Inaccessible by Choice: PDF on MacOS

An accessibility expert pointed out the other day that VoiceOver, Apple’s screen reader for MacOS, doesn’t understand tagged PDF, and therefore doesn’t support accessibility in this extremely common electronic document format.

He’s right. Since there aren’t any screen reader options for the Mac that understand PDF tags, it’s fair to say that AT users on a Mac don’t have the same access to content as do Windows users.

Clearly, this is a problem. The interesting question: is it a PDF problem, or an Apple problem?

One way to answer this question is to simply compare Apple with NV Access, the non-profit developer of NVDA, a screen reader that does a great job with tagged PDF.

 AppleNV Access
Employees72,800 (2012)2
Sighted employeesYesNo
Revenue$156.5 billion (2012)Donations
Screen reader softwareVoiceOverNVDA
Operating SystemMacOS onlyWindows only
Software costComes with MacOSFree
Supports accessible PDFNoYes

If PDF files aren’t “accessibility supported” on MacOS it’s simply because Apple chooses this state of affairs.

Given that 80% of non HTML electronic documents are in PDF format, Apple’s choice is difficult to understand.

I love Apple hardware, but I’m deeply unimpressed that the sometime most valuable company in the world is so blinded by the “not invented here” syndrome that it prefers not to make the modest investment required to make PDF files accessible with its otherwise excellent screen reading software.


11 comments

  1. March 3, 2013 at 11:43

    Amen, brother.

  2. March 4, 2013 at 13:55

    Baffling, to say the least. Perhaps it’s another one of those: “.. it’s a feature!” for Apple users.

  3. March 6, 2013 at 07:45

    It is interesting how little the users of VoiceOver seem to care though.

    I listen to podcasts from VoiceOver users, and it rarely comes up as an issue. I.e. they can read, get around, and use PDFs when they need to.

    In a quick test I couldn’t see any different between a tagged an non-tagged PDF in VoiceOver, but then again, how many tagged PDFs would people come across anyway?

    I’m not saying it’s right to ignore the accessibility features of PDF, but perhaps it simply isn’t a big enough issue for their customers to get it up the priority list.

    Flash used to be the big issue as that was a black-hole for VoiceOver, but the use of flash is diminishing, so that is fading.

  4. March 6, 2013 at 09:19

    Thanks very much for the interesting comment, Alistair. It’s a point that deserves some study.

    Clearly, there are many types of PDFs for which a screen reader like VoiceOver cannot work even in theory. That includes (among other examples):

    • PDFs made up of scanned pages without OCR
    • Forms (documents that include text and form-fields together)
    • …and of course, the big kahuna: PDF files with content ordering that doesn’t resemble the logical order, which is a LOT of PDFs (although not the simplest MS-Word generated files)

    Then there are other features that current-generation VoiceOver users are guarenteed to be without, examples include:

    • Alt. text for graphics
    • Table and list structures
    • Headings

    VoiceOver doesn’t treat tagged and non-tagged PDFs differently for the simple reason that it ignores tags. As such, it’s natural to expect the same performance on tagged and untagged files – in both cases, no access to structural and other information that’s considered necessary for accessibility in other types of content.

    Many AT users (on Windows, admittedly) do use software that does read tags such as JAWS and NVDA – and these users are well-aware that “some PDFs are crap, some work very well”. Of course, as yet the “good” PDFs are still quite rare.

    Indeed, PDF isn’t a “black hole” for VoiceOver, as is Flash. If users have been trained to expect *nothing* from PDF files then more-than-nothing (which VoiceOver certainly delivers) is definitely an improvement.

    PDF can fail accessibility in very subtle ways; far more subtle than HTML. Example: an image without alt. text stands out like a sore thumb on a web page. The AT knows it’s there, presents it to the user.. but then the user gets the file-name instead of any alt. text. They know right away that something is wrong.

    PDF is different. If you try to read a PDF that contains images using a screen reader (like VoiceOver) that doesn’t understand tags then you’ll never even be aware of what you’ve missed (unless the images are referenced in the text, I suppose). Even if your screen reader DOES use tags images may still be “hidden” if they’re (erroneously) marked as artifacts.

    In that sense AT users have a tough time evaluating PDF files or tagging (or software) for quality because in many cases they may not be able to tell what they’re missing.

    That said, I would strongly support a full-up study of how AT users interact with PDF content!

  5. March 6, 2013 at 10:30

    It seems there are probably three main categories of PDF for this discussion:
    1. Those inaccessible to any AT user, at least without significant work (e.g. scanned docs)
    2. PDFs without tags (equivalent for Mac and Windows AT)
    3. PDFs with tags (same as 2 for Mac, much better on Windows AT)

    This is similar to my old ‘4 levels’ article (http://alastairc.ac/2006/07/the-four-levels-of-pdf-accessibility/) but I’ve skipped level 4 in this context.

    I’d argue that 2 is probably the biggest category, but 3 (or 4) should be the aim for organisations.

    A study would be useful, especially if it tested *whether* people would use PDFs they come across, as well as how they interact with them. I suspect in a lot of instances there are alternatives (e.g. HTML), or other ways to achieve what they want without using the PDF, or just using it enough to get an answer.

    VO users are probably used to the general crapness of PDFs (from their point of view), thus propagating the general perception that PDFs aren’t (very) accessible.

    I’ve had clients with accessible PDFs get complaints simply because they used PDFs, the AT user complaining hadn’t even tried it before complaining!

    So whilst I agree Apple should implement better support of tagged PDFs, I think the blame still gets laid at Adobe’s door.

  6. March 6, 2013 at 10:48

    Agreed that for the purposes of this discussion the distinction between well-tagged and auto-tagged PDF files isn’t germane.

    If the user won’t even try the file… what can one say? Certainly, it’s hard to blame the technology (never mind the company who invented the technology) for that!

    I would certainly encourage any organization that goes to the effort of ensuring their PDF files are properly tagged to brag on their accomplishment.

  7. August 11, 2014 at 21:29

    Over another year has gone by, and what Duff lays out here so well is STILL so.
    And while we’re on the topic, why don’t apps have to be accessible to get into the Apple Store?

  8. March 27, 2015 at 13:51

    I believe that readers are misled in this article with respect to pdf accessibility in general on the Mac when using VoiceOver. When attempting to use VoiceOver with a pdf in Adobe Reader, the pdf content area is non-existent. Whereas, using Apple’s Preview or a third party pdf reader like PDFPenPro, the same pdf content area is perfectly readable by VoiceOver. So, VO users do have options for reading pdf’s on the Mac and may be one of the factors why you don’t hear a lot of chatter about the accessibility of pdf’s to VoiceOver users. A scanned image pdf is useless to any screen reader unless some sort of OCR is performed, no matter the platform.

  9. March 27, 2015 at 14:26

    Thanks for your comment, Tim.

    As it stands, VoiceOver on Mac OS X does not support Tagged PDF, and thus does not use the logical reading order and semantic structures essential to accessibility.

    The “options” you refer to are just text extraction – no tables, headings, lists, navigation, etc. That’s not accessibility.

    Oddly enough, Adobe’s just announced that the new version of their Acrobat and Reader software will include support for VoiceOver.


Leave a Reply