Explaining your routine, goals and usual foods every time can make a conversation tedious. Memory lets you save useful context and return to it in later chats.

What may be useful

Your goals, preferences and the routine you describe can help the assistant. For food logging, you can also save foods and meal sets with known nutrition information, making them easier to log again.

For example: “Save this food. I eat it often.” When you log that familiar food again, saved information helps avoid starting with a new estimate. Specify the name and portion: similar foods can have different ingredients.

After saving, check that the entry appears in memory or your saved foods list.

Memory under your control

In the app, open Profile, then AI and “What AI knows about me.” You can review the saved context, correct entries and delete information you no longer need.

Has your goal or routine changed? Update the entry so the assistant does not rely on outdated information. An accidental assumption should not become a permanent fact about you.

Settings include a separate permission for saving health information. Such information requires your consent. The general memory switch and the health-information permission serve different purposes. Check both before sharing context you want to use in future conversations.

Remove a fact or turn memory off

If an individual entry is no longer useful, delete it in memory settings. If you do not want context to be saved, turn off the corresponding setting.

Assistant memory, your metric history and your conversations are different parts of the app. Deleting one saved fact does not delete all your account data.

What memory cannot replace

Memory cannot create missing bracelet measurements. It does not know about a new routine you have not described, and it cannot guarantee accurate portions from a photo. If important context has changed, say so directly.

It helps to review saved entries occasionally: which facts are still correct, which are outdated, and which should be removed. That leaves the assistant with the context you actually want it to consider in future conversations.