How I test AI girlfriend platforms
Ten tests, the same for every platform, each graded from 1 to 5 on a fixed scale. The overall grade is their average plus my verdict, which can move it by up to one point. Every test is run by hand on an account I use, and the measurements, screenshots and recordings behind each grade go in the full test results published with the review.
Tested and written by Miles Foster, Owner and tester, RankDude.
Each platform gets the same session: I sign in, pick three characters, run the chat tests, generate images and a clip from the same prompts, make a recorded call, build a character from the same brief, count the library, and read every plan and credit price at the checkout. Nothing is graded from a feature list or a press kit.
Each review states how the account was obtained: bought at the public price, used on a free tier, or provided by the vendor. A vendor-provided account never changes a grade, and a vendor never sees a review before it is published.
How I rate
Every platform gets one score out of 5.
First, I run the same 10 tests on every platform and give each one a grade from 1 to 5. The average of the graded tests is the starting point; a test that cannot be graded is left out of it, and the review says how many tests the score is based on.
Then I add my verdict: what the tests cannot measure, such as how the platform holds up after the test weeks, how billing and cancelling work in practice, and how it compares with the others in the category. My verdict can move the score up or down by one point at most.
Every review shows how much my verdict changed the score, and why, in its scores table.
What a grade means
| 5 | Excellent |
| 4 | Good |
| 3 | Average |
| 2 | Below average |
| 1 | Poor |
The final score is the average of the graded tests (ten when every test can be graded), to one decimal place, plus my verdict, which can move it by up to one point. If a test couldn't be finished properly it gets no grade, and the review says how many tests the score is based on. Some tests on a free plan use fewer messages, images or calls than usual; I test them again on a paid plan when I can.
The 10 tests
1.Pricing & Value
Every plan and credit pack is read from the checkout, not the marketing page. The cost of one picture, one clip and one minute of call is worked out in dollars. Payment methods are listed as seen at checkout.
- 5: fair unlimited chat with clear, low per-item prices
- 4: good value, one weak spot
- 3: average value
- 2: expensive or unclear
- 1: poor value or hidden costs
2.Character Customization
A character is created from scratch with a fixed brief: origin, age, hair, eyes, body, personality, voice. Each option is counted and the result is checked against the brief.
- 5: every trait adjustable and respected
- 4: most traits adjustable and respected
- 3: presets only, respected
- 2: few options or brief ignored
- 1: no creation
3.Chat Response Speed
At least 20 messages across three characters: short questions, long messages and follow-ups. Each reply is timed from the moment I press Send until the first visible reply text appears, and again until the reply finishes; typing indicators do not count. The median is the grade; the average, fastest, slowest, 95th percentile and any failed messages are reported.
- 5: median under 1 second
- 4: 1.00 to 1.49 seconds
- 3: 1.50 to 2.49 seconds
- 2: 2.50 to 3.99 seconds
- 1: 4 seconds or longer
4.Chat Memory
I give each character my name and three personal facts in the first messages, then ask about them at 15 and 30 messages, on at least three characters. Recall is scored on what the character gets right without a hint.
- 5: everything recalled at 30 messages
- 4: one slip at 30
- 3: recalled at 15, lost by 30
- 2: partial recall at 15
- 1: nothing recalled
5.Character Consistency
A roleplay of at least 20 messages per character with a defined personality and backstory, checked against the profile page. Slips in tone, facts or persona are counted.
- 5: no slips
- 4: one or two small slips
- 3: several slips, persona intact
- 2: persona drifts
- 1: persona lost
6.Image Generation
The same prompts for every platform, safe-for-work and adult, on the free tier and on a paid plan where the account allows. Each result is checked for whether it was generated for the request and whether it matches the character and the prompt. Stored stock files presented as generations are reported as such.
- 5: live generation that honours the prompt every time
- 4: live, minor misses
- 3: live, frequent misses
- 2: stored files or wrong character
- 1: no usable output
7.Video Generation
The same prompts for every platform. Each clip is checked frame by frame for motion, for likeness to the character, and for whether it was generated or served from storage.
- 5: live, fluent video of the right character
- 4: live, minor artefacts
- 3: live, visible artefacts
- 2: stored file or wrong character
- 1: no usable output
8.Voice Call Latency
A recorded call with at least four spoken questions. The gap is measured from the end of each question to the start of the spoken answer; the median is the grade.
- 5: median under 1 second
- 4: 1.00 to 1.49 seconds
- 3: 1.50 to 2.49 seconds
- 2: 2.50 to 3.99 seconds
- 1: 4 seconds or longer
9.Voice Realism
Recorded answers from at least three characters, scored on naturalness, clarity, pacing and intonation, on the same criteria for every platform. Robotic artefacts and unclear words count against the grade. Accent is not scored.
- 5: indistinguishable from a person
- 4: natural with occasional artefacts
- 3: clearly synthetic but easy to follow
- 2: robotic or mispronounced
- 1: unusable
10.Character Variety
Characters are counted, platform-made and user-made separately, and sampled for how different they are in look, personality and style.
- 5: hundreds of distinct characters across styles
- 4: a large, varied library
- 3: a medium library, some repetition
- 2: a small library
- 1: a handful
Documented but not graded
Voice messages in chat, virtual gifts with character reactions, and a social-style feed are checked and described in every review. When they work well they count in the written assessment, but they never add points to the overall grade.
Rules that apply to every test
- Real use of the product on a real account, never promotional descriptions.
- Identical tests, thresholds and sample sizes for every platform in the category.
- Authorised accounts only; no password sharing, no bypassing of age or bot checks.
- Every grade has evidence: a measurement, a screenshot, a recording or a table, in the full test results.
- A grade changes only when a retest produces new evidence; the review records the date and the reason.
Questions about a specific grade go to the contact page. Makers can submit an AI girlfriend platform for testing. The tested ranking is on the AI girlfriend platforms page.