Paywalls are very porous. Depending on your field, you might find many paywalled papers available for free through Google Scholar. I know there are major universities, at least in Europe, that encourage their staff and students to use proxies to log into libraries like ScienceDirect for free.
In my experience I googled up shared PDFs of thousands of paywalled papers in the field of economics. I only encountered a handful of papers that weren't available for free somewhere online so I contacted the authors and they sent me copies of their papers for free.
I'm not aware of any data out there that would suggest that open access papers have a larger audience nowadays.
My intuition is that top paywalled journals have more readers than top open access journals, but average paywalled journals have lesser readership that average open access journals. Mediocre paywalled papers don't seem to generate enough interest for someone to share a copy online so they are the ones that end up truly paywalled.
Sorry for the lack of data references. I just wanted to point out that any data on this question should take into account the field, the rating of the journals, and readers who google up paywalled papers for free.