# Predeclared protocol: scanned and non-fillable PDF filling

Declared before comparative browser runs on 29 September 2026.

- Publisher: Rust Solutions Ltd, owner of one tested product, No Fuss PDF. This is first-party comparative research, not independent certification or a security audit.
- Six task-matched providers: No Fuss PDF, SimplePDF, PDF24, Sejda, PDFescape, Smallpdf. Use the free public editor, signed out, without buying a plan or creating accounts.
- Two synthetic one-page documents with identical artwork: a flat PDF with page text but no interactive fields, and a PDF containing only a rendered image of that page. Neither has form fields. Independently verify the scan has no extractable text before testing.
- Same answer: `Alex Example`. Open the file, add the answer beside Name, finish/export, and capture the actual PDF download. Do not test automatic form detection as a requirement: adding visible text is a valid solution to this job.
- Same desktop Chromium version and 1440x1000 viewport. Each document gets a fresh browser context, accepting no optional cookies where a choice is available. No Safari/mobile comparative ranking or speed league table.
- Independently reopen outputs with Poppler/pdf-lib, check page count, dimensions/rotation and extractable answer text. Visually inspect independently rendered output for a visible answer and any watermark; extraction alone is not sufficient.
- Capture a screenshot at the last reached state and a bounded, sanitised request summary. Record observed PDF-byte/marker transmission and hosts; do not infer a universal privacy/security guarantee from an absence of detected bytes. Never publish raw HAR, cookies, signed URLs or full request bodies.
- Record account/payment gates only if observed in the actual workflow. A provider's advertised free tier is a vendor claim until a completed free export is observed.
- Network challenges, unavailable editor UI or an adapter timeout mean **not completed in this environment**, not “cannot fill scanned PDFs.” Do not bypass CAPTCHA, quotas or account gates. Keep partial runs and limitations visible. No retries to evade provider limits.
- Do not manufacture numerical overall scores. Group verified completions by this task, report ties and missing evidence, and recommend competitors when their observed result fits a reader's needs. Optional Unicode/rotation tests on No Fuss PDF are supplemental, not comparative scores.
- Publish the method, fixture hashes/files, reproducible harness, sanitised results, and permissible evidence. Raw counts describe the tested sample only. No “all PDFs”, “best overall”, “secure”, retention or data-residency inference.
- Outcome is relevant visits followed by tool starts and successful downloads, not the number of pages published. Keep research separate from the product interface; no public usage dashboard/admin link.