Marissa Mayer’s Dazzle AI Launches Photo-Based Personal Assistant

Dazzle AI, the consumer startup founded by entrepreneur Marissa Mayer, announced the launch of Dazzle on September 29, 2026, an AI personal assistant that the company says builds its understanding of each user from photos rather than from a blank prompt.

The debut was announced in “Announcing Dazzle: Pictures, Not Prompts”, a post on the company’s blog by Danielle Kochavy. In it, the company describes Dazzle as an assistant that learns from a user’s photos to become more personalized and proactive, using the context in a camera roll to understand who the user is and what matters to them, while scanning recent photos for things it can help the user get done, buy, fix, or plan.

The Camera-Roll Premise

The post argues that most AI experiences still begin with a blank prompt and suffer from what it characterizes as a cold-start problem, and that assistants drawing on emails and calendars for context skew toward work-oriented tasks. Dazzle starts from the camera roll instead: the post states that 5 billion photos will be taken on smartphones on September 29, 2026, and presents each photo a user takes as a possible prompt — something to buy, fix, add to a calendar, or remember.

According to the company, analyzing those photos lets Dazzle learn a user’s style, interests, hobbies, travel, the people who matter to the user, the things the user owns, and what the user likes, without requiring the user to answer a long list of questions first. The premise, the post states, is that people take pictures and screenshots of tasks they need to do every day.

Features and Availability

The post describes two differentiators: turning a photo the user has just taken into a useful action, and synthesizing contextual information from photos into personalized AI.

Under a feature called Get It Done, Dazzle analyzes photos and screenshots as they are taken and proactively suggests what needs to be done. In the company’s examples, a user who photographs a product such as a jacket can have Dazzle identify it, find where it is available, and buy it, while a user who photographs a damaged floor, a flat tire, or a broken garage door can have Dazzle identify an appropriate professional and arrange the repair. The feature can also add events to a calendar and save QR codes to a wallet.

A second feature, Ideas, offers personalized recommendations based on past photos and screenshots, with the post naming restaurants, concerts, and travel as examples; if the user accepts a suggestion, the company says Dazzle can carry it out within a few taps. Dazzle’s chat runs on frontier models enriched with personal context, according to the company, and can field questions such as what gift to buy a nephew or how to spend a weekend.

Dazzle is available in iMessage and by text, in an iOS app, and on the web. Texting Dazzle a photo of a school flyer, invitation, or appointment card prompts it to extract dates and add them to a calendar; the app learns milestones such as a child’s birthday from past photos and suggests ways to celebrate; and the web version at dazzle.ai handles larger projects such as vacation planning, including destinations, places to stay, experiences, and a full itinerary. The company also offers an MCP connection so that Dazzle’s photo intelligence can be used inside Claude, Codex, and OpenClaw.

Privacy Design

The company says Dazzle is opt-in: users can grant access to a full photo library, select specific photos, or grant no photo access at all. The post states that if Dazzle suspects a photo could be sensitive, it immediately excludes it.

The agent also acts “on your behalf, not as you,” according to the post: each Dazzle agent has its own accounts and tools, works without broad credential-level access to a user’s accounts, and the company says it made a fundamental decision early in development not to impersonate a user.

The company states that personal information is used only to personalize Dazzle for the individual user, not to train public AI models, not for third-party advertising targeting, and not to be sold or rented. It says its agreements with third-party AI providers prohibit those providers from using personal information to train their own models, and it cites encryption in transit and at rest, secure authentication, access controls, and ongoing security testing and monitoring.

Company Background and Seed Funding

Dazzle AI was founded in 2025 by Mayer, whose experience includes Google and Yahoo, and the company states its mission as closing the gap between what people want to do and what they can do with AI. Its About page describes the product as a next-generation personal AI agent powered by context and intelligence gleaned from users’ photos.

On December 23, 2025, the company announced in a funding release that it had raised $8 million in seed funding at a $35 million post-money valuation. The round was led by Kirsten Green of Forerunner, with participation from Kleiner Perkins, Greycroft, Offline Ventures, Slow Ventures, and Bling Capital.

In the funding release, Mayer said Dazzle AI’s goal is to make AI feel simple and to help people get more done, adding: “Now that foundational models have reached a level of consistent excellence, they’ve become a reliable infrastructure. The new frontier is applications—leveraging that power to create real, tangible value.”

The December 2025 release said the funding would allow the company to expand its team and prepare for its first product launch in the coming months. Dazzle is being offered by invitation: users can request an invite through the iOS app, at dazzle.ai, or by texting 742213.