Skip to content

How We Test AI Girlfriend Apps

A good AI girlfriend review isn't a list of features. Any landing page can give you that, and most of them leave out the parts that matter: the message cap, the credit bill, the refund window, the day the character forgets who you are.

So we test AI girlfriend apps the way you'd use them. Same routine for every app, on a real account, over several days. This page walks through it step by step, so you can judge our reviews by the method behind them.

What Happens Before We Score an AI Girlfriend App?

Every app goes through six stages before it gets a review or a place on one of our lists:

  1. A free account, with no card attached
  2. The paid plan most people would actually buy
  3. Several days of chat, with memory checks along the way
  4. Images, voice and calls, priced per use
  5. The terms, privacy policy and refund rules
  6. Cancelling, and what happens afterwards

Each stage feeds a different part of the review. Here's what we do at each one and what we're looking for.

Step 1: How Far Does the Free Tier Really Go?

We sign up as a brand-new user and use the free plan until it stops us. The goal is to find the wall, not to admire the onboarding.

At this stage we note:

A free tier that ends after five messages is a trial, not a free plan, and the review calls it that.

Step 2: Paying for the Plan People Actually Buy

Next we pay. We pick the plan most readers would choose, usually the one the app highlights, and note what the checkout shows versus what the pricing page promised.

Watch out: this is where pre-ticked annual billing, auto-renewal and currency surprises tend to show up, so we screenshot the checkout before confirming. [VERIFY: confirm this is part of your process]

We also record the billing descriptor, the name that appears on a bank statement. For an adult app, plenty of readers want to know that before they pay.

Step 3: Does the AI Girlfriend Remember You After a Few Days?

Chat quality is where AI girlfriend apps differ most, and it can't be judged in one sitting. We spread conversations across several days, because memory problems rarely show up in the first hour.

The Memory Checks We Run

During those chats we plant small details and come back for them later:

Then we ask about them a day or two later, without reminding the character first.

What Counts as Breaking Character

We also note when a companion slips out of the persona it was given:

One slip is normal. A pattern is a con, and it goes in the review with the moment it happened.

Step 4: What Do Images, Voice and Calls Cost Per Use?

On most apps we checked, the subscription is only part of the bill. Images, voice notes, calls and video usually come out of a separate balance of tokens, coins or credits.

So we use each paid feature until we know what one of each costs:

FeatureWhat we measureWhat goes in the review
ImagesCredits per picture, consistencyCost per image, face stability
Voice notesCredits per message, sound qualityWhether it's included or metered
CallsCost per minute, delayReal price of a short call
VideoCredits per clip, lengthWhether it's worth the spend
Character builderFree edits, paid rerollsWhat a finished character costs

Those numbers become the real-cost section of each review, priced for light, regular and heavy use.

Price: a plan's headline number is where we start, never where we stop.

Step 5: Reading the Fine Print So You Don't Have To

Before writing a word of the verdict, we read the documents most users skip. The rules in them often decide whether an app is worth trusting with a card and a few personal conversations.

We check:

  1. The terms of service: content rules, account bans, how disputes work
  2. The privacy policy: what's stored, for how long, and whether chats train the model
  3. The refund policy: the window, the conditions, what's excluded
  4. Age verification: whether there's an actual check or just a checkbox
  5. Payment methods: cards, crypto, app stores, and who processes them

When a policy is vague or silent on something, the review says so. We report what a document says; we don't fill the gaps with guesses.

Step 6: How Hard Is It to Cancel?

The last test is leaving. We cancel the plan we bought and note every step it takes.

Here's what we look for:

An app that makes cancelling hard loses points, however good the chat was.

How Does the Test Turn Into a Score?

Every AI girlfriend app we rate gets one score out of 10, with one decimal. The number sums up the six stages above, weighted toward what matters for that page.

The score is built from these criteria:

CriterionWhich stage feeds it
Chat quality and memoryStep 3
Real monthly costSteps 2 and 4
Free tierStep 1
Images, voice and videoStep 4
CustomizationSteps 3 and 4
Privacy, billing and cancellingSteps 5 and 6

A list built around voice leans harder on calls; a budget list leans on cost. The written review always explains the trade-offs.

Our scores run from 8.0 to 9.9 and compare apps within what we cover. No two apps on the same page share a score, so a higher number always means something specific. The cons stay in the text whatever the score is.

What We Leave Out of Our Testing

Some things never make it into a review, however loudly an app advertises them:

When Do We Retest an AI Girlfriend App?

AI companion apps change fast, so a review is a snapshot, not a final word. We go back to an app when something shifts that a reader would care about:

When a retest changes the verdict, the score changes with it.

Our site uses cookies and similar tracking technologies to personalize our content and analyze our traffic.