026
Every report now says plainly what this tool can and cannot check. It finds some accessibility problems — not all of them: checkers like this one, including the ones built into Adobe Acrobat and Microsoft Word, catch only about 30–40% of the problems in a document, and the rest can only be found by a person opening the file and looking. So a good score means the document passed the checks a computer can run; it does not mean the document is accessible. That now appears on every report at every grade, not just high-scoring ones, with the studies behind those figures linked so you can read them yourself. Nothing about the checks or the scores changed — this is about being straight on what a score can and cannot tell you. What gets checked
ImprovedAugust 26, 2026
Big documents no longer look stuck while they're being checked. Complex, design-heavy PDFs can genuinely take up to a minute — the two veraPDF engine passes are thorough — and the waiting screen used to show one frozen line the whole time. It now walks through the actual checks being run, counts the seconds, and explains what's taking long and why, and the upload area says up front that analysis isn't instant. Nothing about the checks themselves changed — same audits, same scores, just an honest view of the work while it happens. What gets checked
NewAugust 26, 2026
Every PDF report now includes a WCAG second opinion. The veraPDF checker that already validates the PDF/UA standard on every PDF audit now also runs the industry's machine-testable WCAG 2.2 rules — the same subset professional checkers like PAC verify by machine, including text contrast, which this tool's score deliberately doesn't compute. It appears as its own panel on PDF reports: it can flag extra things worth fixing, it plainly says "Did not run" if it ever can't run, and it never changes your score or grade — every previously checked document keeps its exact result. See what's machine-checked
ImprovedAugust 26, 2026
Word and Excel checking just got meaningfully deeper. Colors that come from a document's theme — which is how most Office text is actually colored — are now resolved and checked for contrast, where the checker previously stayed silent on anything not written out explicitly; the same goes for Excel's legacy color palette. Reports also now disclose what a human reviewer should look at: languages used in the document, floating objects whose reading position isn't guaranteed, merged table cells, form fields, and hidden sheets. As with the last release, every change was adversarially reviewed before shipping — ten weaknesses found that way, including a table of contents miscounted as a form control, were fixed first — and every previously checked document keeps its exact score. See what's machine-checked
ImprovedAugust 26, 2026
The PDF checker now sees through three ways a document can look fine while failing its readers: text that displays correctly but extracts as unpronounceable symbols, visible text painted outside the document's reading structure, and form fields that structure never points at — each now scored and explained with the right fix steps. The Matterhorn checklist also moved up the front page as a collapsible section, so you don't have to scroll past everything to find it. Before release, every recent change was deliberately attacked — hostile files, forged data, misleading edge cases — and thirteen weaknesses found that way were fixed first, including one that could have wrongly accused a correctly built form. As always, every previously checked document keeps its exact score; only documents that actually have these problems are affected. See the Matterhorn checklist
NewAugust 26, 2026
PDF reports now include "Your document against the Matterhorn checklist" — a collapsible section showing how your document's findings land on each of the 31 checkpoints of the PDF industry's standard test model, the same checklist explained on the front page. Each checkpoint says, in words, whether issues were found, nothing was machine-detected, human review is needed, or the check did not run — never a tally or a pass/fail stamp, because this view adds nothing to your score and takes nothing from it. It is another way of reading the findings your report already contains, useful when working with remediation vendors who follow the PAC checker or the Matterhorn Protocol. Reports you shared earlier show the new section automatically. How the checklist works
ImprovedAugust 26, 2026
The PDF checker now catches more on its own: formulas without a spoken alternative, footnotes screen readers cannot link to their references, heading labels that mix two conventions, language settings that are not usable codes (like "english" instead of "en-US"), and embedded audio, video, scripts, or switchable layers — each either fixed into the score or flagged for human review, never silently skipped. Every previously checked document keeps its exact score; the new checks only affect documents that actually have these problems. The Matterhorn checklist on this page was also rewritten in plain language, answering why it has that name, what the veraPDF second-opinion checker is, and — plainly — that none of this changes how your score is calculated. See the Matterhorn checklist
NewAugust 25, 2026
The front page now shows the Matterhorn checklist — all 31 checkpoints of the PDF Association's test model for PDF accessibility, the same one professional checkers are built on — and which layer of this tool checks each one: our own audit engine, the open-source veraPDF validator that runs alongside every PDF audit, or human review for the checks no software can make. PDF reports also became more candid about that validator. If it ever cannot run, the report now says "Did not run" in its PDF/UA panel instead of leaving the panel out, because a missing check should never look like a passing one. Nothing about scoring changed, and nothing new is collected or stored. See the Matterhorn checklist
ImprovedAugust 25, 2026
The status page's new "do documents improve?" figures now count public audits only. The automated fleet inventory re-scans the same unchanged documents on a schedule through an internal tier, and those runs were drowning out the picture of documents people actually fix — the typical score change read as zero because most re-checks were fleet re-scans of files that had not changed. Fleet runs are no longer counted there, and neither are older records written before the request tier was recorded, so the figures build from that date. The count of distinct documents checked is unchanged and still includes everything. The data-retention policy (now v1.14) records the clarification in its change log. See the status page
NewAugust 25, 2026
The status page now shows whether documents improve. Two new figures are published there: how many distinct documents were checked (not just how many checks ran), and — for the last 30 days — how many documents were checked more than once, how many of those improved, and the typical score change among re-checked documents. Every figure is a count or a median computed from the usage records the data-retention policy already describes; no file name, content fingerprint, or individual score is shown. When fewer than five documents qualify, the rates and the median are withheld rather than describe a single visitor's documents, and the counts remain published like every other figure on the page. The data-retention policy (now v1.13) records the addition in its change log, and the release was security-reviewed before publishing, with the review recorded in the security log. See the status page
ImprovedAugust 22, 2026
The service now keeps a daily record of what it did, on its own server, for the people who audit it. Each night the previous day's usage records — which documents and pages were checked, with what result — are written to one file, kept for the same 365 days as the records themselves and then deleted. Audits that could not be completed are now recorded too, with a one-word reason and never the error text. The service also keeps its own error log for 30 days so a fault can be diagnosed quickly. These records hold the same fields the usage record always has — nothing about who made a request — and nothing is downloadable from this site; the data-retention policy (now v1.12) describes exactly what they hold. Read the data-retention policy
SecurityAugust 20, 2026
The software this site is built on has been updated to close published security flaws. These were in third-party code the service depends on rather than in the accessibility checks themselves, so nothing about how your documents are checked, stored, or deleted has changed — and no document was affected. One of the flaws was reachable on the live site rather than only possible in theory, which is why this update was made promptly and is recorded here. The full technical account is in the security log, and the plain-language version is in section 10 of the data-retention policy. Read the security review log
ImprovedAugust 20, 2026
Long updates in this notice are now shortened. This box used to print an entire update — sometimes a dozen lines — directly above the upload area, which pushed the tool itself down the page. It now shows the opening few sentences and offers a link to read the rest. Nothing is lost: the What's New page still carries every update in full, word for word, and it is linked from the header and the footer of every page. Updates short enough to fit are shown whole, with no link.
ImprovedAugust 20, 2026
PDF reports describe links and images more accurately. A link's text is now read from the link itself rather than from the line of text around it, so a vague link such as “here” is caught even when the rest of the sentence is descriptive — and a link split across two lines is no longer flagged by its first word. Links that have no tag at all (common for links inside Word text boxes — screen readers simply never reach them) are reported as a finding of their own, with the page number. The reading-order check names the pages that read out of sequence. And when an “image” missing alt text is really a box of text — a sidebar, a timeline, a chart title — the report now says so and tells you to change its tag instead of describing it, because alt text on a box of text would hide the text from screen readers. Saved reports keep their original wording until the document is checked again.
NewAugust 18, 2026
Web-page audit reports have their own report page now. Alongside documents, this service also audits web pages (for example, the pages that link to a PDF) and hands back a shareable report link — but those links opened a “Page not found” error, because the report page itself had never been built. It exists now: score and grade, every issue found with the exact elements affected and how to fix each one, plus the checks that still need a human eye. Links shared before today work too — including ones generated months ago — with nothing to re-run.
ImprovedAugust 17, 2026
The reading-order check is fairer to documents with images. Programs like Excel and Word draw a page's images last, no matter where the image belongs on the page — and the check was counting that drawing order against otherwise perfectly organized documents (a logo at the top of the page, for example). The comparison now judges only the order of the text, which is what a screen reader actually reads; genuinely out-of-order text is still flagged exactly as before. Saved reports keep their original evaluation until the document is checked again — if a past report docked reading-order points and the document has images, re-running the audit shows the corrected result.
ImprovedAugust 17, 2026
Checking several files at once is easier to follow. You can now upload up to five files together (the limit was three), and when they finish, each file gets its own clearly visible report card at the top of the page — grade, score, and filename — so switching between reports is one click. Previously the row for switching files was easy to miss, and some visitors thought only their first file had been checked. Nothing about the checks themselves changed.
ImprovedAugust 17, 2026
The font-embedding check now agrees with Adobe's preflight tools. Some documents were losing points for “fonts not embedded” when the flagged fonts never actually display any text — one common case is a font used only to draw the spaces between words, another is leftover font entries that Acrobat's own repair process leaves behind without any page using them. Neither can affect how a document looks or reads aloud, so neither is flagged anymore. A font that genuinely displays text without being embedded is still flagged exactly as before. Saved reports keep their original evaluation until the document is checked again, so if a past report flagged fonts, re-running the audit is the way to see the corrected result.
ImprovedAugust 16, 2026
Scores from saved audit results now always match their report page. When a document that was audited before is checked again — usually by an automated inventory — the tool answers from its saved copy, and that answer used to carry the score computed on the day of the first audit, even though scoring has been refined since. Those saved answers are now re-scored under the current rules before they're returned, exactly as report pages already do, so a score in an inventory or spreadsheet and the report it links to can no longer disagree. If an older exported list still shows a different number, the report page's score is the one to trust — re-running the check refreshes the list.
NewAugust 16, 2026
Fix instructions now match the program your PDF was made with — and every report shows what that program was. If a PDF records that it was laid out in Adobe InDesign, as many annual reports are, the action plan now gives real InDesign steps instead of Word menus, verified against Adobe's current documentation. A new “About this document” card above the plan shows everything the file records about itself — the program that made it, the author, when it was created and last changed, the page count — with “Not set” where the file is silent. The card ends with a plain sentence naming which program the fix steps are written for and why. How the audit works
NewAugust 16, 2026
Fix instructions now match the program your PDF was made with. Many documents — annual reports especially — are laid out in Adobe InDesign rather than Word, and until now the report's “fix the source document” steps assumed Word menus. When a PDF records that it came from InDesign, its action plan now walks through the real InDesign workflow — tagged export, heading styles, image descriptions, reading order — verified against Adobe's current documentation. Everything else, and every older report, reads exactly as before. How the audit works
ImprovedAugust 15, 2026
After an automatic repair, the results now account for every issue the audit found: each one is marked fixed, improved, or — stated plainly — no change, with its before-and-after score, so nothing the repair couldn't touch quietly disappears. Fix steps also now describe the problem that was actually found: a document whose only text issue is unembedded fonts is told to embed the fonts, not to run scanned-document recovery on perfectly readable text. And the audit report and the repair results use the same plain-language name for each finding, so the two pages can be read side by side. How the audit works
ImprovedAugust 14, 2026
Reports with a high score now say, right under the grade, what accessibility professionals keep having to explain: even a high score is not a guarantee. Automated checks can confirm accessibility structure is present — they cannot tell you the document truly works with a screen reader, because that is a judgment only a person can make. The new notice shows the two halves of the job side by side — the half software finished, and the half that always needs a person — along with how many accessibility criteria were never machine-checked at all for your document. It appears wherever the score does: on the report, the printout, the downloaded file, and the shared link. Lower grades carry a one-line version of the same reminder, so a person is in the loop at every grade. How the audit works
PrivacyAugust 14, 2026
The site now counts page views with Plausible, a privacy-friendly, open-source analytics tool that ICJIA runs on its own server — no commercial analytics provider, no ad network, no tracker. It sets no cookie and stores no IP address, and it cannot recognize you tomorrow or follow you to another site: visits are linked only within a single day, by a code that is scrambled every 24 hours. It sees which pages are visited — never anything about the documents you check, because audits don't pass through it at all. Exactly what is recorded, where it lives, and for how long is documented in the data-retention policy, now at v1.7. Read the data-retention policy
ImprovedAugust 11, 2026
Every fix step in every report has been re-checked, word for word, against the current versions of Microsoft Word and Adobe Acrobat (verified August 2026). Adobe redesigned Acrobat's entire menu system in 2023, so each Acrobat step now shows the current path first with the older “classic” path in parentheses — whichever version you're on, the step matches your screen. Each card of fix steps also names the exact app versions it was written for and tells you how to recognize which Acrobat you have. And if a menu item still doesn't match what you see, the card says who to contact — IDS at ICJIA — to get your software brought current. How the audit works
PrivacyAugust 9, 2026
The sign-in system has been removed — and it was never actually in use: login was never switched on, so nobody ever needed an account and no audit was ever tied to one. With the tool now in wide use, that unused machinery was deleted outright rather than left sitting in the code — it stays free and open, with nothing to register for. The service also no longer stores who uses it: the database columns for email address, IP address, and browser identifier were deleted, along with the data they held. What an audit leaves behind is metadata about the file — its name, score, grade, and date — never the file, and now about nobody. The full accounting is in the data-retention policy, at v1.6. Read the data-retention policy
NewAugust 9, 2026
Reports now show your document's full heading outline — every heading with its actual text, for both PDF and Word files. Word reports also list any paragraphs that only look like headings (bold, large text) so you can restyle them with real Heading styles. And the technical detail on each report card starts open in the detailed view, so the specifics — which image, which heading, which table — are visible without extra clicks. Each card's toggle still collapses the detail if you prefer the summary. How the audit works
ImprovedAugust 8, 2026
Scores are now calibrated strictly to the accessibility standard: style-guide advice (like “use exactly one top-level heading”) still appears on your report, but no longer lowers your grade. If you re-audit a document, its score may be a little higher than an older shared link of the same file — nothing changed about what passes or fails the standard. Reports are also more upfront about limits: they now list more of what no automated tool can check, and they warn you when every image in a document is marked decorative, since screen readers skip those silently. How the audit works
NewAugust 8, 2026
Every report now has a big “Printer-friendly action steps” button. It opens a clean page in a new tab with your fixes written out in full — each one showing both how to fix it in the original Word or PowerPoint file and how to fix it in Adobe Acrobat, so whoever ends up doing the work has what they need. Print it, or save it as a PDF, and work from it beside the document instead of keeping this page open. The same button appears after auto-remediation, where it lists what the automatic fixes could not repair. Reports also always open in the visual step-by-step view now, and the Visual/Detailed chooser above every report is much easier to find. How the audit works
NewAugust 7, 2026
Every report now ends with a checklist of what a person still needs to check by hand — including reports that score 100. Automated checks can confirm that accessibility structure is present, but not that it is right: a picture described as “image” passes the alt-text check and tells a screen reader nothing. So each check your document passed now comes with the one judgment the tool could not make. Nothing in that list is a failure — it is the difference between a document that is structured correctly and one that actually reads well. How the audit works
ImportantAugust 7, 2026
How scores and grades are calculated has changed, and many documents will score differently than before. A document is no longer punished for being simple: checks that do not apply to it — table markup in a document with no tables — now count as passing rather than being dropped. A document's score can no longer rise above what its worst unresolved problem allows: a minor item holds it at 89, a moderate problem at 79, a critical problem at 69. The letter still comes from the same scale it always has (90 an A, 80 a B, 70 a C, 60 a D), so the number and the letter always agree. Reports you shared before today show their corrected score when reopened. How scores are calculated
ImprovedAugust 7, 2026
The report headline is clearer. Because the letter grade is a weighted average while the publish-or-not verdict counts blocking issues, a document could previously be described as "Excellent" and "not ready to publish" in the same sentence. Now, when something blocks publication, that leads: "Not ready to publish — 2 critical issues." The fix steps also read more plainly — jargon like "structure tags" and "OCR" is explained where it appears, and the Acrobat instructions now say plainly that they need Acrobat Pro, not the free Reader. Two categories that genuinely cannot be fixed inside a PDF now say so instead of pointing you back to a source file you may not have. See what the audit checks
NewAugust 7, 2026
Audit reports now open in a visual, plain-language view: your grade front and center, a color-coded severity summary, and a numbered action plan that walks through fixes one step at a time — written for document authors, not developers, with instructions for Word, PowerPoint, Excel, and Acrobat. Nothing was removed: the Visual/Detailed toggle at the top right of any report switches to the complete technical report, and your choice sticks on your device. Downloaded HTML reports and printouts carry the same new layout with every detail expanded. How the audit works
PrivacyAugust 5, 2026
Shared report links have always stopped working after one year. Starting today, the stored report itself is also deleted: the service's cleanup sweep now permanently removes a shared report's data about 30 days after its link expires, so nothing you shared outlives its usefulness by more than a month. The data-retention policy (v1.4) states the exact lifecycle, and its § 8a explains precisely what a shared report contains. Read the data-retention policy
TransparencyAugust 5, 2026
The data-retention policy now includes a dated, line-by-line verification of what is and isn't stored — every database table and every write path checked against the public source code, with the evidence published as its own section (§ 8a). The verification also tightened one statement rather than quietly rewording it: a report you choose to share can quote short strings from your document (such as image alt text and link labels) inside its findings; a plain audit stores none of that. Separately, the service's usage records are now backed up nightly on the server, integrity-checked, keeping only the five newest snapshots — and the status page now shows when the last backup completed. Read the storage verification (§ 8a)
SecurityAugust 5, 2026
The whole tool has had a fresh independent security review — attack surface, sign-in, file handling, the checking engines, and what the service publishes about itself. Nothing critical or high-severity was found. Two low-severity issues were identified and fixed the same day, both about how an uploaded file's name is stored internally; neither exposed any document you have uploaded, and no report or stored data was affected. The full write-up, including what was checked and found sound, is in the project's public security log. The data-retention policy itself was reviewed end-to-end the same day and brought fully current (policy v1.2): it now states plainly what the usage log records and that entries are deleted after 365 days. Read the data-retention and security notes
ImprovedAugust 5, 2026
The status page now shows what kinds of file were audited — PDF, Word, PowerPoint and Excel — alongside the scores. It also clears up a genuinely confusing label: two different columns were both called “other” and meant opposite things. One now reads “Unrecognized extension”, meaning the document was checked normally and we simply could not tell its file type from its web address, and the other reads “Other file types”, meaning the upload was refused and could not be checked at all. View the status page
NewAugust 4, 2026
The status page now also shows what people bring that the tool cannot check at all — the older Office formats and CSV data files — over the last 30 days and since the tool launched. It sits below the grade distribution and is counted entirely separately from it, because a file that was refused was never assessed: it has no score, so including it among the graded documents would misrepresent both numbers. One thing to read carefully: these are attempts rather than documents, so the same file uploaded twice counts twice. View the status page
ImprovedAugust 4, 2026
Uploading an older Office file — .doc, .xls, .ppt or .rtf — now gets you a real answer instead of a list of what the tool accepts. These formats cannot store the headings, alt text and table information an accessibility check looks for, which is why Word and Excel disable their own accessibility checkers for them too. The tool now names the format, gives the Save As steps, and is clear that converting brings your content across but not accessibility structure — so expect to still add headings and alt text afterwards. It also recognizes an old file that has simply been renamed with a newer extension. Spreadsheet exports (.csv and .tsv) get a different explanation on purpose: there is nothing in a CSV for this tool to check, and that is not a fault. For raw data CSV is often the right format, and its accessibility depends on the page that links it rather than on the file itself.
NewAugust 4, 2026
The status page now shows how documents checked here have actually scored — how many earned an A, a B, and so on — over the last 24 hours, the last 30 days, and since the tool launched. It is a quick way to see how common serious accessibility problems really are. One caveat matters enough that it is printed on the page itself: these are files visitors chose to upload, often because they already suspected something was wrong, and the same file may be uploaded more than once. So the figures describe what people bring here to check, not the overall state of any organization’s documents. View the status page
ImprovedAugust 4, 2026
If you click a link while a document is still being checked, the page now asks first — leaving cancels the check and discards the report. You will only see this while a check is actually running; ordinary clicks are unaffected. The status page also has a link back to the audit tool now, so it is no longer a dead end, and the “Status” link in the header opens in this tab instead of a new one. View the status page
ImprovedAugust 3, 2026
The status page at /status is now readable without any browser add-on. It shows the same information as before, laid out as a colour-coded, collapsible outline instead of a single block of unformatted text, with a link to view the raw data if you prefer it. Nothing about what is published has changed — still only totals and yes/no answers, with no file names, no email addresses, and nothing identifying anyone who has used the service. View the status page
ImprovedAugust 3, 2026
The top of the page is simpler. There is now a “Status” link showing whether the audit service and each of its checking engines are running, and the “Analyze” link has been removed — click the site title on the left to clear your results and start a new file. The title works with the keyboard as well as the mouse, which the old version did not. View the status page
NewAugust 3, 2026
There is now a status page at /status showing whether the audit service is running, whether each of its checking engines is working, and how many documents have been audited — broken down by PDF, Word, PowerPoint and Excel. It is plain data, updated continuously, for anyone who wants to confirm the tool is healthy or see how much it is being used. It contains only totals: no file names, no email addresses, and nothing identifying anyone who has used the service. Past updates like this one are now kept under “What’s New” in the header and footer, so dismissing this notice no longer loses them. View the status page
ImprovedJuly 26, 2026
Both PDF/UA panels now sit below the issues you need to fix, and neither can be mistaken for a clean bill of health. The “PDF/UA-1 signals” card used to appear at the very top of your report, right under the score — above the critical issues. PDF/UA-1 essentials are structural markers (is the file tagged? are the fonts embedded?); they cannot tell whether your alt text is meaningful or your reading order makes sense, which is what the accessibility grade measures. So a document can meet every PDF/UA-1 essential and still have critical issues that block publishing — and when that is the case, the card now says so directly.
ImprovedJuly 26, 2026
Your report now leads with what actually blocks publishing. The critical issues and their fix steps appear directly under your score, above the technical PDF/UA-1 (veraPDF) panel — which is informational. That panel can show a green "Pass" on a document that still has critical WCAG issues to fix, because the two checks answer different questions: PDF/UA-1 verifies the file's formal tagging, while your WCAG grade is what decides whether people can actually use the document. When both are true, the panel now says so in plain language instead of just showing a green tick.
ImprovedJuly 26, 2026
Accuracy and integrity fixes to how documents are judged — every one of them catching a real barrier the audit used to miss. A PDF carrying an EMPTY set of accessibility tags (the structure was there, but nothing was inside it) was being reported as properly tagged with no detected failures; it is now treated exactly like an untagged file, because that is what it means for someone using a screen reader. Images that were never tagged at all — invisible to a screen reader, and worse than an image with a missing description — now count against the score instead of being skipped. Re-auditing a document may therefore give a different score than before: reports you saved earlier keep the score they were given at the time, and the fresh audit is the correct one.
NewJuly 22, 2026
Audit results now include a PDF/UA-1 (ISO 14289-1) machine-check verdict from veraPDF — a Pass/Fail badge with the exact technical checkpoints that failed, shown alongside your accessibility grade. It's the automatable equivalent of the checks in Adobe/PAC-style PDF/UA tools, covering the machine-verifiable requirements (full conformance still needs a human to confirm alt-text quality and reading order).
ImprovedJuly 22, 2026
PDF scoring accuracy fix: some PDFs — often those exported from design tools like Adobe InDesign — carry leftover 'phantom' tags (images, lists, and tables) that aren't part of the document a screen reader actually reads. The audit was counting those as real content — flagging images as missing descriptions, lists as 'incomplete structure', and tables as broken — and unfairly lowering the score. The tool now ignores tags that aren't connected to the live document structure, and recognizes when a page's images are deliberately marked as decorative (and need no description), so well-built documents are no longer penalized.
ImprovedJuly 19, 2026
Major accuracy upgrade across all four formats: the audit now catches PDFs whose security settings block screen readers outright (they used to pass), checks real background colors before judging contrast (no more false alarms on white-on-dark headers or designed slide layouts), reads Excel link text from the linked cells, recognizes custom Word heading styles and per-run slide languages, audits Word headers/footers/footnotes and Excel chart sheets, and reports 'needs manual review' instead of guessing whenever the evidence can't be resolved from the file.
NewJuly 2, 2026
Now supporting Microsoft PowerPoint (.pptx) and Excel (.xlsx) files — upload a presentation or workbook for the same WCAG 2.1 AA accessibility audit as PDFs and Word documents, with findings and fix guidance tailored to each app.
NewJuly 1, 2026
Now supporting Microsoft Word (.docx) files — upload a Word document for the same WCAG 2.1 AA accessibility audit as PDFs, with findings and fix guidance tailored to Word.

Back to the audit tool