How we test what we publish
Last reviewed: November 2026
This page describes the method behind the guides on ImageTools Pro — what we actually run, what we read from the code instead of guessing, and what we cannot claim. If a sentence on this site sounds more precise than the work behind it, this page is the place to check whether that precision is earned.
1. Every guide starts with a real run
Before a tool page or guide is published, we open the tool in a desktop browser, load a real file and run the settings the article talks about. We record the output size in bytes — not an approximate percentage — and we look at the result to confirm nothing was cropped, stretched, blurred or flattened by accident.
The files we use are ordinary ones, chosen because they break things in different ways:
- phone and camera JPEGs (several megabytes, 12 to 24 megapixels, with EXIF and GPS data);
- full-page PNG screenshots (1920 pixels wide, flat colours, large file);
- a logo PNG with a transparent background;
- a multi-page PDF, to check page order and the page limit;
- an animated GIF, to see what a converter does with the frames.
Every number printed in a guide comes from one of those runs, with the file and the settings stated next to it, so you can repeat the same operation and compare.
2. Limits are read from the source code, not copied from another blog
Upload limits, accepted formats, page counts and quality defaults are read directly from the code that runs the tool. On this site that means, for example: a 20 MB ceiling on most tools, 25 MB on the EXIF remover, 50 MB on the target-size compressor; JPEG quality defaulting to 0.9 on conversion and 0.92 on resizing; the PDF-to-image tool rendering at most 50 pages; GIF input being read as its first frame only; and a transparent PNG turning into a white background when it is exported to JPEG.
Quoting the code rather than a marketing page is deliberate: when the two disagree, the code is what your file will actually experience.
3. What we do not test
We do not run an automated cross-browser suite, and we do not maintain a device lab. Our reference environment is Chromium on desktop; Safari, Firefox and mobile browsers are not systematically verified, and we say so rather than implying universal support. If a guide promises behaviour that depends on a browser feature, we test it where we can and describe the result as observed in Chromium.
We also do not claim third-party certification, an editorial board or years of professional experience. This is an independent project maintained by the people who build the tools; the about page says the same thing in fewer words.
4. How we check the privacy claim
The tools decode your file with the browser's own APIs and process it on your device. When we review a tool, we confirm that the image is read locally and that no request carries its bytes; the only download our background remover performs is its model file from a CDN, which contains no image data. That is why the same tools keep working when the network drops after the page has loaded.
This is also why we can be direct about the trade-off: local processing means the speed of your device sets the limit. A very large photo on an older phone will be slower here than on a service that uploads it to a server farm.
5. External facts carry a date
Figures that come from someone else — a platform's recommended image size, a format specification, a documented browser limitation — are only stated when we can point to the source, and the date we checked it is written next to the figure. Platform recommendations change; a figure without a date is a figure you cannot trust.
When a number cannot be verified, it is left out. We would rather publish a shorter guide than repeat a claim we cannot stand behind.
6. Review rhythm
Guides are revised when the tool changes, when a browser change alters the result, or when a reader reports a mismatch. There is no fixed publishing calendar and no word-count target: a page is rewritten because something in it became wrong or unclear, not to keep a schedule.
7. Reporting a mistake
If a step no longer works, a limit has changed or a sentence overstates what the tool does, write to us through the contact page with the URL and what you saw. Reports are read by the people who maintain the code, and a confirmed mismatch is fixed in the guide and, where needed, in the tool.