Google Tests AI Mode With Gmail And Photos Access

Google is piloting a new capability called Personal Intelligence that lets its AI Mode draw on context from Gmail and Google Photos to craft tailored answers. The opt-in feature, powered by the Gemini 3 model, promises responses that reflect a user’s plans, preferences, and past activity—while keeping emails and images out of model training, according to the company.

What Personal Intelligence Actually Does

Personal Intelligence aims to collapse the gap between generic AI replies and the nuanced reality of a user’s life. Ask for a winter coat recommendation, and it can factor in where you’re flying next (from your ticket in Gmail), the typical weather on arrival, and what you actually like to wear (gleaned from Google Photos) before suggesting options. Planning a family trip? It can infer each person’s interests from shared photo history and propose an itinerary that fits.

It also leans into more playful prompts. You can ask it to write the title of a biopic about your life or describe an ideal day, and it will use memories and artifacts from your account to make the output feel personally grounded.

How It Works Across Google Search And Gemini

The feature debuted in the Gemini app and is now available in AI Mode within Search as a Labs experiment. Access is rolling out to Google AI Pro and AI Ultra subscribers who opt in. Within AI Mode, Personal Intelligence tries to infer intent and context automatically, which can make seemingly simple prompts—“Where should we eat tonight?”—surprisingly specific, factoring in past preferences, dietary notes in messages, or recent reservations.

The system runs atop Gemini 3, Google’s latest multimodal model. In practice, that means it can reason across text and images, and it’s better at fusing cues from disparate sources—like an emailed confirmation and a batch of recent photos—into a single reply.

Privacy Controls And Data Use For Personal AI

Google says it will not directly use your emails or photos to train Gemini’s underlying models. Instead, the company may use specific prompts and responses to improve feature quality over time. The feature is disabled by default and requires explicit permission to connect Workspace data (including Gmail) and Google Photos.

The opt-in approach mirrors how other tech firms are handling personal-context AI. Privacy groups such as the Electronic Frontier Foundation and the Future of Privacy Forum have cautioned that any assistant with access to personal archives must be transparent, revocable, and tightly permissioned. Google’s setup allows feedback on each answer via a thumbs icon and granular control over connected sources, but users should still review what’s shared and periodically audit settings.

Availability And How To Enable It On Your Account

Eligible subscribers will receive invites over the coming days. If you don’t see it, you can manually enable it by opening Search, tapping Profile, choosing Search Personalization, selecting Connected Content Apps, and toggling on Workspace and Google Photos. It remains opt-in and can be turned off at any time.

Google notes that the feature performed well internally but can still misread context or produce errors. When it misses, downrating a response helps tune future results. As with any AI assistant, double-check time-sensitive or high-stakes recommendations.

Why This Matters For Google And Everyday Users

The move is strategic. Gmail and Google Photos each serve well over a billion users, and tapping those troves responsibly could make Google’s assistants feel meaningfully smarter without requiring users to change habits. If Personal Intelligence lands as intended, it could nudge AI from novelty toward necessity—speeding up shopping choices, travel planning, and personal admin that generic chatbots tend to handle poorly.

It also positions Google against increasingly personalized rivals. Apple is weaving device-level context into Apple Intelligence, while enterprise tools from Microsoft and OpenAI are experimenting with permissioned data connections. The differentiator will be trust: the assistant that’s most helpful without overstepping will likely win sustained use.

Early Limitations And Real-World Examples

Even with better context, AI can still hallucinate or overweight weak signals. A photo of a beach wedding doesn’t mean you prefer coastal vacations; a forwarded itinerary might be a friend’s trip, not yours. Expect growing pains—especially with shared family libraries, group emails, and ambiguous prompts.

Used carefully, the upside is tangible. A parent could ask for quick meal suggestions based on past family favorites spotted in Photos and calendar gaps in Gmail. A business traveler might get packing guidance keyed to meeting locations, dress codes mentioned in email threads, and the weather on arrival. The goal is less typing, fewer tabs, and more on-target answers.

The Bottom Line On Google’s Personal Intelligence

Personal Intelligence is Google’s boldest step yet toward AI that understands the user as much as the query. With strict opt-in, explicit data connections, and visible feedback tools, it’s a cautious but consequential expansion. If Google balances personalization with privacy—and if Gemini 3 delivers consistent accuracy—AI Mode could become the default way many people ask Google for help.