Google Gemini Personal Intelligence Explained: What It Does With Your Data
Key Takeaways
- Personal Intelligence is Google’s opt-in personalisation layer for Gemini. It lets the chatbot draw on your connected Google apps, including Gmail, Calendar, Drive, Photos, Maps, YouTube and Search, to give more personal answers.
- Nothing is connected by default. You choose whether to turn it on, and you can pick which apps Gemini may read, and disconnect them at any time.
- It first reached users in the United States who subscribe to Google AI Pro or Ultra, in beta, and it began rolling out on Android Auto on 20 October 2026. Cars with Google built-in are promised it in the coming months.
- It is not available in the European Economic Area, Switzerland, the United Kingdom or Nigeria, and it does not work with work, school or supervised accounts. You must be 18 or over and have a personal Google account.
- Google says it does not train directly on your inbox or photo library, but it does train on your prompts and Gemini’s responses, and its own help pages say content from connected apps may be used for service improvement, generative AI training and human review.
- Human reviewers may see material from connected apps. Google says it filters sensitive data before review, but independent analysis has questioned whether that filtering is adequate.
- You can turn the feature off, delete your chat history, use Temporary Chats that are not used for personalisation, and ask Gemini what information it used for an answer.
What Personal Intelligence Actually Is
Personal Intelligence is a feature inside Google’s Gemini chatbot that changes what the chatbot is allowed to know about you. In its basic form, Gemini answers from its training and from what you type into the chat. With Personal Intelligence switched on, it can also reach into your own Google apps and use what it finds there to shape its answers.
Google describes it as an optional personalisation layer. That wording matters, because it tells you two things at once: the feature is meant to make Gemini more useful to you personally, and it is not part of the core chatbot. You can use Gemini without it, and turning it on is a separate choice.
The idea is simple enough to state. Instead of you searching through thousands of photos or old emails yourself, you ask Gemini, and it scans the connected services to retrieve the specific information. Josh Woodward, vice president of Google Labs, Gemini and AI Studio, has framed the premise as context: the more Gemini knows about your life, the better it can help you.
The execution is where the questions start, because the feature needs deep access to your private communications and memories to work at all. That trade-off, between a genuinely more helpful assistant and the privacy cost of giving an AI model that level of access, is what this article explains.
What It Can Read
Personal Intelligence can draw on a wide set of Google services. According to Google’s own help pages and its announcements, the connected sources include:
- Google Workspace, including Gmail, Calendar and Drive. For this to work, smart features in other Google products need to be switched on.
- Google Photos, including your photo library. In some countries, using Photos requires Face Groups with your face selected, plus the “Estimate missing locations” setting.
- Google Maps, including places and location-related information.
- YouTube, including your watch history.
- Google Search, including Search history and related services such as Shopping, News, Flights and Hotels.
That is a broad sweep of a person’s digital life. Your emails hold conversations, bookings, receipts and account details. Your photos hold faces, places and dates. Your search and YouTube history hold interests and habits. Your calendar holds where you will be and with whom. Put together, these sources give Gemini a detailed picture of a person, far beyond what a chatbot normally sees.
It is worth being precise about how the access works. Gemini does not read everything at once. It uses tool calling, meaning it invokes tools such as Gmail search by itself when it judges that information from a connected app would help answer a question. Whatever the search turns up is inserted into the conversation’s context, the text the model takes into account when generating its answer. Independent testing reported on GitHub has shown that Gemini can fetch emails and files in the background on its own, even when the user never asked it to search a particular app.
Google says Gemini will try to show where it pulled information from, so you can verify an answer. That source referencing is an important control, because a personalised answer that quietly draws on a misread email is worse than a generic answer that admits its limits.
How It Differs From Gemini’s Memory
Personal Intelligence is often confused with Gemini’s memory, and the confusion is understandable, because Google has used overlapping names for related settings. The two are different things, and the distinction matters for privacy.
Gemini’s memory, also called personal context on some accounts, is what the chatbot learns from your past chats with it. Google added this in August 2025, and it is on by default. It means Gemini can remember things you told it in earlier conversations and use them later. On some accounts, the setting appears under the name Personal Intelligence rather than personal context, which adds to the confusion.
Personal Intelligence, as a distinct feature, goes further. It is not about what you told Gemini. It is about what is already sitting in your Google apps. It lets Gemini reason across your Gmail, Photos, Calendar, Drive, Maps, YouTube and Search, pulling specific details without you explicitly asking it to search a particular app.
There is also a memory summary, a written summary of what Gemini understands about you, which you can view, edit and delete. As of November 2026, that is available in the United States, for Google AI plan subscribers, in English only. You can also ask Gemini directly, for example by asking “What do you know about me?” or “Did you use any info from past chats?”, to see whether an answer drew on your history.
Keeping the two apart helps you make a clearer decision. Turning off memory stops Gemini learning from your chats. Turning off Personal Intelligence stops it reading your apps. They are separate switches, and you may want one on and the other off.
Who Gets It, and Who Does Not
Personal Intelligence is rolling out gradually, and eligibility is narrower than many people assume.
The feature first reached paid subscribers in the United States on Google AI Pro and Ultra plans, in beta, with plans to expand globally and to free users later. It is also rolling out on Android Auto, where it began on 20 October 2026, and Google says cars with Google built-in will get it in the coming months.
To be eligible, you need to be 18 or over, to have a personal Google account rather than a work, school or supervised one, and to have Keep Activity switched on. If you run your business on a Google Workspace work account, you will not see the feature there at all, because work, school and supervised accounts are excluded.
It is also not available in the European Economic Area, Switzerland, the United Kingdom or Nigeria. That exclusion covers a large part of this site’s audience, which is worth stating plainly. If you are in the UK or the EU, you cannot switch Personal Intelligence on at the time of writing, and Google has not given a date for when that might change. The feature also only runs in the Gemini mobile app, the web app at gemini.google.com, Gemini in Chrome and Gemini on a smartwatch, and it is not available in certain features such as Gems.
Because the rollout is gradual, even eligible users may not see it immediately. If you meet every condition and still do not see the option, that is expected, not a fault with your account.
How to Turn It On, and How to Turn It Off
Personal Intelligence is opt-in. Google states that apps are not connected until you turn the feature on, and that it can be switched off at any time in Gemini settings.
On the web, the process starts at gemini.google.com. If you are eligible, a card appears inviting you to opt in. You can then choose which apps to connect. You do not have to connect everything. If you want Gemini reading your Photos but not your Gmail, that is an option, and you can change your choices later. Connected apps can be disconnected in Gemini settings, under the Connectors section, and app connections can also be managed under Settings, then Personal context or Personal Intelligence, depending on how the setting is labelled on your account.
Turning the feature off is equally straightforward, but there is a detail that catches people out. Disconnecting an app, or deleting information in that app, does not delete information already stored in Gemini Apps Activity. That activity must be managed separately. If you want a clean break, you need to deal with the activity log as well as the connection itself.
Your chat history is managed under Activity on gemini.google.com. There you can turn Activity off, delete it, or set it to auto-delete after 3, 18 or 36 months. Eighteen months is the default. Chats that are reviewed by humans are kept for up to three years, which is a separate retention period from the activity log itself.
Temporary Chat is the most private way to use Gemini. It is a private thread that is not saved to history and is not used for personalisation or training. It is kept with your account for 72 hours for service and safety purposes, then discarded. If you want to ask Gemini something without it affecting what the chatbot knows about you, Temporary Chat is the right tool.
What Google Says About Privacy
Google’s privacy position on Personal Intelligence has several layers, and they do not all point in the same direction. It is worth setting them out carefully, because the gap between them is where most of the concern sits.
First, Google says the feature is off by default, and that you choose whether to enable it and which apps to connect. That is the headline control, and it is real.
Second, Google says Gemini will not train directly on your inbox or photo library. The company frames this as a privacy advantage, on the grounds that your sensitive data already lives on Google’s servers rather than being sent to a third party. As Google has put it, because the data already lives at Google securely, you do not have to send sensitive data elsewhere to start personalising the experience.
Third, and this is the part that needs attention, Google does train on your prompts and on Gemini’s responses. That means that if you ask a question containing sensitive personal information drawn from your email, that prompt can be used for training, and it can be reviewed by humans. Google’s support documentation says connected data may include sensitive or confidential information and may be used for service improvement, generative AI training and human review under the applicable settings and notice.
Fourth, Google acknowledges the feature is not perfect. Users may encounter what the company calls over-personalisation, where Gemini leans too heavily on personal context and gives answers that are wrong, or that assume too much. The feature is in beta, and Google’s own materials describe it as something still being refined.
Fifth, Google says Gemini will try to show where it pulled information from, so you can check its work. That is a useful safeguard, but it depends on you actually checking, and on the references being accurate.
Taken together, Google’s position is that your raw inbox and photo library are not themselves training data, but that anything you ask about them, and anything Gemini says in reply, can be. That is a narrower promise than many users assume when they hear “we don’t train on your data”, and it is worth understanding before you switch the feature on.
The Human Review Question
The most sensitive part of Personal Intelligence is not what the model does with your data, but who else might see it. Google’s help pages now state explicitly that content from connected apps can end up in front of human reviewers. Those reviewers may be Google employees or outside contractors who read and rate conversation logs to improve the AI.
That is a change from an earlier position. When Google launched the Workspace extensions for Gemini in 2023, a blog post on its responsible approach to generative AI made a clear commitment: personal content from Gmail, Docs and Drive would not be seen by human reviewers. According to analysis posted on GitHub and discussed on Hacker News in November 2026, that very post was changed on 16 October 2026, and Google’s help pages now say something different.
The GitHub analysis, by a researcher using the handle thisisme6598, also reported a test suggesting Gemini does not even need to be asked about your files to pull them in. The test showed Gemini fetching emails and files in the background, and quoting two to four sentences verbatim from a file, often including personal data, account details or amounts. In some cases the complete text of an email or file landed unredacted in the conversation context.
Google says it automatically filters sensitive data before review, using regular expressions, which are search patterns that detect things like credit card numbers, and named entity recognition, a technique that finds names, places and organisations in text. The GitHub author considers that inadequate. Narrative text cannot be reliably anonymised this way, he argues, because account numbers, confirmation numbers and transaction data slip through, and even the combination of a first name, a city and an event can identify a person.
We have not independently reproduced the GitHub test, and we report its findings as claims from that analysis rather than as our own verification. But the underlying point is not in dispute: Google’s own documentation now says connected app content may be human reviewed. If you connect apps containing confidential information that you would not want a reviewer to see, that is a real risk to weigh, and the advice from privacy researchers is straightforward. Do not connect apps to Gemini if they contain material you would not want a human reviewer to read.
The Android Auto Rollout
On 20 October 2026, Google announced that Personal Intelligence was beginning to roll out on Android Auto, as part of a wider announcement that Gemini itself was now available globally on Android Auto and in cars with Google built-in.
In the car, the feature works the same way it does on the phone. With permission, Gemini can retrieve details and reason across connected Google services, including Gmail, Calendar, Photos, Maps, YouTube and Search. Google’s examples include finding a flight gate or navigating to an appointment, using information drawn from the driver’s own apps.
The car rollout has the same conditions as the phone rollout. It is opt-in, limited to personal Google accounts of people 18 and older, and not available in the European Economic Area, Switzerland, the UK or Nigeria. If you already turned Personal Intelligence on in the Gemini phone app, it carries over to the car automatically. Cars with Google built-in, meaning vehicles that run Google’s software directly without a phone, will get it in the coming months, and Google has not given a specific date.
There is one practical point worth noting about the car. A spoken request in a car is often heard by passengers, and a personalised answer that draws on your email or calendar may say things aloud that you would not want others to hear. That is a small detail, but it is the kind of thing that matters in practice, and it is worth thinking about before you enable the feature in a shared vehicle.
A Quiet Reversal Worth Knowing About
There is a revealing footnote to the Personal Intelligence story that says something about how Google is handling the feature. For about three weeks in October 2026, Google’s help article for Tasks in Gemini, the feature formerly known as Gemini Spark, listed opting in to Personal Intelligence as a requirement. That meant that, for a short period, handing Gemini access to your Gmail and Photos was the price of entry to a separate feature.
Google has since taken that requirement back off the list. The Internet Archive’s capture of the help article from 25 October 2026 still carries the line, but the live page on 9 November 2026 does not. The requirements list now has four items, being 18 or over, having a personal Google account, having Google AI Pro or Ultra, and having Keep Activity on, and Personal Intelligence is not one of them. Google announced neither the addition nor the removal.
The episode matters for two reasons. First, it shows that Google is still working out where the boundary sits between Personal Intelligence and its other features, and that the boundary can move without an announcement. Second, it confirms that Personal Intelligence is, at the time of writing, a separate choice again, rather than something bundled into another feature. Handing Gemini your Gmail and Photos is optional, and it should stay that way unless you decide otherwise.
Should You Turn It On?
There is no single right answer, because the trade-off depends on what you use Gemini for and how much you care about the data involved. But there are some clear guidelines that follow from the facts above.
Turn it on if you are in an eligible region, you have a personal Google account, and you want Gemini to answer questions that genuinely need your own information, such as finding a booking in your email, or identifying a place in your photos. Start with one app rather than all of them, and see whether the answers are actually better before you connect more.
Think carefully before connecting Gmail or Drive if those apps hold work documents, financial records, medical information or anything else you would not want a human reviewer to read. The risk is not that Google will publish your emails. It is that a reviewer, or a prompt that ends up in a training set, will carry a fragment of your private life somewhere you cannot control.
Use Temporary Chat for anything sensitive, even with Personal Intelligence on. It is the one mode that is not saved, not used for personalisation and not used for training, and it costs you nothing to use it.
Check the activity log regularly, and set auto-delete to the shortest period you are comfortable with. Three months is the shortest option Google offers, and it is a reasonable default if you do not need a long history.
And if you are in the UK, the EU, Switzerland or Nigeria, the decision is made for you for now. The feature is not available, and there is no date for when that changes. That may be frustrating, but it also means you have time to watch how the feature develops in the United States before it reaches you.
For readers who want to understand Gemini more broadly, our earlier coverage is a useful starting point. Our guide to Gemini 4 Carbon explains Google’s frontier model, our piece on Gemini 4 Hydrogen pricing and benchmarks covers the faster model in the same family, and our guide to five things to never tell ChatGPT sets out practical privacy rules that apply just as well to Gemini. Our overview of Google AI Pro in 2026 explains the subscription that Personal Intelligence currently requires.
What We Could Not Verify
- We could not verify from Google’s own pages the exact date Personal Intelligence first became available to US Google AI Pro and Ultra subscribers. Reporting puts the US beta earlier in 2026, with the Android Auto rollout dated 20 October 2026, but Google has not published a single clear launch date for the feature as a whole, and we have not stated one as fact.
- The GitHub analysis of Gemini fetching emails and files in the background, and of the changes to Google’s 2023 blog post, is a third-party claim. We have reported it as such and have not independently reproduced the test.
- Google has not published figures for how many users have enabled Personal Intelligence, how often human reviewers see connected app content, or how accurate the source references are. We have not invented any such figures.
- The exact list of connected apps varies slightly between Google’s announcements and secondary reporting. TechCrunch’s report, for example, names Gmail, Calendar, YouTube, Maps and Search history without Photos, while Google’s post and other coverage include Google Photos. We have reported the fuller list from Google’s own pages and noted the difference.
- We could not verify whether or when Personal Intelligence will become available in the UK, the European Economic Area, Switzerland or Nigeria, or to free users. Google has said it plans to expand globally and to free users, but has given no dates.
Sources
- Google, “Personal Intelligence”, Gemini help pages — Google’s own description of the feature, its connected apps, eligibility and controls.
- Google, “Get hands-free personalized help on the road with Gemini”, The Keyword, 20 October 2026 — the Android Auto rollout announcement.
- Lilach Bullock, “Gemini Personal Intelligence: What It Reads, Who Gets It”, updated 10 November 2026 — https://www.lilachbullock.com/gemini-personal-intelligence/
- Fello AI, “What Does Gemini Know About Me? 7 Prompts to Check” — https://felloai.com/7-powerful-prompts-that-reveal-what-google-gemini-knows-about-you/
- Tokenstrom, “Gemini can show your private emails and Drive files to human reviewers” — https://tokenstrom.com/posts/gemini-can-show-your-private-emails-and-drive-files-to-human-reviewers/
- NewsPoint, “Explained: Google Gemini’s Personal Intelligence, and how your emails, photos, and search history will help AI answer questions better” — https://www.newspointapp.com/english/trending/explained-google-geminis-personal-intelligence-and-how-your-emails-photos-and-search-history-will-help-ai-answer-questions-better-gadgetsnow/articleshow/145048200fcc538f6d0221ee12a920b1f32677e3
- Access All GPT, “Google Brings Gemini Personal Intelligence to Android Auto” — https://www.accessallgpt.com/research/google-gemini-personal-intelligence-android-auto-cars-rollout-privacy-limits
- Plain AI Daily, “Gemini Spark Is Now Tasks in Gemini: What Changed and Do You Get It?” — https://plainaidaily.com/guides/gemini-spark-explained