Accessibility vs. compliance (reprise)

7 June 2021 | Ted Page

This post is a response to a recently published article that argues for privileging standards compliance over accessibility when remediating PDFs (What to do when a screen reader doesn’t read a remediated document correctly,, 2 June 2021).

The CommonLook article asserts that the most important aspect of the process of PDF accessibility remediation is that “the document perfectly complies with the necessary standards.” (emphasis added)

Contrary to the above statement, we here at Add most definitely do not believe that standards compliance is even necessary for accessibility, let alone the most important consideration.

Add vs. ‘Cdd’?

Our company is called “The Accessible Digital Documents Company”. For good reason, it is not called “The Compliant Digital Documents Company”.

As will be seen, if you focus solely on compliance, your documents may well end up being less accessible than they otherwise might be.

When compliance and accessibility conflict

The CommonLook article’s central contention is that you should: “remediate to the standard, not the assistive technology.” The example it gives of a potential conflict between the two is as follows:

…many screen readers will say ‘Link’ when they land on a Reference tag, even if there’s not really a link there. But if the tester isn’t aware of this, it can be confusing and possibly flagged as an issue, even when it’s not.

But it is a problem

In our view, it clearly is a problem if a screen reader announces something as a link when it isn’t. This may happen, for example, if you use a (standards-compliant) Reference tag for an unlinked footnote reference. We at Add don’t do this, and never would, for the simple reason that it is, as the CommonLook article concedes, confusing. And, it adds nothing of value.

The above is actually a relatively trivial example. A far more consequential example would be that if you were remediating an exam paper and you used the (standards-compliant) Formula tag for any equations or other mathematical expressions that it might contain, that content would not be read at all by NVDA. There is a clear choice here: accessible or compliant. You can’t have both.

The long view vs. the short view, and when it really matters

If you are tagging the UN Declaration of Human Rights, and hence you expect the document to have a long shelf life, there could be an argument for privileging compliance over more immediate accessibility. I say “could” because, as the CommonLook article concedes, standards evolve over time, so what is compliant today may not be in a few years’ time. But, more importantly, when you are fixing an exam paper that has a shelf life of less than a day, the argument for immediate accessibility over compliance is more than compelling. It’s unassailable.

Not all documents, of course, are the UN Declaration of Human Rights or exam papers. However, based on more than 17 years’ experience of remediating PDFs, day-in day-out, I would argue that most of them have a short, or short-to-medium shelf life.

Remediating to a specific screen reader

Finally, the CommonLook article argues that “remediating to a specific screen reader can be detrimental to the overall accessibility of a document” (again emphasis added). I have no problem with this statement. However, it doesn’t help the author’s case at all, for the simple reason that the exact opposite is also true.

For example, if an author has used proper roman numerals (as opposed to the letters i, v and x etc) then these will work fine in JAWS but not in NVDA. Would I fix them (with Actual text) for NVDA? Yes, of course I would. Why? Because (1) such roman numerals are almost indecipherable in NVDA and (2) there is no downside to doing so. (See also, by the way, the Formula tag issue above—exactly the same principle applies).


Because of the differences in emphasis and approach—compliance vs. end-user focus, not all PDF accessibility remediation companies are the same. Those commissioning accessible PDF remediation should be aware that there are genuine differences in the results that each produces.