Home / How We Test AI Tools

How We Test AI Tools

The same 5-criteria framework, applied to every tool we review — so a score means the same thing across every guide on this site.

Adrian Cole, Lead AI Tools Reviewer at TeknexiaHub

Adrian Cole

Lead AI Tools Reviewer at TeknexiaHub

Star ratings without a rubric are just vibes. We built the Teknexia AI Tool Evaluation Framework so every review scores tools the same way — you can compare a 4-star Jasper review to a 4-star Copy.ai review and know "4 stars" means the same thing in both.

What we actually do before publishing a score

We don't score a tool from its pricing page or a 10-minute demo. Every review starts with a real task — a blog draft, a video edit, a brand-voice training cycle — run through the tool the way a working creator would use it, not the way a vendor's onboarding flow wants you to use it.

That means: real prompts on real briefs, tracking how many edits a draft needs before it's publishable, testing the free tier or trial limits honestly, and using the tool for at least several days before writing a word of the review.

Two ways we evaluate a tool

Some tools have no usable free plan or trial worth testing. Rather than fake a hands-on test or skip a tool entirely, every review uses one of two clearly labeled methods — we never blur the line between them.

// Hands-on tested

We used it ourselves

We ran the tool's free plan or trial through a real task — a draft, an edit, a training cycle — for at least several days before scoring it. All 5 framework criteria are judged from that direct use.

// Research-based

No usable free access exists

We score the same 5 criteria from the vendor's own documentation and pricing pages, official product demos, and aggregated third-party reviews (G2, Capterra, Trustpilot, relevant forums). For example, Ease of use is judged from demo walkthroughs and repeated user feedback, not direct manipulation.

Both methods use the same 5-criteria framework and the same 0-10 scale — only the source of the data changes, and it's traceable: every review states which method applies. We clearly label which method applies to each review, so you always know what you're reading.

The Teknexia AI Tool Evaluation Framework

Every score you see on this site breaks down into these 5 criteria, each out of 10.

Ease of use — /10
Time to first usable output, how much of the interface you need to learn before the tool earns its keep.
Output quality — /10
How much editing a draft needs before it's publishable, judged against our own edits, not the vendor's demo.
Workflow integration — /10
Browser extensions, API access, CMS/doc integrations, team handoff — does it fit how you actually work.
Pricing value — /10
Cost relative to output, not cost in isolation; a $59/month tool can score higher than a $19/month tool if it replaces two subscriptions.
Beginner friendliness — /10
Can someone with zero prompt-engineering experience get a usable result in the first session.

Why we publish all 5 numbers, not just an average

A single overall score hides the tradeoff that actually matters. A tool can be excellent at output quality and still be the wrong buy if pricing value or beginner friendliness drags the average down for your specific situation. Publishing all 5 lets you weight them yourself instead of trusting our weighting.

When we re-test

Last framework update: July 16, 2026

Tools change fast. We re-run the framework on any tool that ships a major pricing or feature change, and we note the last-updated date at the top of every review so you know how current the score is.

Where affiliate links fit in

We use affiliate links on some recommendations, disclosed on every page that has one. Scores are set before we decide whether a link is affiliate — see our Editorial Policy for the full breakdown of how that separation works.

Read our full affiliate disclosure →

Want to see our framework in action?

See the 5 criteria applied to a real, published review.