Gemini Spark, Google's personal AI agent, can now work directly with your Google Photos library. Google Photos lead Shimrit Ben-Yair announced the addition on X on September 3, 2026 (US time), and Google's help center now lists the supported actions. Searching and curating photos and videos, enhancing image quality, generating collages, creating and sharing albums, and extracting text from images can all be chained together in one request. The rollout is aimed at Google AI Pro and Ultra subscribers in the US and will take place gradually over the coming weeks.

Six Kinds of Actions Combined Into One Request

Google's help page organizes what Gemini Spark can do in Google Photos into six categories. The first is search and curation: Spark can find photos and videos by subject, location, date or event, pick out the best shots and filter out duplicates. The second is editing and creation, covering quality enhancements, quick fixes, and the generation of collages or stylized images.

The third is album creation and organization. Spark can create new shared or private albums and save curated photos into them. The fourth is album sharing, where Spark adds or sends links to shared albums through Connected Apps such as Gmail, Google Docs and Messages. The fifth is scheduled automation, which puts recurring chores like building a monthly highlights album or sorting photos of receipts on a schedule. The sixth is text extraction: Spark finds the relevant images, pulls the text out of them, summarizes or reformats it in a Google Doc, or uses the information to carry out another task.

What makes this distinctly Spark is that these actions are not just available individually but can be combined into a single task. In one of Google's examples, a single prompt asks Spark to find the best 15 photos from last week's trip to Hawaii, enhance the beach shots, put them into a shared album called Aloha 2026 and draft an email with the link. Other examples include a recurring task that collects the top food photos every weekend, builds a collage and adds it to an album called What my camera ate this week, and a cross-app request that checks a photo of a local summer concert series against Google Calendar availability to work out which night the user can attend.

Originals Stay Untouched, and Sharing Requires Confirmation

Because a photo library is packed with personal information, Google spells out several safeguards in how the feature behaves. First, Spark never overwrites the original photo. When an edit is needed, it makes a copy, edits that, and saves it to Google Photos as a new image. Second, before actions that reach outside the library, such as creating a shared album or sending an email, Spark always asks for the user's permission. Newly created albums are private by default, and Spark asks again before sharing one.

At the same time, Google's general guidance for Spark points to the risks that come with handling Connected Apps and user data. The help center describes the possibility of prompt injection, where malicious instructions hidden in a web page, email or document could lead the agent to take unintended actions, and notes that Spark may share personal information with websites while completing a task. Google says user supervision remains essential for important tasks. Given that this means handing over an entire photo library, checking the results as you go is the realistic approach at first.

Requirements: US, 18 or Over, English Only

Several conditions apply to the Google Photos integration. Users must be 18 or over and located in the US (some US regions are excluded). They also need to be eligible for Gemini Spark itself, which means signing in with a personal Google Account, holding a Google AI Pro or Ultra subscription, and having Gemini Apps Activity turned on. Google Photos must also be connected to Gemini Apps as a Connected App in advance.

For now the feature is English only and available solely in the Gemini mobile app and on the web at gemini.google.com. To use it, switch to Spark from the menu in the Gemini app, type what you want done with your photos or videos into the text box, and submit. Specifying a time turns the request into a schedule, and typing a slash lets you call an existing skill. Asking Spark “What Google Photos skills do you have?” returns a list of the supported actions.

Google is enabling the feature gradually, so it may not appear immediately even for users who meet the requirements. No timeline has been given for availability outside the US.

Spark Has Kept Adding Connections Since Its May Debut

Gemini Spark launched in the US on May 19, 2026 as a personal agent feature with schedules and skills. Availability expanded on June 17, though several countries including Japan were still excluded at that point. On July 14, Spark reached Ultra subscribers in every region where Gemini Apps are supported except the European Economic Area, Nigeria, Switzerland and the UK, with support for all Gemini Apps languages. On July 16 it opened to Pro subscribers in the US.

Its connections have grown steadily too. Alongside Workspace services such as Google Calendar, Gmail, Docs, Sheets, Slides, Drive, Keep and Tasks, Spark supports third-party apps including Canva, Instacart, OpenTable, Dropbox and Zillow, and it can connect custom apps through a Model Context Protocol (MCP) server URL. Among these, Google Photos is one of the most personal data sources yet, and the addition signals Google's intent to establish Spark as an everyday organizer.

Spark also caps the number of tasks running at once at 15, and usage counts against the compute-based limits of Gemini Apps. Users who register many recurring schedules will want to keep that cap in mind.

Summary

On September 3, Google added Google Photos control to its personal AI agent Gemini Spark. Six kinds of actions, namely search and curation, enhancement and collage generation, album creation and sharing, scheduled automation, and text extraction from images, can be combined into a single request, with originals left untouched and confirmation required before sharing or sending email. The feature is for Google AI Pro and Ultra subscribers aged 18 or over in the US, English only for now, and is rolling out gradually over the coming weeks in the Gemini mobile app and on the web. No date has been given for Japan, so further announcements will be worth watching.