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:
- Signing up and going through onboarding, including any age-verification or content-preference prompts.
- Creating or selecting a character and running multi-turn conversations to see how the app responds over time.
- Checking whether the app remembers details from earlier in a conversation or session (see our persona memory guide for what this actually means).
- Reviewing the pricing page, in-app purchase flow and cancellation process as presented to a new user.
- Reading the app's own privacy policy and terms of service for stated data practices and content controls.
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This is important: our scores are not a statistical aggregation of user reviews, and we do not publish review-count or star-rating structured data for these scores, because we have not collected a real, verifiable set of user reviews ourselves.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:
- We note what each app's own privacy policy says about data retention, chat logging and third-party sharing — we don't audit their servers, we read and summarize their published policy.
- We note what age-verification step (if any) a new user encounters during signup, based on our own account-creation experience.
- We keep our own content at a safe-for-work level across this blog and our comparison pages, and we clearly label where a covered app includes mature/18+ content options.
- See our dedicated guide: Privacy & Age Verification, Explained.
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.
Related on CompanionRank
Best AI Companion — our current rankings Companion Comparison — feature tables What Is Persona / Long-Term Memory?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.