CompanionRank
Home / Blog / Methodology
Trust & Methodology

How We Review AI Companion Apps

An honest, plain-language account of what we actually test, what we don't, and how our scores get made.

Last verified: July 2026Author: CompanionRank Editorial TeamReading time: ~7 min
TL;DR: We test what's reachable on free tiers, demos and public documentation, and we clearly label anything we couldn't verify ourselves as "vendor-reported." Scores are our own editorial judgment across a fixed rubric, not an aggregation of user star ratings — we don't publish an AggregateRating schema because we don't collect user review data. We re-check pricing and features roughly every quarter and mark the verification date on every page.

Why we're telling you this

Most "review" sites in this category present ranked lists without explaining how the ranking was produced. We think that's a problem, especially for apps that involve personal conversation, emotional attachment and payment information. This page exists so you can judge our work on its actual merits — what we checked, what we assumed, and where the gaps are.

What we test hands-on

For every app we cover, our editorial team creates an account on the free or trial tier where one exists and spends real time in the product. That includes:

We take dated screenshots and written notes during these sessions, which is what our scoring is based on.

What we mark as vendor-reported and unverified

Some claims sit behind a paywall we haven't purchased, behind enterprise-only features, or are simply statements the company makes about its own backend (for example, specifics of a model architecture or exact context-window size). When we can't independently confirm something, we say so directly in the text — phrases like "the app states" or "according to [app]'s help center" are a deliberate signal, not filler. We do not present a vendor's marketing claim as our own tested finding.

How our scores work

We score apps across a small number of consistent categories: character/persona depth, memory behavior, roleplay flexibility, image or media generation quality (where applicable), pricing clarity, and content/safety controls. Each category score is our team's editorial judgment based on the hands-on session described above, applied the same way across every app we cover.

This is important: our scores are not a statistical aggregation of user reviews, and we do not publish review-count or star-rating structured data for these scores, because we have not collected a real, verifiable set of user reviews ourselves. A number that looks like "4.8 out of 5 based on 12,000 reviews" should come from an actual review collection system — ours doesn't work that way, so we don't claim it does.

Verification cadence

Pricing changes, features get renamed, and policies get rewritten. We aim to revisit each page on a roughly quarterly cycle to re-check prices, feature claims and policy links, and sooner if we're notified of a major product change. The "Last verified" date at the top of every guide and review reflects the most recent check — if it's been longer than two quarters, treat pricing specifics with extra caution and check the app's own site directly.

Our stance on privacy and age-appropriateness

AI companion apps sit in a category that can involve sensitive personal disclosure and, on many platforms, adult-oriented content. We think users deserve straight information here, not hype:

Corrections

If you spot something on this site that's outdated or wrong, we want to know. We do not accept payment in exchange for a specific score or ranking position; where we have an affiliate relationship with an app, it's disclosed in the footer of every page and never changes the scoring rubric applied to that app.

Frequently asked questions

Do you actually pay for and use these apps?

For core features we test on free tiers and publicly available demos where one exists. Where a feature is only available on a paid tier we couldn't access, we label the claim as vendor-reported and unverified rather than presenting it as something we confirmed ourselves.

What does "vendor-reported" mean on your site?

It means the claim comes from the app's own marketing, help docs, or app-store listing, and we have not independently reproduced it. We flag these clearly instead of blending them in with hands-on findings.

How do you calculate your scores?

Scores are our editorial judgment across a fixed set of categories, applied the same way to every app. They are opinions based on our testing notes, not an aggregation of user ratings or a certified benchmark.

How often do you re-check your reviews?

We aim to re-verify pricing, feature lists and policy pages roughly every quarter, and sooner if we notice a major app update. The visible "Last verified" date reflects the most recent check.