Perplexity adds memory to its AI assistants
Perplexity is adding memory so its assistants can remember preferences, interests, and conversations between sessions. The feature works with different models, includes incognito mode, and lets you disable memory or limit data use.
Perplexity has added memory to its AI assistants so they can remember your preferences, interests, and previous conversations. The idea is simple: you should not have to explain who you are, what you need, or how you prefer to work every time you open a chat.
The feature is part of the new personalization tools the company announced for Perplexity and Comet Assistant. The system analyzes the context of your conversations and saves useful details, such as your favorite brands, dietary needs, or topics you ask about frequently.
What changes when AI has memory
Until now, many assistants could search your history, but they treated it as another dataset to help predict a response. Perplexity says its approach retrieves relevant information from stored memory and uses it directly to prepare each response.
In practice, you could ask for things like:
- Sneaker recommendations that take an injury or your training for a race into account.
- A book for your flight to New York based on your tastes.
- Gift ideas for different members of your family.
- A reminder of advice you received the previous week about a problem at work.
The goal is to keep you from having to repeat the context. If you have told Perplexity that you are training for a marathon, for example, that information can influence future sports equipment recommendations.
Memory you can control
Memory is not enabled as a black box with no options. Perplexity lets you turn it off whenever you want. In addition, both memory and search history are automatically disabled in incognito mode, so those sessions are not saved.
The company also says that the data is encrypted and offers an option to prevent your data from contributing to model improvement through its AI data retention settings.
This matters because memory can make an assistant more useful, but it can also store personal information you do not want to keep. Being able to review the options and turn the feature off will be just as important as having it available.
Context works with any model
One of Perplexity's bets is that memory will not be tied to a single model. You can switch between fast models, reasoning systems, or specialized tools while keeping the same personalized context.
That means a conversation does not lose all its value when you choose another model. Your previous work and preferences remain available for the next task, based on the information Perplexity considers relevant.
For you, the change could mean fewer repeated explanations and responses that better fit your situation. But it is also worth keeping an eye on what the system remembers, how long it keeps it, and which tasks use that data. Memory makes the assistant more convenient, but its usefulness will depend on whether you remain in control of the information you share.