How We Test and Update Our Content

Our tests, comparisons and data studies follow the same rules. This page states them so that you can check whether we kept them. It belongs to AI Icon Creator, and we say plainly when a rule does not apply to a particular article.

Rules before results

For a test of tools, we write down the briefs, the scoring and the rules before the first run, and we publish them with the results. If we change a rule, we log the change with its date and say whether we had already seen a score. Our test of eight AI app icon generators shows what this looks like, including the changes we made along the way.

How we test tools

• Same inputs for everyone. Every tool gets the same briefs and default settings, and we keep the first result for each brief without re-rolling.
• Free tiers only. In comparisons we use what a tool offers for free and pay for nothing. When a free tier runs out, we say what we could and could not test, and we keep that tool out of the ranking.
• Terms first. Before we touch a tool we read its terms for clauses on benchmarking, competing products and automated access. If they forbid the way we would test it, we do not test it, and we say why.
• Scores from evidence. What can be read from files, screens and written terms is scored that way. The parts that need judgement are scored blind by more than one scorer, and we publish each scorer's marks.
• Our own product too. We test AI Icon Creator in the same way, disclose that we make it and list its weak points.

How we collect data

• Public sources. For studies of app icons we use public data, such as Apple's public chart feed, and we write down the date we collected it.
• Statistics, not other people's icons. We analyse icons and publish the statistics. We do not republish third-party icons.
• Method and limits. Each study explains how we measured, how we checked our measurements, for example with a second reviewer or a spot check, and what the numbers cannot tell you. When we can, we publish the data as a CSV file.

How we use AI

We use AI assistants to help run tests, process data and draft articles. The team decides what to test, sets the rules and is responsible for what we publish. Where an AI does a step that affects a result, such as reviewing or scoring icons, the article says so and says how we checked its work.

How we source facts

• Platform rules come from the platforms' own documentation, from Apple, Google and web standards, and we link it.
• Facts about other companies, such as prices, free allowances and terms, carry the date we read them on the company's own pages. We quote only short phrases from terms.
• Independent sources are labelled by kind, for example independent review, forum thread, user reviews or vendor blog, because a vendor's own blog is not independent.
• Reviewed, not tested. When we only read a tool's documentation, we write "we did not test it".

How we update and correct

• We recheck the prices, free allowances and terms in our comparisons every quarter, and we change a post's updated date when we change the post.
• When a change matters to the reader, we say what changed at the bottom of the article.
• If we got something wrong, email hello@aiiconcreator.com. We check it, correct the article and note the change.

What we do not do

• We do not accept payment for inclusion or ranking.
• Our tests and comparisons contain no affiliate links.
• We do not present a review of documentation as a test.
• We do not publish a claim about another company that we cannot source.