Skip to main content

Anthropic's New 'Reflect' Feature Gives Claude a Mirror for Your AI Habits

Anthropic just dropped a new feature that might make you think twice about how you talk to your AI. Called Reflect, it's essentially a mirror for your Claude interactions—a way to look back at your usage habits and figure out where you're nailing it and where you could use a little polish.

Starting now, free, Pro, and Max plan users who have the memory function turned on can access Reflect through the settings page on web and desktop apps. Pick a time frame—1 month, 3 months, 6 months, or 12 months—and the system will whip up a personalized summary of your chat activity. Think key topics you've discussed, patterns in how often you use certain features, common task types, and even your most productive hours and what you were working on during those times. Future updates promise a visual timeline of usage duration, the ability to set quiet periods, and reminders about how long you've been at it.

Image

Four Dimensions of 'Checkup': From Using AI to Mastering It

Reflect doesn't just hand you a bunch of stats. It evaluates your AI collaboration habits across four dimensions: task delegation, goal description, result identification, and responsible caution. Each dimension comes with specific examples from your own chats and targeted suggestions for improvement. In other words, it's like getting an 'AI usage health check report' that tells you which tasks you assigned correctly, which instructions were a bit fuzzy, and which results you should double-check before trusting.

Anthropic's message is clear: improving AI isn't just about the model itself. Your own 'questioning literacy'—how well you ask questions and evaluate answers—plays a huge role in the final outcome. While most of the industry is focused on what AI can do, Anthropic is taking a rare inward look at how well users are doing. It's a refreshingly human-centric approach in a field that often forgets the person behind the keyboard.

Clear Privacy Boundaries: Health Data Stays Out

On the privacy front, Anthropic has drawn some firm lines. Reflect won't peek at incognito chat content or extract underlying files from connected tools. For example, if you ask Claude to summarize your inbox, the report might note that you frequently summarize emails, but it won't include the actual email content. And any conversations involving health integration tools are completely excluded from the analysis. So your sensitive data stays yours.

Key Points

  • Reflect is now in testing for free, Pro, and Max users with memory enabled.
  • Reports cover 1, 3, 6, or 12 months of chat activity, highlighting topics, patterns, and peak usage times.
  • Four evaluation dimensions: task delegation, goal description, result identification, and responsible caution.
  • Privacy safeguards: incognito chats and health data are off-limits.
  • Future updates will include a visual timeline, silent periods, and usage reminders.