Review methodology

This page documents the exact process AIpedia follows to evaluate every AI tool in our encyclopedia. We publish it so that readers can decide, independently, whether our conclusions apply to their situation. If you think any step should be tightened, we'd like to hear from you via the contact page.

1. Tool selection

We only add tools that meet all of the following criteria:

• The tool is commercially available (not vapourware) and has a documented home page or product page.
• The tool uses AI as a core — not incidental — feature. Plugins that simply wrap an existing chatbot are not listed as stand-alone products.
• We can reasonably verify the vendor's identity and contact details.
• The tool falls into one of our published categories (see the home page) or justifies a new category.

2. Hands-on testing

Whenever the free tier or a trial is available, an editor signs up and runs at least three representative tasks covering the tool's primary use case. For paid-only tools we rely on the vendor's documented demos, reviewer access where offered, and cross-referenced third-party reviews — and say so in the verdict.

• Response quality and accuracy on a fixed benchmark prompt/task.
• Setup time from signup to first output.
• Stability during typical workloads and around advertised limits.
• Clarity and completeness of the free/trial experience.

3. Our rating rubric (1.0 – 5.0)

The editorial rating consolidates four weighted factors:

Individual factor scores are averaged with the weights above to a single 1.0 – 5.0 rating, rounded to one decimal place. The rating is intentionally coarse: we want readers to rely on the written pros, cons, and verdict for the detail, and use the rating as a quick comparative anchor only.

4. Writing the review

Every review follows the same structure so that readers can compare tools side by side without re-learning the page layout: what it is, who it's for, pricing, features, pros and cons, how to get started, alternatives, FAQ. First drafts may be prepared with AI assistance, but every section is edited, fact-checked, and has its claims cross-referenced against primary sources by a human editor before publication.

5. Update cadence

• Every 6 months: every featured tool is reviewed for pricing, feature, and positioning changes.
• Immediately: after a major release, pricing change, ownership change, or public incident.
• On reader report: if a reader submits a correction we verify and (if accurate) update within two business days.

The Last updated date at the top of every review reflects the most recent editorial pass, not merely a re-publish.

6. Conflict-of-interest controls

• We disclose affiliate and advertising relationships in every review that contains commercial links. See the disclosure page.
• Editors do not accept payment, equity, or free access in exchange for favourable treatment.
• Advertisers have no pre-publication preview rights.
• When our founder's own products are mentioned (MixCast, AIOPulse, UGCast), the relationship is disclosed inline.

7. What we publish and what we don't

If our assessment of a tool is below 2.0 after testing, we prefer to delist it rather than publish a low-score review, unless the tool is large enough that readers will search for it anyway. In that case we keep the review visible and explain why the score is low. This prevents our index from growing with low-value listings and protects users from landing on ghost-town pages.

8. Feedback

This methodology is a work in progress. If you believe a step should be added, tightened, or removed, tell us through the contact page. Reader feedback has shaped this document more than anything else.