Google isgiving an AI upgradeto its Collections feature — basically, Google’s own take on Pinterest, but built into Google Search. Originally a name given toorganizing images, the Collections featurethat launched in 2018let you save for later perusal any type of search result — images, bookmarks or map locations — into groups called “Collections.” Starting today, Google will make suggestions about items you can add to Collections based on your Search history across specific activities like cooking, shopping or hobbies.

The idea here is that people often use Google for research but don’t remember to save web pages for easy retrieval. That leads users to dig through their Google Search History in an effort to find the lost page. Google believes that AI smarts can improve the process by helping users build reference collections by starting the process for them.

Here’s how it works. After you’ve visited pages on Google Search in the Google app or on the mobile web, Google will group together similar pages related to things like cooking, shopping and hobbies, then prompt you to save them to suggested Collections.

For example, after an evening of scouring the web for recipes, Google may share a suggested Collection with you titled “Dinner Party,” which is auto-populated with relevant pages from your Search history. You can uncheck any recipes that don’t belong and rename the collection from “Dinner Party” to something else of your choosing, if you want. You then tap the “Create” button to turn this selection from your Search history into a Collection.

These Collections can be found later in the Collections tab in the Google app or through the Google.com side menu on the mobile web. There is an option to turn off this feature in Settings, but it’s enabled by default.

The Pinterest-like feature aims to keep Google users from venturing off Google sites to other places where they can save and organize things they’re interested in — whether that’s a list of recipes they want to add to a pinboard on Pinterest or a list of clothing they want to add to a wish list on Amazon. In particular, retaining e-commerce shoppers from leaving Google for Amazon is something the company is heavily focused on these days. The company recently rolled outa big revamp of its Google Shopping vertical, and just this month launched a way toshop directly from search results.

The issue with sites like Pinterest is that they’re capturing shoppers at an earlier stage in the buying process — during the information-gathering and inspiration-seeking research stage, that is. By saving links to Pinterest’s pinboards, shoppers ready to make a purchase are bypassing Google (and its advertisers) to check out directly with retailers.

Meanwhile, Google is simultaneously losing traffic to Amazon, which nowsurpasses Google for product searches.Even Instagram, of all places, has become a rival, as it’s now a place to shop. The app’s Shopping feature is funneling users right from its visual ads to a checkout page in the app. PayPal, catching wind of this trend, recentlyspent $4 billion to buy Honeyin order to capture shoppers earlier in their journey.

For users, Google Collections is just about encouraging you to put your searches into groups for later access. But for Google, it’s also about getting people to shop on Google and stay on Google, no matter what they’re researching. Suggested Collections may lure you in as an easy way to organize recipes, but ultimately this feature will be about getting users to develop a habit of saving their searches to Google — and particularly their product searches.

Once you have a Collection set up, Google can point you to other related items, including websites, images and more. Most importantly, this will serve as a new way to get users to perform more product searches, too, as it can send users to other product pages without the user having to type in an explicit search query.

The update also comes with an often-requested collaboration feature, which means you can now share a collection with others for either viewing or editing.

Sharing and related content suggestions are live worldwide.

The AI-powered suggested collections are live in the U.S. for English users starting today and will reach more markets in time.