How We Test Our Tools

The short version

Each tool is run automatically in a real browser (headless Chrome) against the same build that goes on the website. A script feeds it real files, waits for the download and then inspects the file it produced. A tool only "passes" if its output is correct, not just if it finishes without an error.

1. Real test files

Before each test run, a script creates a fresh set of test files so that results can be repeated exactly:

• PDFs with known content: 2-page, 3-page and 120-page documents, a landscape document and a password-protected PDF;
• photos of 1200 × 800 and 4000 × 3000 pixels in JPEG, PNG and WebP;
• a logo on a plain white background, a portrait for passport photos and an image of printed text for OCR;
• problem files: an empty file and a text file renamed to .pdf.

2. Checking the output, not just the screen

For each tool the script saves the real download and checks it with separate software:

• PDF tools: page count, page order, rotation and the text on each page, read with independent PDF libraries. For the password tool we confirm the file refuses to open without the password and opens with it.
• Image tools: format, pixel dimensions and file size; for the background remover, the transparency of specific pixels.
• QR codes: the downloaded PNG is decoded and must return exactly the link we typed.
• OCR: the recognised text must contain the sentence in the test image.
• Converters: known conversions such as 1 mile = 1.609344 km and 100 °C = 212 °F must appear exactly.
• Generators: JSON exports must contain what we typed, and "Print / Save as PDF" must open the browser's print dialog.

3. Things that go wrong

Each file tool is also given an empty file, a fake PDF, a password-protected PDF or a non-image file. We check that a clear message appears instead of a silent failure or a broken result. We also load every tool at a 375-pixel-wide phone size, check nothing spills sideways off the screen, and check the tool can be reached with the Tab key.

4. Watching the network

ToolMany says your files are processed in your browser. To check that, the test records every network request made while a tool is working. Any request that carries data (an upload) is flagged. In the latest run no tool sent any request with a body. The OCR tool downloads its recognition engine and language data from a public CDN the first time it is used; that is a download of code, not an upload of your image.

5. What testing found and changed

Testing is only useful if it changes things. Problems found in the latest round and fixed:

• the PDF password tool did not really encrypt; it was rebuilt with AES-256 encryption;
• two PDF tools depended on loading a library from a third-party CDN and could fail on slow connections; they now use a bundled copy;
• printed passport photos came out at the wrong physical size, and the download link could vanish;
• unit conversions used rounded factors, and several tools accepted broken files without any message;
• the PDF merger accepted password-protected and damaged files without warning, and listed unreadable files as having one page.

Tool pages now describe the limits we actually observed, and numbers quoted on them (file sizes, dimensions) come from these tests.

What the tests do not cover

• Real phones, Safari and Firefox: the automated tests use Chrome, with a phone-sized window for layout.
• Printed output: we check that the print dialog opens, not the paper result.
• Very large files: the largest inputs are a 120-page PDF and a 4000 × 3000 photo. Bigger files depend on your device's memory.
• Your exact requirements: a passport office, tax authority or employer's system may have rules no general tool can check.

If a tool does not behave as its page describes, please tell us. See also our editorial policy.