How we rate companion apps

AArtur PuleevUpdated 2026-10-29How we test

Who maintains this edition

The Companion Guide NL editorial team writes and updates this edition for readers in the Netherlands. We compare AI companion and adult chat apps based on what each provider publishes: pricing pages, terms of service, privacy policies, help-centre articles and app-store listings. Every figure on a review or comparison page shows the date we checked it, and every page shows when it was last updated. This page explains how we calculate ratings and how the site makes money.

Published facts, no hands-on tests yet

We calculate scores from each app's published terms, pricing, policies and store listings on the dates shown. We have not yet used paid test accounts, so no score here reflects our own experience of memory, conversation quality or generated images.

That is why every total is capped at 4.5 out of 5. The whole site uses one /5 scale, and the score an app gets in its review is the same score it carries in every hub, card and comparison table.

Eight criteria and their weights

• Pricing transparency: 15%
• Free tier: 15%
• Privacy: 15%
• Reputation: 15%
• Cancellation and refunds: 10%
• Content rules: 10%
• Age check: 10%
• Platforms and language: 10%

Each criterion gets a score from 1 to 5. If a provider publishes nothing on a criterion, we leave it out rather than guess. We then rescale the remaining weights, so the total is a weighted mean of the criteria we could score. An app needs at least three scored criteria to get a total.

Pricing, free use and cancelling

Pricing transparency starts at 2. An app gains a point for published plans in a confirmed currency, another for naming the descriptor that appears on a bank or card statement, and one more for documenting what each token or credit action costs (or for not using tokens at all). It loses a point if the introductory price differs from the renewal price, and another if it publishes no plans.

Free tiers are scored on what they allow:

• 5: unlimited free text or at least 50 messages a day
• 4: at least 20 messages a day
• 3: a smaller daily allowance, or a free trial of three days or longer
• 2: a one-off total, or a free plan that requires a card
• 1: no free plan

Nomi.ai, for example, lists 50 free messages a day (checked 2026-10-28). Candy.ai's 5 free messages are a total, and no daily renewal is stated (checked 2026-10-28).

Cancellation and refunds starts at 2. An app gains a point each for documented cancellation steps, a statement of what access remains after cancelling, and a published refund policy. It loses a point for auto-renewal without a documented reminder, or for a known renewal trap.

Content rules, privacy and age checks

Content rules starts at 2. An app gains a point for a public rules page and another if those rules explicitly ban minors and real-person likenesses. It gains a third point if adult content sits behind an opt-in toggle or a stated 18+ gate.

Privacy starts at 2 and stays between 1 and 5:

• +1 for a linked privacy policy
• +0.5 if the policy is dated within the last 18 months
• +1 for a documented way to delete your account
• +0.5 for a stated retention period
• +0.5 for an operator based in the EU/EEA, UK, Switzerland or Canada
• +1 for confirmed end-to-end encryption
• −1 if the policy allows chats to be used for model training

Age checks are scored by method. Document, biometric or third-party verification scores 5, as with Character.AI's use of Persona (checked 2026-10-28). Email or account registration plus a stated 18+ policy and enforcement notes scores 3. Self-declaration alone scores 2, and no check scores 1.

Platforms, Dutch interface and reputation

Platforms and language starts at 1. A web app, an iOS listing and an Android listing each add a point. If there are no store apps, an installable web app adds 0.5. On this edition, a Dutch interface adds one more point. Character.AI lists Dutch among its interface languages (checked 2026-10-28). Replika's iOS listing does not include Dutch (checked 2026-10-28).

Reputation uses the best public rating we can find on Trustpilot, the App Store or Google Play: 4.3 or higher scores 5, 4.0 scores 4, 3.5 scores 3, 3.0 scores 2, and anything lower scores 1. We subtract a point for a documented incident, such as a data breach or a removed review profile, and another for regulator action in the past 24 months. Store ratings feed this criterion only. We never present them as our own score.

Coverage labels

• Full data: all eight criteria scored.
• Provisional: five to seven criteria scored, with the 4.5 cap applied.
• Limited data: three or four criteria scored. These totals are capped at 3.5 and should be read as a floor, not a verdict.
• No score: fewer than three criteria scored.

Totals change as missing facts get documented.

Dates, euros and BTW

Each fact has its own check date. When we refresh a page, the updated date changes and we replace any figure that has moved.

We show euro prices only where a provider or the Dutch App Store lists them. The Dutch App Store, for example, lists Janitor+ at €14.99 a month or €99 a year (checked 2026-10-29). We did not complete a checkout, so we show no BTW breakdown for those prices. Prices published only in US dollars are labelled USD and not converted, and we do not treat a dollar sign without a stated currency as euros. We never add or subtract 21% BTW ourselves.

iDEAL, SEPA and PayPal are common in the Netherlands, but we only say an app accepts them if the app documents it. OurDream's help pages, for instance, state that PayPal is not supported (checked 2026-10-28).

How the site is funded

Some links on this site are affiliate links. If you subscribe through one, we may earn a commission at no extra cost to you. Affiliate links never change a score. Rankings follow the criteria above, so an app with no affiliate programme can rank above one that pays.

What paid test accounts will add

Once we start using paid test accounts, we will add three experience criteria: memory in practice, conversation quality and support response. At that point we will remove the 4.5 cap on totals, and pages will show the date of each test.

Apps we never list

Some services are excluded from every hub, table and recommendation on this site. This includes any app whose features involve sexual imagery of real, identifiable people rather than fictional adult characters. Excluded apps get no score, and if we mention one, it is only to say that it is not listed.