The counter uses PDF.js to read the text layer already stored in each page, then applies Unicode-aware text segmentation in your browser. The per-page table helps you spot unusually dense pages, empty pages, or sections that may have been missed.
PDF files describe positioned text fragments rather than the paragraphs you see on screen. Headers, footers, tables, columns, and references are included when they exist in the text layer. The tool does not guess which sections a publisher or university excludes from an official limit.