When we're talking about "how many visitors" we're talking about our Unique Devices data.  This is an estimate and the way it's computed restricts us to only knowing high level numbers at the project or project family level (like de.wikipedia or "all wikipedias").  So if an estimate for number of visitors to a specific subset of pages is required, we don't collect that data.

If what's needed is "how many visits", then we have Pageview data.  This is broken down per page, and available as a bulk download, through an API, etc.  So if you can compile a list of pages that have some feature (like equations for example), then it's possible to cross-reference that with pageview data and get the answer.  To compile this list in the specific case of equations, you may be able to use the templatelinks table of the project you're interested in.  These are mirrored to the cloud db replicas.  So if equations generally use templates, then you can search for pages with links to those templates in that table, and that would be the list of pages you're interested in.  With those page titles / page ids you can then query the pageview data.

Hope this helps explain a bit more our data, but feel free to follow up with questions.