Google Drops Paywall on Personal Intelligence, Expanding Gemini to All US Users

Google on Tuesday opened its Personal Intelligence feature to all free-tier users in the United States, giving its Gemini AI assistant permission to scan Gmail inboxes, Google Photos libraries, YouTube watch histories, and more than a dozen other data sources to generate personalized responses. The feature launched in mid-January as a paid beta limited to AI Pro subscribers at $19.99 per month and AI Ultra at $249.99. Two months later, the paywall is gone.

Personal Intelligence is now live in AI Mode in Search for all US users and rolling out to the Gemini app and Gemini in Chrome, Google said Tuesday. Google did not say when other countries would get access. And the feature works only with personal Google accounts. Workspace business, enterprise, and education users are shut out.

The expansion converts Google's consumer product suite into a feeding system for its AI assistant. And it drops the price of admission to zero at the exact moment when every major AI company has landed on the same conclusion: personalized assistants create stickier users than generic ones.

Twelve services, one toggle

Personal Intelligence connects Gemini to a user's existing Google services. Gmail and Drive. Maps and Calendar. Photos, YouTube, Search history, even Google's travel booking tools. All of it pipes into the system. Ask Gemini a question, and it pulls context from whichever connected app has the answer. You skip the part where you explain your whole situation from scratch.

Google's blog post offered polished consumer scenarios. Ask for a bag recommendation, and Gemini checks your recent shoe purchases to suggest options with matching hardware colors. Can't remember your printer model when it jams? Gemini pulls the receipt from Gmail. Headed somewhere on vacation and want a real itinerary? It checks your hotel booking against your photo library, figures out what you liked last time, and builds something around your food preferences and travel habits.

Those examples read like marketing copy, and they partly are. But strip away the polish and the underlying infrastructure has real technical ambition. Personal Intelligence runs on Google's Gemini 3 family of large language models, which supports a context window of one million tokens, according to SiliconAngle. A typical user's Google account contains far more data than that limit allows. Google's solution is a technique called context packaging. It filters your account data before it reaches the model, selecting only the pieces relevant to what you actually asked. Everything else gets dropped. Your entire digital life does not fit in one prompt, and Google is not trying to make it fit.

Activating the feature takes several deliberate steps. In AI Mode, users tap their profile, select Search Personalization, then Connected Content Apps, and finally Connect Workspace and Google Photos. In the Gemini app, the toggle lives under Settings > Personal Intelligence. Everything is off by default. Google repeated the opt-in language three times across its blog post.

But the default-off framing masks a design incentive. The Gemini interface is likely to surface prompts and nudges steering users toward enabling Personal Intelligence. The feature works best when it knows the most. Google's two-month paid beta looks less like a subscription play and more like a stress test before flipping the switch for everyone.

The privacy fine print

"Gemini and AI Mode don't train directly on your Gmail inbox or Google Photos library," Google stated Tuesday. That sentence does careful work. The word "directly" shoulders the entire claim.

What Google does train on: the specific prompts users type into Gemini or AI Mode, and the model's generated responses. If you ask Gemini to summarize your last ten work emails about a project deadline, the prompt and the resulting summary both enter the training pipeline. Your inbox itself stays untouched. The AI's interpretation of your inbox does not.

This gap between user perception and technical reality will matter as the feature scales to hundreds of millions of accounts. Most people scrolling past Google's privacy statement on their phone will assume their personal data stays entirely outside AI training. The actual boundary sits in a different place, and Google seems content to let the ambiguity stand.

When Allison Johnson at The Verge tested Personal Intelligence during the January paid beta, she found that "Gemini can analyze my interests and make some pretty good guesses about what I'd be interested in; it's the details where AI gets lost." Useful enough to keep coming back. Not refined enough to trust without checking. That combination, good guesses paired with rough edges, is a product that trains its users to engage more, not less.

Google made a point of saying users can disconnect apps and revoke access whenever they want. Fair enough. But the training data generated from prior prompts and responses persists even after you flip the switch off. Revoking access stops Gemini from reading new data. It does not unwind what the model already learned from your questions about your own information. The off switch governs the future. It does not reach backward.

There is also a practical asymmetry in the opt-in design worth noting. Enabling Personal Intelligence takes five taps through nested menus. Understanding what you consented to requires reading Google's privacy notice, its Gemini-specific data policy, and the terms governing each connected app. The friction is front-loaded in the wrong direction. Turning it on is easy. Knowing what "on" means takes effort most users will skip.

The personalization arms race

Google is not building Personal Intelligence in isolation. Every major AI company has reached the same strategic conclusion over the past twelve months: generic chatbots are commodity products, and personalized ones are the path to retention that justifies the infrastructure spend.

When Google first launched Personal Intelligence in January, it joined a market already shaped by OpenAI's ChatGPT memory features and Meta's AI assistant, which debuted last spring with built-in access to Facebook and Instagram user histories. OpenAI pushed further still in February, rolling out personalized advertising inside ChatGPT for its 800 million users. That move collapsed the distance between personalization as a service and personalization as an ad-targeting engine.

Apple plans to add comparable capabilities to Siri later this year, connecting the assistant to emails, messages, files, and photos across Apple devices. Microsoft took a different route, integrating its Copilot assistant with consumer Microsoft 365 last January and opening up documents, emails, and calendars to cloud-based AI processing. Neither has matched the breadth of what Google is attempting.

Google's advantage in this race is structural. No other company controls as many consumer data sources under a single account system. Gmail, YouTube, Maps, Photos, Search, Drive, Calendar, Chrome browsing data, travel bookings. The list of connected services runs past a dozen. Apple routes Siri queries through on-device processing wherever possible, keeping personal data off cloud servers. Google takes the opposite approach, running Personal Intelligence entirely through its cloud infrastructure, where the full capacity of Gemini 3's million-token context window can be applied. More power, less privacy architecture.

The risk mirrors the advantage, almost perfectly. A privacy incident involving Personal Intelligence would not damage one product in isolation. It would corrode trust across every Google service connected to the system. Gmail and Photos and YouTube are not standalone apps for most users. They form the infrastructure of daily digital life. Connecting them all to a single AI reasoning layer concentrates both value and vulnerability in the same place.

What Google gains from giving it away

Dropping the paywall two months into a paid beta reveals where Google's priorities actually sit. This was never a subscription feature. It was a proving ground.

The economic logic runs in one direction, and it has nothing to do with monthly subscription fees. The more data Gemini absorbs from a user's Google account, the harder it becomes for that person to switch to a competitor. Someone new can sign up for ChatGPT or Claude in minutes. But neither assistant can instantly access years of email threads, a photo library tagged with faces and locations, YouTube viewing patterns, or navigation history from Maps. That accumulated personal context is the real product Google is protecting, and the company just removed the last barrier to building it.

Google appears emboldened by the beta results. Alphabet's stock edged higher Tuesday afternoon following the announcement, according to TipRanks, though the broader market was also up and isolating the signal is difficult. The speed of the free-tier rollout, just two months after the paid launch, suggests the beta generated the engagement numbers Google wanted to see.

Business and enterprise users remain locked out, and that exclusion is telling. Google kept Personal Intelligence away from business accounts, and the reason is probably less about product readiness than about legal exposure. Enterprise customers come with compliance requirements, data residency rules, and legal teams that actually read contracts before clicking Accept. Consumer accounts don't carry that baggage. Different liability, different rollout.

For the average US user staring at their Google profile settings, the proposition boils down to a single toggle. Turn on Personal Intelligence, and Gemini gets noticeably more useful. Connect more apps, and it gets smarter still. Give it access to everything, and switching to another AI assistant starts to feel like relocating to a city where nobody knows your name.

Google spent two months charging $19.99 to prove the feature worked, then gave it away for free the moment it was confident enough people would turn it on. The off switch exists. So does the bet that most users will never touch it.

Maria Garcia

Los Angeles

Tech culture and generative AI reporter covering the intersection of AI with digital culture, consumer behavior, and content creation platforms. Focusing on technology's beneficiaries and those left behind by AI adoption. Based in California.