Editorial Policy

The short version

We do not write about a tool we have not run, or publish a number we have not measured. When a test produces an unflattering result, that result gets published too. Our own revenue has no influence on anything we write, and errors get corrected in public rather than quietly deleted.

We test what we write about

Every guide that describes a tool is written after using that tool on real files. When an article says an image went from one size to another, those are the actual bytes from an actual run, not an estimate and not a figure repeated from somewhere else.

Where a result depends on the input, and with images it usually does, we say what we tested: the source file, its dimensions, the settings used. A number without its conditions is close to useless, and publishing one invites readers to expect a result we never promised.

We show the limits, not just the wins

It is easy to publish only the flattering test. We do not. If compressing to a target degrades an image visibly at some point, the guide says where that point was. If a tool fails on a category of input, the guide says which category.

The same applies to our own tools. Several tool pages carry an explicit section on what the tool will not do. That is deliberate. A guide that only reports successes is marketing wearing a guide's clothes.

Sources and references

Where a claim comes from outside our own testing, such as a platform's required image dimensions or a browser's format support, we link to the primary source rather than to another article summarising it. We prefer official documentation, standards bodies and platform help pages, and we check a link resolves before publishing it.

Where sources disagree, which happens often with social platform dimensions, we say so rather than picking one silently.

Keeping guides current

Image formats, browser support and platform requirements change. Guides that depend on those facts are revisited, and a guide that has been substantially revised carries an updated date. Where a specification changed after we published, we prefer to update the article rather than leave a stale claim standing next to a fresh date.

We also remove or merge our own content when it stops earning its place. Overlapping articles that said much the same thing have been consolidated rather than left up to pad a page count.

Corrections

We correct errors. If a factual mistake changes the meaning of a passage, we fix the text and, where the change is material, note what changed rather than pretending the original never existed. Silently rewriting history is how a site loses the right to be believed.

If you spot an error, tell us through the contact page. Corrections are treated as bugs, not as suggestions, and they take priority over new writing.

Money does not influence editorial

This site runs no advertising at all. It is funded by affiliate commissions on a small number of tools we recommend, and that funding buys no influence over what we publish. No partner reviews an article before it goes live, no article is written to please one, and no product gets a favourable mention in exchange for anything.

We do not run sponsored posts, paid guest posts, or paid link insertions. If that ever changes, any such content will be labelled clearly at the top of the article. Where we mention a third-party product, it is because it is genuinely relevant, and we say plainly when a competitor does something better than we do.

Affiliate links

Some articles contain affiliate links, which earn us a commission if you buy something after clicking. They never change your price, and they never change our answer. Where the best recommendation is a free tool, an open-source project or a competitor that pays us nothing, that is what we recommend. Commission rate never affects the order of a comparison, and any article containing an affiliate link carries a disclosure box above the first recommendation. The full rules are in the affiliate disclosure.

How we use AI

We use AI tools in the course of building and writing, in the same way most software teams now do: drafting, editing, reviewing code. We are open about that rather than pretending otherwise.

What we do not do is publish generated claims. Facts, measurements and test results in our guides come from tests we ran, and a model is never the source of a number on this site. Illustrations that were generated or drawn rather than photographed are described as illustrations, and we never present a generated image as a screenshot of a real result.

Images in our articles

Screenshots are captured from the real tool doing the real thing described in the text. Photographs used to illustrate a point are either ours or openly licensed, and licence details are respected. Diagrams are original. We do not fabricate a before and after.

Who writes this

Articles are published under the ImgCruncher team byline. The people who write the guides are the people who build the tools, which is why the guides can say what a tool does badly. There is no outsourced content desk and no ghost-written filler.

What we will not publish

• An article about a tool nobody on the team has used
• A measurement we did not take
• A near-duplicate of an article we already have, written for a slightly different search phrase
• A screenshot that misrepresents what the tool actually produced
• Content designed primarily to earn a commission rather than to answer a question

Feedback

Corrections, disagreements and requests all go to the contact page. Related: about us, disclaimer.