Claude Reflect Explained: What It Shows, Do You Get It, and How to Turn It Off

What changed (August 25, 2026): the setting that switches this off has moved for most people. Anthropic rebuilt Claude's memory today, and because Reflect is switched on and off by the memory setting, the path on this page is no longer the only one. On the new system the toggle is "Generate memory from chats" in Settings > Memory; on the older one it is still "Generate memory from chat history" in Settings > Capabilities. Anthropic's own memory help article gives the tell: "If you see Settings > Memory, you're using the new memory experience." Both paths are documented below because the migration is gradual and Anthropic still publishes both. Note that Anthropic's Reflect help article has not been updated for the change -- it was last revised 2026-07-09 and still names only the older path, so this page now follows the memory article for the setting and the Reflect article for everything else.

What changed (August 24, 2026): one reassurance on this page was too broad, and three facts were missing. Re-read against Anthropic's help article today. Availability, the memory requirement and the all-or-nothing off switch all hold unchanged. But this page said files from connected tools "are not pulled in" without the other half of Anthropic's sentence: the raw email or file is not pulled in, but anything Claude wrote in the chat about it can be -- ask Claude to summarise your inbox and that summary is in your recap. Corrected below. Also added: your Claude Mobile activity is counted even though the page itself is desktop-only, the recap is only generated when you actually open it, and the feature is still rolling out gradually.

Think of it as screen time, but for your AI chatbot. On July 9, 2026, Anthropic launched Reflect, a monthly recap that tells you what you have been using Claude for, when you reach for it most, and how you tend to work with it. It is free, it landed on the Free plan as well as Pro and Max, and you turn it on by doing nothing -- if memory is already on, it is already collecting. The catch is that last part, and the fact that the company grading your AI habit is the company selling you the AI. This page covers what the recap shows, whether you get it, and how to switch it off.

Key Takeaways

• Reflect is a monthly recap of how you use Claude, not a new model or a chatbot feature. It shows your topics, your most active day, your peak hour, and your total conversations.
• It is free and it includes the free plan. Free, Pro, and Max all get it in beta, in every region where Claude is offered. There is no extra charge.
• It is not on your phone. Reflect only appears in Claude on the web and the desktop app. It is also not available on Team or Enterprise plans.
• The catch: it only works if memory is on. Reflect appears only when Claude's memory setting is enabled, so getting the recap means letting Claude retain more about you.
• Turning it off is all-or-nothing. There is no dedicated toggle for the recap -- the only way to kill it is to switch memory off entirely, in Settings > Memory on Claude's new memory system or Settings > Capabilities on the older one.
• It skips your most sensitive chats. Incognito conversations, anything using a health integration, and activity in Claude Cowork and Claude Code are left out.
• But your phone counts, and connected apps partly count. Claude Mobile activity is included even though the recap page is not on mobile, and while raw emails and files from connected services stay out, what Claude wrote about them in the chat does not.
• It grades you. The recap frames observations around four "AI fluency" skills, which means it is also quietly telling you where to use Claude more.
• The honest verdict: a genuinely interesting mirror, and the quiet hours and break reminders are real. But it is a usage report from the company that profits from your usage. Read it that way.

What Is Claude Reflect, in Plain Terms?

Reflect is a dashboard that summarizes your own Claude history back to you. Anthropic describes the monthly recap as showing "how you've been using Claude -- the topics you spent time on, when you tend to reach for it, and where you might try something new." It is the same basic idea as your phone's screen time report or a music app's year-in-review: it does not add any new ability to Claude, it just reflects your existing behavior back at you as charts and numbers.

The everyday version: at the end of the month, Claude hands you a short report card. It tells you how many conversations you had, which day of the week you leaned on it hardest, which subjects came up most, and where you might try something you have been ignoring. Alongside that, Anthropic added optional controls -- quiet hours and break reminders -- that let you tell Claude to leave you alone at certain times.

Do You Get Claude Reflect? Availability by Plan

Yes, almost certainly -- this is one of the rare AI features that shipped to the free tier on day one. Unlike agent features that stay locked to premium plans for months (Google's Gemini Spark still needs a paid AI Pro or Ultra subscription), Reflect went straight to everyone on a personal plan. Here is who gets it:

Two limits matter more than your plan. First, platform: Reflect only shows up in Claude on the web and in the Claude desktop app. There is no recap in the mobile app, so if you only use Claude on your phone, you will not see it -- though your phone usage is still counted in the recap you would see elsewhere. Anthropic's wording: "Your activity on Claude Mobile is included in your recap even though that page isn't available on mobile yet." Second, memory: the recap "only appears when memory is turned on," per Anthropic's help center. Region is not a barrier -- it is available everywhere Claude is offered.

To find it: open Claude on the web or desktop, go to Settings > Reflect. It shows the past month by default, and a dropdown switches the view to this month so far, the past 3 months, the past 6 months, or the past year. A refresh icon next to the dropdown regenerates the recap for whichever range you are looking at.

One caveat that is easy to miss: Anthropic says the feature "is rolling out gradually. If you don't see it yet, check back later." So an eligible account with memory on can still find nothing under Settings.

What Does the Recap Actually Show You?

It is mostly numbers about your habits, plus some commentary about your technique. The recap opens with a written summary, then gives you three headline stats and a set of breakdowns:

Those four skills are Anthropic's "AI fluency" framework: Delegation (deciding whether and how to hand work to AI), Description (explaining what you actually want), Discernment (judging whether the output is any good), and Diligence (taking responsibility for what you do with it). In practice this is the part of the recap that reads like a performance review -- it tells you which of the four you are weak at.

The Catch: You Have to Turn Memory On

This is the trade nobody mentions in the headlines. Reflect needs a long view of your chat history to spot patterns, so it only appears when memory is switched on. That means the price of a report about how much you use Claude is letting Claude remember more about you across conversations. If you deliberately keep memory off for privacy reasons, you do not get a recap -- and if you want the recap, you have to accept memory.

Whether that is a fair trade depends on why you turned memory off in the first place. If you never turned it on and do not care either way, this costs you nothing. If you keep memory off on purpose, a monthly usage chart is a thin reason to reverse that decision. There is no middle setting.

How to Turn Claude Reflect Off (and Set Quiet Hours)

There is no off switch for the recap by itself, which is the most annoying thing about it. Per Anthropic's help center, the recap has no dedicated toggle. The only way to disable it is to switch Claude's memory off, which kills memory and the recap together -- you cannot keep one without the other.

Where that switch lives depends on which memory system your account is on, after Anthropic's August 25, 2026 memory rebuild. Check your settings: if you see a Memory section you are on the new system, and if memory instead sits under Capabilities you are on the older one.

The break controls are separate, and they survive independently:

Looking at the recap is free in the sense that matters on a capped plan: Anthropic says viewing it "doesn't count toward your usage limits." Only a conversation you start from one of its suggestions counts, like any other chat.

Usefully, turning the recap off does not wipe your chat history, and it does not cancel any quiet hours or break reminders you already set. Both nudges are dismissible when they appear -- they are reminders of your own stated preferences, not hard blocks. If you like the idea of putting limits on AI features you did not ask for, our guide to turning off Gmail's AI features covers the same instinct in Google's products.

What It Does Not See

More than you would expect, and this is the genuinely reassuring part. Anthropic carved several categories out of the recap entirely:

• Incognito chats are excluded completely.
• Any conversation that used a health integration is left out of your insights entirely. Anthropic names Apple Health and Health Connect as examples.
• Claude Cowork and Claude Code activity is not counted.
• Raw files and emails from connected services are not pulled in -- but read the next paragraph, because this is narrower than it sounds.

The connected-services carve-out has a second half worth knowing before you rely on it. Anthropic's exact wording: "When Claude pulls content from a connected service like Gmail or Google Drive, your recap doesn't incorporate the raw email or file. It may incorporate anything Claude wrote in the chat about it. For example, if you ask Claude to summarize your inbox, the emails themselves aren't pulled into your recap, but the summary Claude wrote is." So the documents stay out; Claude's description of the documents does not.

There is also a real limit on when the recap exists at all. Anthropic says that even with memory enabled, "your recap is only generated when you navigate to the Settings > Reflect page; if you don't land on that page, Claude won't create a recap for you." Nothing is being compiled about you in the background unless you go and look.

On sensitive subjects, Anthropic makes three specific commitments rather than a vague one: distress-related topics "never lead your recap", they are "never itemized with counts or percentages", and "where appropriate, support resources appear alongside your recap". So the recap is a summary of your topics and rhythms, not a searchable transcript of everything you have ever said.

Is Reflect Actually on Your Side?

Be a little skeptical, because the incentives do not line up. A tool that shows you how much you rely on Claude is being built by the company that wants you to rely on Claude. TechCrunch's read is that Reflect "subtly makes the case for why you should keep using it" -- laying out everything Claude helped with makes the product look indispensable, and the recap nudges you toward features like Projects that pull your work deeper into Claude and make switching harder.

That does not make it useless or dishonest. The quiet hours and break reminders are real controls that genuinely limit the product, and a company purely optimizing for engagement would not ship them. But hold both thoughts at once: a screen time report from Apple has no reason to want you on your phone more, while a usage report from Anthropic has every reason to want you in Claude more -- and it does tell you "where you might try something new." Treat the numbers as facts and the advice as marketing.

Should You Use It?

Look at it once, then decide. If memory is already on, the recap costs you nothing and a single glance at your topic breakdown and peak hour is a genuinely interesting mirror -- most people are surprised by what they actually use AI for versus what they think they use it for.

Skip it if you keep memory off on purpose. A usage chart is not worth reversing a deliberate privacy choice, and there is no way to get one without the other.

Do set the quiet hours if AI has crept into your evenings. That feature is the most useful thing in this release and it works whether or not you care about the recap.

If you are still working out which assistant deserves your money or your attention in the first place, our comparisons of the best free AI chatbot and which AI chatbot is worth paying for are the place to start. For what Claude's underlying models can actually do, see our explainers on Claude 5 Fable and Claude Sonnet 5.

The bottom line: Reflect is a free, mildly fascinating look at your own AI habits, with real controls attached and sensible privacy carve-outs. Just remember who is holding the mirror.

Sources

• Anthropic, "A new way to reflect on how you use Claude", July 9, 2026.
• Claude Help Center, "See your monthly recap".
• TechCrunch, "Anthropic's new Claude feature is quietly selling you on AI", July 9, 2026.