ChatGPT's Computer History Logs Your Clicks and Keystrokes on Mac

OpenAI has rolled out a new feature for ChatGPT's macOS desktop app called Computer History — and it does exactly what the name implies. According to ChatGPT's Computer History tracks your clicks and keystrokes, the feature logs your on-screen activity and turns it into a timeline that ChatGPT and Codex can reference when you ask them something. It learns how you work, suggests automations, and can even pick up tasks you left unfinished.

This isn't a quiet background update. It's a meaningful shift in what a desktop AI assistant can actually see — and it's already drawing comparisons to Microsoft's Windows Recall. The difference, OpenAI says, is in what it doesn't capture.

What Computer History Actually Does

The feature builds a running timeline of your computer activity — your clicks, keystrokes, the apps you open, the documents you touch. ChatGPT and Codex can then pull from that timeline when you make a request. Think of it as giving the model memory of your actual workday, not just the conversation you're having right now.

In a demo video cited by The Verge, Dominik Kundel from OpenAI's Developer Experiences team showed the app looking up the last document he edited, checking whether it had been shared via Slack, and giving a recap of how he spent his morning. That's genuinely useful context — the kind a good assistant would actually have.

OpenAI says Computer History does not capture images, videos, or audio. Instead it relies on what the company calls 'events' — structured data about what you did, not a visual record of your screen. That's the key distinction from Windows Recall, which leaned heavily on screenshots and drew intense privacy scrutiny.

The Privacy Controls Built In

Computer History is opt-in, which matters. You have to turn it on deliberately — it doesn't start logging by default. Once active, you can exclude specific apps and websites from being tracked. You can also delete individual entries if you want more granular control over what's in the timeline.

Ari Weinstein, Product and Engineering manager at OpenAI, confirmed on X that Computer History will automatically ignore content in incognito or private browser tabs. So your private browsing stays private — at least from this feature.

That's a reasonable set of guardrails. But the honest reaction here is: reasonable guardrails and complete peace of mind aren't the same thing. Logging keystrokes and clicks across your workday is a significant level of access, even if the data never leaves in screenshot form.

What This Means for Marketers and Agency Owners

Here's the real question worth sitting with: how much of your workflow do you want an AI model to see? For solo founders and small teams, a feature like this could be genuinely powerful. If you're constantly juggling client work across multiple tools — research docs, Slack threads, draft briefs — having ChatGPT remember the context of your morning could cut a lot of friction.

For agencies, the calculus gets more complicated. Client data flows through your machine constantly. Drafts, keyword research, competitive audits, proposal docs — some of that is confidential. Knowing you can exclude specific apps and websites helps, but it puts the responsibility on you to configure it correctly. One slip and you're logging something you shouldn't be.

There's also an AI visibility angle worth thinking about. OpenAI is building a model that understands how you actually work — your real patterns, not just what you tell it in a prompt. That's a meaningful leap toward AI systems that can proactively surface relevant information or suggest the next step in a workflow. That's useful. It's also a significant shift in how much context a commercial AI product holds about you personally.

And yes — if you're advising clients on AI tool adoption, this is the kind of feature that deserves a real conversation, not just a quick 'sure, sounds fine.' Privacy policies, data handling, and what employees are opting into on work machines all matter here.

What to Do Now

1. If you're using ChatGPT's macOS desktop app, check your settings now. Computer History is opt-in, but it's worth knowing where that toggle lives before a curious teammate turns it on.
2. Map out which apps and workflows touch client data on your machine. If you do enable Computer History, those are the apps you'll want to exclude from tracking immediately.
3. Talk to your team before anyone opts in. This is a workplace data decision, not just a personal preference. Set a policy before the feature sets the precedent for you.
4. Review any AI tool policies you already have in place. Computer History is a strong case study for why those policies need to cover not just what AI generates, but what it observes.
5. Watch for Codex integration closely. The fact that Codex can reference the activity timeline — not just ChatGPT — suggests this is heading toward agentic workflows where the AI acts on your behalf based on what it's seen you do. That's powerful and worth staying ahead of.

Background and Context

Microsoft's Windows Recall landed with a thud when it was announced. The concept of an AI that maintains a searchable record of everything you've done on your computer is genuinely useful — but the screenshot-based approach felt invasive, and Microsoft faced enough pushback to delay and significantly rework the feature.

OpenAI's Computer History is clearly positioned as a cleaner version of that idea. No screenshots, opt-in controls, private tab exclusions, deletable entries. Whether that's enough to earn broader trust is something the market will decide. But the direction is unmistakable: AI assistants are moving from reactive (answer my question) to proactive (I already know what you need).

That shift has real implications for how search, content, and AI-assisted work evolve over the next few years. If a model knows your full working context, the nature of how you query it changes completely. You stop explaining everything from scratch. That changes what 'a good AI answer' even looks like — and ultimately, it changes what content has to do to be useful inside those answers.

If you're already thinking about AI search visibility — how your content gets cited in tools like ChatGPT and Perplexity — features like Computer History are worth watching. They signal that the AI layer is getting more personal, more contextual, and more woven into daily work. Tools like AI visibility tracking can help you understand where your content is already showing up in that layer, and where there are gaps.