Ask Gemini in Google Search to plan a weekend for your parents and it might suggest a slow museum, a stroller‑friendly garden and an ice cream parlor — because it has already seen the parking reservation in your Gmail, holiday selfies in Google Photos and a search for “easy hikes for seniors.” That scenario, which testers are already reporting, captures why Google’s new Personal Intelligence feels both genuinely helpful and quietly intrusive.
What Google rolled out
On Jan. 22 Google expanded Personal Intelligence into AI Mode in Search, letting opt‑in users connect Gmail and Google Photos so Gemini can “reason across” that private material and tailor responses. The feature is initially available to Google AI Pro and AI Ultra subscribers in the U.S. as a Labs experiment, and Google says it won’t train Gemini directly on your inbox or photo library — the model accesses context to answer queries but training is confined to limited signals like prompts and replies.
Google’s product team frames this as Search finally meeting the messy, specific shape of your life. In practice that means itineraries informed by past bookings, shopping suggestions that favor brands you already wear, and reminders that surface because receipts and confirmations sit in your email.
For Google’s writeup of the feature see the company blog: Personal Intelligence in AI Mode in Search.
Small conveniences, big implications
The conveniences are immediate and palpable. Testers report the assistant pulling a car insurance renewal date from an email, spotting a license plate in Google Photos, and factoring in family details (a new baby, for instance) when planning travel. That’s the promise of a “personal assistant” that truly remembers.
But this is also part of a broader push: Google has been tying Gemini into more of its ecosystem — from Deep Research that can plumb Gmail and Drive for work documents to agentic booking features that act on your travel and appointment confirmations — making the assistant more proactive and capable over time. Those efforts are showing up in product experiments across search and workspace tools, not in isolation. For more on how Gemini’s research features are expanding into user files, see the recent coverage on Gemini’s Deep Research and Workspace integrations. And Personal Intelligence complements other AI Mode capabilities like automated booking and agentic actions that Google has been testing in AI Mode.
How Google says it protects data — and where uncertainty remains
Google emphasizes opt‑in controls and says the system uses steps to “filter or obfuscate” sensitive details from conversations. It also claims the underlying Gemini 3 model doesn’t train directly on raw Gmail or Photos content. Still, the company acknowledges mistakes can happen: inappropriate inferences, wrong connections, or recommendations that feel off.
There are practical safety mechanisms — ability to turn the integrations on or off, follow‑ups and corrections inside AI Mode, and a thumbs‑down feedback toggle. But control depends on users noticing and acting. The smarter an assistant gets at predicting your needs, the more you have to trust its assumptions.
Why regulators and privacy hawks will pay attention
This rollout arrives in a climate where scrutiny of big tech’s control over personal data is already high. Google’s unique advantage — decades of search logs, calendar entries, email receipts and photos tied to individual accounts — is precisely what makes Personal Intelligence powerful. It’s the same concentration of data that critics and regulators worry about when they talk about market dominance and privacy risk.
Expect questions about data minimization, retention, and whether those inferences could be repurposed for ad targeting or other uses down the road. Google’s assurances about not training on inboxes will matter to researchers and lawmakers; transparency about what signals are used, and clear user controls, will decide how many people actually opt in.
Practical tips if you try it
- Treat it like a feature that deserves a quick audit: check which apps are connected, and toggle off anything you don’t want the assistant to read.
- Use the feedback tools when responses are wrong; that’s how Google says the system will improve.
- Don’t feed it truly sensitive queries (health, legal or security matters) even if those items live in your account — AI answers can be persuasive but imperfect.
Google’s Personal Intelligence is a striking example of the tradeoffs built into modern AI: convenience born from concentrated data. For some, that tradeoff will feel like a revelation — Search that remembers. For others, it will feel like surrender.
Try it with eyes open. The feature is optional, and Google’s rollout will tell us more about how quietly — or loudly — personal assistants will become part of everyday search.