How we check this · last full check 13 October 2026

How we check this

Two parts. A score for how good each app is to use, and five checks for whether it'll let you down later. Everything else on the site hangs off those two.

The score

Every app we've reviewed gets six scores out of 5: conversation, customizing her, how far it goes, voice and pictures, privacy, and value for money. They add up to one number, weighted: conversation 25%, customizing 20%, how far it goes, voice and pictures, and privacy 15% each, value 10%. Conversation counts most because it's what you'll spend most of your time doing. Apps we haven't reviewed yet say so instead of getting a made-up number.

The five checks

• Says who runs it. A company is named in the privacy policy or terms.
• Honest price. No cheap first month that renews higher, no weekly billing.
• Delete button. You can delete your account yourself, without emailing anyone.
• Keeps your chats out of training. The policy says your chats don't train their AI, or only if you opt in.
• Still the app it was. Not shut down, offline or turned into something different.

Each one is ✓, ✗ or ?, and ? means the policy doesn't say. We never fill in a guess.

What we read

• The privacy policy and terms, found through the links on the app's homepage.
• The public price page. If prices only show up after sign-up, we use the US App Store listing instead.
• For apps that shut down or moved: whether the domain still resolves, the app's own pages, app store listings, and news coverage.
• For the laws: the law itself, or write-ups from law firms and news outlets we'd trust.

What we write down

For each of the 46 apps: the company and its country, whether your chats train their AI, whether data goes to ad partners, how you delete your stuff, the minimum age and how they check it, refunds, auto-renewal, how to cancel, whether disputes go to arbitration, and the price. Where we can, we keep a short quote (under 25 words). The link is always there.

The colours

• Green: they promise something good for you, like "we do not use your chats to train models".
• Red: they say the opposite, like "all fees are non-refundable".
• Grey: vague, conditional, or not in there. We never fill in a yes or no ourselves.

Two calls we make on purpose. "Anonymized" use of your chats still counts as training. And sharing with ad partners, targeted ads, or "selling" data under US privacy law all count as ad sharing.

What we don't do

• Sign up, pay or log in to check something. If it's behind a login, the field stays empty.
• Claim to know what happens on their servers. A policy tells you what they promise, not what they do.
• Publish street addresses or the names of individual people.
• Let any app or affiliate program see, change or approve what we write.

What's still missing

Some sites block unfamiliar browsers, and some policies live at addresses we couldn't find. Those apps show "not mentioned" or a note, never a guess: CrushOn.AI.

Mistakes and updates

We re-check every app regularly and every page shows the date. If something's wrong or out of date, email desk@aigirlfriendadvisor.org with a link and we'll fix it. Apps can write too. We update when their public documents change.

Using our data

It's free under CC BY 4.0. Credit "AI Girlfriend Advisor" with a link and you're good. Download it here.