Crossref Similarity Check URL points to PDF reader and not full-text URL

We applied for the IThenticate plagiarism service. They required us to have deposited full-text links for Similarity Check for more than 90% of your journal-level article DOI metadata.

However, we noticed that when depositing via OJS, the full-text URL included in the metadata points to the PDF.js viewer, not the PDF, resulting in an invalid link for the IThenticate crawler.

How do we modify the OJS crossref deposit metadata to automatically include the PDF link?

In our case, it should be this link:

Instead of this one:

What application are you using?