How we review AI tools.

Our AI and productivity coverage runs across three sites: AI Productivity Guides, Best Productivity AI, and AI Productivity Tools. Different formats, same rules. Below is how those rules work.

What we test by hand

A review does not run until someone has used the product for real tasks, not a five-minute click-through of the marketing site. That means creating an account, running the same set of representative prompts or workflows we use across competitors, checking the output for accuracy, and noting where the tool breaks, stalls, or quietly does less than the pricing page implies. Screenshots and examples in a review come from that session, not from the vendor's press kit.

For AI writing and research assistants specifically, we check the same core questions each time: does it cite sources correctly, does it say when it is unsure rather than guessing with confidence, and does the free tier actually resemble the paid one or exist mainly to get you to upgrade.

How we verify pricing

Pricing gets pulled directly from the vendor's own pricing page, not from a comparison chart somebody else built. Every price in a review is dated to the day it was checked, because AI pricing moves often and a number from three months ago can be flatly wrong. Where a vendor uses usage-based or credit pricing that is hard to summarize in one figure, we say so and show the math instead of forcing it into a clean number that misleads.

How often reviews get refreshed

Reviews get revisited on a standing schedule, and sooner than that when a vendor ships a pricing change, a major feature update, or a model swap that changes what the tool can do. A review that has not been checked in a while says so at the top, with the date of the last pass, rather than sitting there looking current when it is not.

Ad-supported, and what that means

These sites run on advertising and, on some pages, affiliate links. If you click through to a vendor and sign up, we may earn a commission. That relationship is disclosed on the pages where it applies, and it does not decide which tools get covered or how a review comes out. A tool that pays nothing and a tool that pays a referral fee are held to the same test.

No pay-for-placement

We do not sell better placement, a higher score, or a friendlier writeup. A vendor can buy an ad next to a review. A vendor cannot buy the review itself, and none of our editors take briefings that come with strings attached.

Who writes these

Every review carries a named author who has actually used the product and will answer for what the piece says. Our contributors and their beats are listed on the authors page, along with the editor who reviews the work before it publishes.

A timing note worth flagging

Readers sometimes ask why our AI tool reviews now mention things like model disclosures or content labeling. Part of the answer is regulatory. Under the EU AI Act, the transparency obligations in Article 50, covering AI system disclosure and the labeling of synthetic content, become applicable on August 2, 2026, according to Sidley Austin's Data Matters compliance tracker (June 24, 2026). Tools that serve EU users are adjusting how they disclose AI-generated output ahead of that date, and our reviews are starting to note which vendors are ready and which are not.

Something look off in a review? Outdated pricing, a feature that changed, a claim that does not hold up, tell us through our contact page and we will check it.