Methodology
How we test AI girlfriend apps
Every app we rank goes through the same paid test and the same 100-point rubric. Each review shows its test record so you can see what we did.
Last updated
The short version
- We buy the plan with our own money. We never review a free trial as if it were the full product.
- Each app gets at least 7 days of use, with the same scripted tests on the same day of the test.
- We score seven areas on a fixed 100-point rubric. Affiliate partners get the same rubric as everyone else.
- We record the price we paid, the date and the device. That record sits at the top of the review.
What we buy
We subscribe to the plan most readers would choose, usually the cheapest monthly plan that removes message limits. If an app sells tokens or credits for images or voice, we buy a pack and log what each feature cost. When a plan changes price during a test, the review shows both prices and the date of the change.
The tests
Memory test
On day 1 we give the companion 10 simple facts, such as a pet's name and a job. We ask about each fact after 24 hours and again after 7 days. The score is the number of facts recalled without a hint.
Conversation test
We run three fixed scripts: everyday small talk, a long story thread, and a disagreement. We read each one for character consistency, repetition and whether the companion moves the conversation forward.
Pricing and cancellation test
We add up what a month actually costs once tokens are included. At the end of the test we cancel through the app's own flow and record how many steps it takes and whether a renewal charge follows.
Privacy test
We read the privacy policy for training use, data sharing and retention. Then we delete one chat and the whole account and check that both are gone.
The rubric
| Area | Points | What we measure |
|---|---|---|
| Conversation quality | 20 | Stays in character, follows a thread over 50+ messages, varies its replies. |
| Memory | 20 | Recalls 10 seeded facts after 24 hours and again after 7 days. |
| Pricing honesty | 15 | Real monthly cost including tokens or credits, renewal terms, how hard it is to cancel. |
| Privacy and data | 15 | What the privacy policy allows, whether chats can be deleted, account deletion that works. |
| Media features | 15 | Images, voice and video: quality, limits per plan, cost per item. |
| Controls and safety | 10 | Age checks, content settings, report and block tools. |
| Support | 5 | Response time and answer quality to one billing question and one technical question. |
| Total | 100 |
What we do not do
- We do not publish explicit images or explicit chat text. Reviews describe adult features in plain terms.
- We do not test or describe any feature that depicts a real person without consent, or anyone under 18.
- We do not invent quotes, users or results. If a test did not run, the review says so.
- We do not change a score because an app pays us more.
Updates
AI companion apps change prices and features often. We re-check prices and plan limits every 90 days and re-run the full test once a year, or sooner after a major change. Every review shows its last test date.
Changelog
- : rubric published.