How we test AI companion apps

Most reviews in this category are written from the landing page. Ours start with the documents that decide what you'll actually pay.

1. The paper trail

For every app we read and date the terms of service, privacy notice, refund policy, subscription page and help-center billing articles. We record the operator's legal name and registration, the governing law, and the billing name that appears on card statements.

2. The real yearly cost

We turn every billing term (weekly, two-weekly, monthly, quarterly, yearly) into a 12-month total, and note what the subscription doesn't include: tokens, credits, coins and add-ons.

3. Hands-on testing

A reviewer pays for the app with their own card, uses it for at least a week, records what features cost in credits, then cancels and requests a refund within the published window. We keep dated screenshots of checkout, balances, cancellation and support replies.

4. Public complaints

We collect complaints from PissedConsumer, the BBB, Trustpilot, app-store reviews, Reddit and help-center comment threads. We link every one, report sample sizes, and never call an app a "scam" unless a regulator or court has.

5. Scores

Scores run from 0 to 10 and combine value, pricing clarity, refunds and cancellation, privacy terms, user-reported experience and, once tested, hands-on quality. Commissions never affect scores, and apps that pay us nothing are ranked by the same method. Scores marked "provisional" don't yet include hands-on testing.

6. Updates and corrections

We re-check prices at least monthly and show a "last updated" date on every page. Corrections go to [email protected].