Subscore

User ControlTesting methodology

User Control measures how much control you have over the personal information stored by an AI girlfriend app.

We check whether you can delete conversations, close your account, request deletion of personal data stored outside the visible account, opt out of AI training, and download a copy of your data.

These controls matter because deleting something from the screen does not always mean the company has deleted it from its systems.

The current live database contains five User Control evidence tests. Consent Controls appears in an older methodology description but is not an active evidence definition in the live scoring system.

5 evidence groups
5 scored tests

User Control is organized into 5 evidence groups. Each group contains one scored test: Delete Chats, Delete Account, Delete Personal Data, Training Opt-Out, and Export Data.

How the User Control score is calculated

Every test receives a score from 0 to 10.

The live database stores the following weights: 25 for Delete Chats, 25 for Delete Account, 12 for Delete Personal Data, 12 for Training Opt-Out, and 12 for Export Data.

These weights add up to 86 rather than 100. The scoring engine automatically rescales them when calculating the final score.

After rescaling, the public percentages are 29.07% for Delete Chats, 29.07% for Delete Account, and 13.95% for each of the remaining three tests.

Delete Chats and Delete Account have the largest effect on the score. User Control makes up 28% of the Privacy score.

29.07%Delete Chats
29.07%Delete Account
13.95%Delete Personal Data
13.95%Training Opt-Out
13.95%Export Data

Weighted evidence groups (combined 100%)User Control score28% of Privacy score

View exact calculation

The stored weights are 25, 25, 12, 12, and 12, totaling 86. Percentages are rounded for display. The calculation engine uses the original stored weights and rescales them proportionally.

User Control makes up 28% of the Privacy score. Privacy makes up 10% of the overall performance score.

Scored tests and weights

Scored test	How much it counts	See scoring
Delete Chats	29.07%	View
Delete Account	29.07%	View
Delete Personal Data	13.95%	View
Training Opt-Out	13.95%	View
Export Data	13.95%	View
Total	100%

How the score is calculated

Each test gets a score from 0–10
Test score × how much it counts
Example: 10 × 29.07% = 2.91 points
We do this for every test
We add all the points together
Final User Control score
Counts for 28% of Privacy

Example calculation

We multiply each test score by how much it counts. We then add all the points together.

Scored test	Test score	How much it counts	Calculation	Points added
Delete Chats	10.00	29.07%	10.00 × 29.07%	2.91
Delete Account	5.00	29.07%	5.00 × 29.07%	1.45
Delete Personal Data	10.00	13.95%	10.00 × 13.95%	1.40
Training Opt-Out	5.00	13.95%	5.00 × 13.95%	0.70
Export Data	10.00	13.95%	10.00 × 13.95%	1.40
Final User Control score	—	100%	Add all points	7.85/10

Standard scoring

Yes

The control is clearly available and works without an important restriction.

Limited

The control exists, but an important restriction makes it less useful. Examples include only some data can be deleted, support must be contacted, the export is incomplete, or the training opt-out only covers some content.

No

The control is clearly unavailable.

Unknown

We could not confirm whether the control exists. Privacy Unknown results are excluded from the calculation rather than automatically receiving 0. Unknown does not mean the app is safe or that the control is available.

Not Applicable

A test can be marked Not Applicable when it genuinely does not apply. For example, Training Opt-Out may be Not Applicable when the company clearly states that private chats and photos are never used for AI training. When a test is Not Applicable, we remove it and spread its weight across the remaining tests.

Manual adjustment

In rare cases, we may adjust a score when the calculated result is clearly misleading. We always record the reason.

Why User Control matters

AI girlfriend apps can store highly personal information.

This may include private conversations, uploaded photos, generated images and videos, character preferences, relationship details, payment and account information, and information saved by the AI as memories.

A delete button does not always remove all of this information.

Deleting a conversation may only remove it from your chat list. Closing an account may leave backups, payment records, or other stored information behind.

That is why we test several different controls.

Delete Chats checks whether individual conversations can be removed. Delete Account checks whether users can close the full account without fighting through support.

Delete Personal Data goes further by checking whether users can request deletion of information that may remain outside the visible account.

Training Opt-Out checks whether users can stop their content from being used to improve AI models. Export Data checks whether users can obtain a copy of the information connected to their account.

A strong platform should make these controls easy to find and easy to understand.

What good User Control looks like

A high User Control score means users can remove their conversations, close their accounts, request full data deletion, control AI training, and download their information. A lower score means important controls are missing, hidden, or only available through a slow support process.

Main benchmarks

GoodLimitedWeak

Chat deletion

YesLimitedNo

Account deletion

YesLimitedNo

Data rights

YesLimitedNo

Minimum useful standard

Delete Chats: Yes
Delete Account: Yes
Delete Personal Data: Yes
Training Opt-Out: Yes
Export Data: Yes

Common red flags

Deleting a chat only hides it temporarily
Account deletion requires several emails to support
The company does not explain how to delete stored personal data
Training is enabled automatically with no opt-out
The export contains only basic profile information
A data export request receives no result within 30 days
Important privacy controls are spread across several confusing pages

How we test User Control

We use a paid test account and complete the User Data Controls test session.

We create three chats and try to delete each one.

We then check the account settings for a direct account-deletion option and count the number of steps required to request deletion.

We review the account settings, privacy policy, terms, and help pages to determine whether users can request deletion of personal information stored outside the visible account.

We also check those sources for an AI-training opt-out.

Finally, we request an export of the test account’s data and record whether it arrives within 30 days.

For every test, we save the result, relevant notes, and supporting evidence.

Limitations and edge cases

We cannot directly inspect the company’s internal databases.

When a chat disappears, we can confirm that it is no longer visible in the account. We cannot independently prove that every backup copy was immediately erased.

The same limitation applies to account deletion and personal-data requests. We can test the available process and review what the company promises, but we cannot see every internal system.

Some companies are legally required to keep certain information after account deletion. Payment records, fraud-prevention records, and legal records may be stored longer than normal account content.

That does not automatically mean deletion controls are bad. The company should clearly explain what remains, why it remains, and for how long.

Export requests can also take time. The live methodology gives the company up to 30 days to deliver the export.

Our results reflect the controls, policy, and account settings available on the recorded test date.

Evidence groups

User Control has 5 evidence groups made up of 5 scored tests.

Delete Chats measures whether you can remove individual conversations from your account.

This lets you remove a private or unwanted conversation without deleting your full account.

Delete Chats

29.07%of Delete Chats (29.07% total)

Whether you can remove individual conversations from your account.

How we test

We create three separate chats. We try to delete each conversation through the normal controls available to users. We record how many chats were deleted, whether restrictions applied, and whether the chats remained removed after reopening the account.

What counts

Deleting one full conversation
The conversation disappearing from the chat list
The deletion remaining after refreshing or reopening the app
A normal deletion control available to users

What does not count

Deleting individual messages while the conversation remains
Archiving or hiding the chat
Muting notifications
Removing the character while keeping the chat history
A deleted chat returning after the page is refreshed

Result shown

Yes — all 3 chats were deleted successfully

Scoring

We use the result categories below.

Result	Score
Yes — all three chats could be deleted	10/10
Limited — only some worked or important restrictions applied	5/10
No — chats could not be deleted	0/10
Unknown — result could not be confirmed	Excluded

In this example, Delete Chats scores 10/10.

Evidence group 2 of 5

Delete Account measures whether users can request account deletion directly through the account settings.

A user should not need to search through legal pages or argue with customer support just to leave the platform.

Delete Account

29.07%of Delete Account (29.07% total)

Whether users can request account deletion directly through the account settings.

How we test

We open the test account’s settings and look for an account-deletion option. We count every required step from opening the settings to reaching the deletion request or final confirmation. We also record whether the option is inside the account, support must be contacted, extra identity checks are required, and the company explains what happens to stored data.

What counts

A clearly labelled Delete Account option
An in-product deletion request
A normal process available to paying users
Clear confirmation of what the request will remove

What does not count

Logging out
Cancelling the subscription without deleting the account
Removing the app from the device
Deactivating notifications
A policy saying deletion is possible without explaining how

Result shown

Limited — deletion was available, but support had to be contacted

Steps required: 6

Scoring

We use the result categories below.

Result	Score
Yes — deletion can be requested directly in account settings	10/10
Limited — deletion exists but requires extra help or important restrictions apply	5/10
No — no usable account-deletion process was found	0/10
Unknown — the process could not be confirmed	Excluded

In this example, Delete Account scores 5/10.

Evidence group 3 of 5

Delete Personal Data measures whether users can request deletion of personal information that may remain outside the visible account.

This is broader than deleting chats or closing the account.

Delete Personal Data

13.95%of Delete Personal Data (13.95% total)

Whether users can request deletion of personal information that may remain outside the visible account.

How we test

We review the privacy policy, account settings, help pages, and data-request instructions. We check whether users can request deletion of stored personal information beyond the content they can remove themselves.

What counts

A personal-data deletion request
A privacy request form
A clear email or contact method for deletion requests
Clear instructions explaining which data can be removed
A process available to normal users

What does not count

Only deleting visible chats
Only deleting the account profile
Cancelling the subscription
Asking the AI character to forget something
A vague statement saying users have rights without explaining how to use them

Result shown

Yes — users can submit a request to delete personal data stored outside the visible account

Scoring

We use the result categories below.

Result	Score
Yes — a clear personal-data deletion process exists	10/10
Limited — only some data can be deleted or important restrictions apply	5/10
No — no personal-data deletion process is available	0/10
Unknown — the company does not clearly explain the process	Excluded

In this example, Delete Personal Data scores 10/10.

Evidence group 4 of 5

Training Opt-Out measures whether users can stop their chats, photos, or other content from being used to train or improve AI systems.

This test focuses on user choice. Whether the company uses private content for training in the first place is scored separately under Data Use.

Training Opt-Out

13.95%of Training Opt-Out (13.95% total)

Whether users can stop their chats, photos, or other content from being used to train or improve AI systems.

How we test

We check the account settings, privacy settings, privacy policy, help pages, and training or model-improvement controls. We look for a clear option or request process that lets users opt out.

What counts

A training toggle inside the account
A clear opt-out request form
A documented email process
An opt-out covering chats, images, or other private content
Clear information about what the opt-out changes

What does not count

Turning off marketing emails
Disabling personalized advertising
Deleting one chat
A vague privacy setting with no explanation
A policy saying data may be used without providing a choice

Result shown

Limited — users can opt out of future training, but the control does not cover data already collected

Scoring

We use the result categories below.

Result	Score
Yes — a clear and useful training opt-out exists	10/10
Limited — the opt-out covers only some data or important restrictions apply	5/10
No — training may occur and no opt-out exists	0/10
Unknown — training choice is not clearly explained	Excluded
Not Applicable — private content is clearly not used for training	Removed and weights rescaled

In this example, Training Opt-Out scores 5/10.

Evidence group 5 of 5

Export Data measures whether you can request and receive a copy of the information connected to your account.

This can help users understand what the platform stores and keep a copy before deleting their account.

Export Data

13.95%of Export Data (13.95% total)

Whether you can request and receive a copy of the information connected to your account.

How we test

We request an export of the paid test account’s data. We record how the export is requested, whether support is required, when the export arrives, what file formats are included, and whether important account information is present. The live methodology allows up to 30 days for the export to arrive.

What counts

A self-service data export
A formal data-access request
A downloadable account-data file
An export delivered within the test period
A file that opens and contains account information

What does not count

Copying chats manually
Taking screenshots
Exporting only one conversation when a full account export was requested
Receiving an empty or broken file
A policy saying users can request data without providing a working process

Result shown

Yes — the account export arrived after 8 days

Format: JSON and HTML

Scoring

We use the result categories below.

Result	Score
Yes — a useful account-data export was received	10/10
Limited — the export was incomplete or important restrictions applied	5/10
No — no usable export was available	0/10
Unknown — the export process could not be confirmed	Excluded

In this example, Export Data scores 10/10.

Back to Privacy