How to Read App Store Privacy Labels Before Downloading an AI Companion
An AI companion may feel like a private conversation, especially when it uses a character's name and remembers what you have said. But an intimate-looking interface tells you very little about which data the app collects, what travels to servers, or how long information is retained. A better first step is to inspect the app's App Store privacy section and then check what the full policy actually promises.
Apple's privacy labels are helpful starting points—not independent security certifications. Apple explains that developers provide the information about their practices, including those of third-party partners; its own consumer privacy materials describe the labels as self-reported. Apple's developer guidance and consumer privacy overview describe the system.
Find the label before installing
On an iPhone, open the app's App Store listing and scroll to the App Privacy section. You may see categories such as “Data Used to Track You,” “Data Linked to You,” “Data Not Linked to You,” or an indication that the developer reports no data collected. Exact labels and presentation can vary.
Then tap through to the detailed explanations. The headings are not enough on their own: two apps could both list “User Content” and still process it for different purposes.
For a companion app, focus on five kinds of information:
| Data type or question | Why it matters in an AI chat app |
|---|---|
| User Content | Messages, uploaded images or voice may contain intimate details |
| Identifiers | Used to associate activity with an account or device |
| Purchases | May reveal subscription status or transactions |
| Usage Data | Can describe how often and when you interact |
| Diagnostics | May help fix crashes but can still transmit technical information |
These are examples of topics to examine, not a claim that every companion app collects all five.
“Data not linked” doesn't mean “never collected”
A common mistake is treating “Data Not Linked to You” as a guarantee that no information leaves the device. It generally refers to how collected data is associated with identity under the label system. It's different from “Data Not Collected,” and neither label by itself answers every question about server processing, encryption, retention or security incidents.
Similarly, a privacy label does not automatically prove that conversations are end-to-end encrypted. That claim would need separate technical and policy evidence. Check the developer's full privacy policy before sharing sensitive information.
Five questions the label cannot answer completely
1. Does chat text go to an outside AI provider?
Many generative apps use remote inference services. If an app routes requests to another provider, the privacy policy should explain relevant third-party processing, purposes and protections. Don't assume that “private chat” means locally processed chat.
2. Can employees access conversations?
Some services may review selected content for troubleshooting, support, safety or model evaluation. Others may restrict access or use automation. The answer is app-specific. Search the policy for “review,” “human,” “service provider,” “retention” and “training,” and ask support when language is vague.
3. What happens when you delete a message?
Removing text from the interface does not always mean immediate deletion from backups, logs or the provider's systems. Look for retention windows, deletion exceptions and instructions for account erasure.
4. Are voice recordings and transcripts handled differently?
An app could store a transcript but not the raw audio, or vice versa. Voice features deserve their own explanation. Ask whether clips are stored, processed transiently or shared with speech providers.
5. Does deleting the app cancel the subscription?
No—removing an app is not the standard process for ending an App Store subscription. Apple instructs customers to open Settings → [your name] → Subscriptions and cancel the relevant subscription. Apple's current instructions.
Permissions versus privacy labels: two separate checks
An app's privacy label describes reported practices; iOS permissions control access to device features such as the microphone and photos. An app generally needs permission to use protected hardware. You can review or change it in Settings → Privacy & Security. Apple explains how.
A voice-call app may reasonably request microphone access when you start a call. A text-only chat asking for constant camera access deserves a closer look. Permission requests should match the feature you're using.
The two-minute privacy audit
Before signing up, do this:
- Open the App Privacy section and identify the reported data categories.
- Open the developer privacy policy and search for “chat,” “voice,” “photos,” “third party,” and “delete.”
- Check whether the app requires an account and what account deletion means.
- Compare requested permissions to features you've actually chosen.
- Avoid uploading addresses, IDs, passwords, private third-party photos or other sensitive material unless you understand why it's necessary.
- Look for a clear way to contact the developer about privacy questions.
A trustworthy service can still collect data; the key is that it explains the collection clearly and limits it appropriately.
A realistic example of reading between the lines
Suppose App A reports collecting user content and says its provider processes it to generate responses. Its policy names a retention period and account-deletion process. App B advertises “100% confidential” but offers no meaningful explanation of server processing or deletion.
You cannot declare A perfectly secure just from its transparency. But you do have more information with which to make an informed choice. Vague promises are not stronger protections than specific explanations.
What about age ratings?
Age ratings tell you about an app's content suitability, not whether its data is protected. A high age rating is not a privacy audit; a lower rating is not a safety guarantee. Treat age suitability, moderation and data handling as different questions.
For youth-related risks, see AI companions and teen safety. For device-level controls, see 12 iPhone privacy checks.
Checking our app
You can inspect the current privacy disclosures for AI Girlfriend: Chat & Voice on its App Store listing. This website belongs to its developer, so we encourage the same scrutiny of our app as any other provider. Don't rely on this article as a substitute for the live listing and privacy policy.
Compare two privacy policies without needing to be a lawyer
Read the parts that describe what, why, with whom, how long, and how to delete. Write each answer on one line. If a policy says “we may share data with partners” but doesn't identify purposes or categories of processors, that's a question to ask support. If it explains that audio is processed transiently to produce a reply and transcripts are retained for a defined period, you have a more specific basis for a decision.
Keep the exercise proportional. You may not need to review every paragraph before asking for a fictional café scene. But if you intend to discuss private relationships or upload identifiable photos, that extra diligence becomes worthwhile.
Finally, remember that a privacy label is a snapshot of declared practices. Revisit it when an app substantially changes features or adds voice, camera or image uploads. The data footprint of a text-only product may change when it adds media capabilities.
FAQs
Does the App Store independently verify every privacy claim? The categories are supplied by developers; they are not a universal third-party audit.
Does “Data Not Linked to You” mean zero collection? No. It describes reported linking practices, not necessarily absence of collection.
Can I remove microphone access afterward? Yes. Review app access in iOS Privacy & Security settings.
Should I put sensitive personal secrets in an AI chat? Only after carefully evaluating the service's practices; minimizing sensitive disclosures is a safer default.
