What Is ChatGPT Computer History? OpenAI's Screen-Tracking Feature Explained
ChatGPT's Computer History feature tracks your desktop activity to build context and suggest automations. Here's how it works and what it means for privacy.
What is ChatGPT Computer History?
Computer History is a new ChatGPT desktop feature that watches what you do on your computer, across apps and websites, and uses that activity to build context, answer questions about your day, and suggest tasks it could automate for you. Unlike ChatGPT’s existing memory, which only stores what you type into it, Computer History observes what you actually do: which documents you opened, what you typed, which folders you moved files into, and which sites you visited. OpenAI says it does this without recording your screen or audio, relying instead on activity summaries.
TL;DR
- Computer History is a ChatGPT desktop feature (Mac only at launch) that tracks app and browser activity to build a running record of your work, not just what you tell the chatbot directly.
- It requires a paid plan (Pro, Business, or Enterprise) and is currently unavailable in the EU, UK, and Switzerland.
- The standout use case is automatic skill creation: after you finish a task, you can ask ChatGPT to turn your recent activity into a reusable “skill” without manually explaining each step.
- It can also run an automation audit, scanning your recent activity and telling you which tasks are “highly automatable” versus which need a human.
- Users control scope through an include-only or exclude list of apps, and can pause or clear history at any time, though the default settings require active management to avoid over-sharing.
- The feature draws an inevitable comparison to Microsoft Recall, which faced heavy backlash in 2024 for similar screen-tracking behavior, though the two systems store and process data differently.
- The bigger story is about context: giving an AI assistant sequential, real-world visibility into your workflow is a different kind of input than prompts, memory, or connectors, and it may be the missing piece for making agentic AI genuinely useful to non-technical users.
Seven tools to build an app. Or just Remy.
Editor, preview, AI agents, deploy — all in one tab. Nothing to install.
How does Computer History actually work?
Once enabled in ChatGPT’s desktop settings under the Integrations tab, Computer History runs in the background and logs activity across the apps and browser tabs you’ve allowed it to see. It doesn’t just note that you opened a document. It captures the sequence of what you did: opening a file, editing it, saving a PDF, moving that PDF into a new folder, then attaching it to an email. That sequence is what makes the feature different from a simple activity log or browser history.
You can then ask ChatGPT questions like “What have I done today?” and get a summary of your session. More usefully, you can ask it to turn a completed workflow into a reusable skill. Instead of manually describing every step you took, ChatGPT reconstructs the steps from what it observed and packages them into something you can reuse the next time a similar task comes up.
It can also proactively suggest automations. Asked which of the day’s tasks could be automated, it can flag items as “highly automatable” or “low automation potential,” and it appears to factor in the actual content of what you were working on, including to-do lists buried inside a document, not just the fact that you opened the document.
How do you enable and scope it?
Computer History lives in ChatGPT desktop settings, under an Integrations category. Turning it on prompts a permissions screen explaining that ChatGPT can summarize activity across apps and websites without recording your screen or audio, and reminding you that you’re responsible for pausing it during sensitive activity like video calls or text conversations with other people, unless those people have consented.
From there, you choose a scope in one of two ways: an exclude list (everything is tracked except the apps you name) or an include-only list (nothing is tracked except the apps you explicitly allow). The exclude model is the default, which means it captures everything unless you remember to carve out exceptions, an approach that puts the burden on the user to think of every sensitive app in advance. Switching to include-only and manually adding only the apps you actually want tracked, such as a browser, a text editor, or a file manager, is the more conservative approach.
You can pause, clear, or adjust this at any time, and OpenAI’s own settings language recommends excluding categories like health or finance apps by default.
Is Computer History worth the privacy tradeoff?
That depends entirely on how much you trust the scoping tools and how disciplined you are about using them. The core tension is real: this feature only becomes powerful once you grant it access to things like your main browser, which for most people means email, banking sessions, work documents, and personal browsing all funneled through one window. Once Chrome is included, a huge share of daily digital life is visible to the system.
Everyone else built a construction worker.
We built the contractor.
One file at a time.
UI, API, database, deploy.
The counterargument is that this is opt-in, not automatic, and the scoping controls are genuinely granular. You can limit tracking to a single folder and a text editor if that’s all you’re comfortable sharing, and still get some of the benefit. The feature is also currently gated behind paid plans and geography, meaning OpenAI isn’t rolling this out to its entire user base by default.
The honest answer is that the value of the feature scales with the access you give it, and so does the risk. A narrow setup (say, just a project folder and a note-taking app) is low-risk and moderately useful. A broad setup that includes a full browser session is where the real power (and the real exposure) shows up.
How does this compare to Microsoft Recall?
Microsoft Recall, announced in 2024, took periodic screenshots of a user’s activity and used on-device AI to make that history searchable. It was met with an immediate and severe backlash over security and privacy concerns, enough that Microsoft delayed and reworked the rollout multiple times.
Computer History is conceptually similar in that both systems build a memory of what you did on your machine rather than what you told an AI directly. The stated mechanics differ: OpenAI says its feature summarizes activity without capturing screen recordings, while Recall’s original design was built around screenshots. Whether that distinction meaningfully changes the privacy calculus is a matter of ongoing debate, and it’s likely to be litigated in public opinion the same way Recall was.
One difference in timing matters too. Recall launched before the current wave of agentic AI tools, when the practical use case for “the AI remembers what you did” was mostly search and recall. Computer History arrives at a moment when AI agents can act on that context: creating skills, drafting automations, and chaining tasks together. That changes the stakes on both sides. The convenience case is stronger, and so is the surface area for something to go wrong.
Why does this matter for AI automation?
Large language models are ultimately as good as the context they’re given. Prompts, memory, connectors to other apps, and now something like Computer History are all just different mechanisms for feeding an AI more relevant information about what you’re trying to do. What Computer History adds is something none of those other mechanisms provide on their own: a sequential, real-time record of how a piece of work actually unfolded, not a static snapshot or a manually written prompt.
That sequencing is what allows the automation-suggestion behavior to work at all. An AI that only sees a finished document has no idea how you got there. An AI that watched you open a file, edit it, save it, and email it can reconstruct the workflow and offer to package it as a reusable skill. That’s a meaningfully different capability than memory or file access alone, and it’s part of why this feature has generated more discussion than a typical settings update.
Frequently Asked Questions
What is ChatGPT Computer History?
It’s a ChatGPT desktop feature that observes your activity across chosen apps and websites to build context about your work, answer questions about what you did, and suggest tasks you could automate.
Does Computer History record my screen or audio?
Remy is new. The platform isn't.
Remy is the latest expression of years of platform work. Not a hastily wrapped LLM.
OpenAI states that it summarizes activity across apps and websites without recording screen video or audio. It relies on activity summaries rather than screenshots or recordings.
Who can use Computer History right now?
It’s available in the ChatGPT desktop app on Mac only, and requires a Pro, Business, or Enterprise plan. It is not currently available in the EU, UK, or Switzerland.
How is this different from Microsoft Recall?
Both track user activity to build a searchable or usable memory of past work. Recall relied on periodic screenshots; OpenAI describes Computer History as using activity summaries rather than screen captures. Both raise similar privacy questions about how much of a user’s digital life should be visible to an AI system.
Can I control what Computer History sees?
Yes. You can choose an exclude list (track everything except named apps) or an include-only list (track nothing except named apps), and you can pause or clear the history at any time from settings.
