How Google Discovery Ads Reach People Across Feeds and Surfaces

Google Discovery Ads are a paid advertising format that appears in the discovery feeds of Google properties, such as the YouTube home feed, the Gmail promotions and social tabs, and the Discover feed in the Google app. Unlike search ads, which respond to a specific query, these ads reach people who are browsing and may not be looking for anything in particular. For readers who want to study this format in a structured way, the Micro Diploma: Google Discovery Ads offers a path to understanding how these campaigns are built and measured. The format matters because it lets advertisers reach people earlier in their decision process, when interest is forming rather than when a purchase is already planned.

A Discovery campaign is built from a set of assets rather than a single fixed ad. An advertiser supplies images, short videos, headlines, descriptions and a business name, and the system assembles combinations and tests them against audiences. Targeting relies on signals such as past viewing behavior, search activity across Google, demographics where permitted, and similarity to people who have already engaged with the advertiser. The system then decides which asset combinations to show and to whom, based on the goal the advertiser selects, such as conversions or a target cost per action. Because the machine learning component adjusts over time, the quality and variety of the input assets strongly affect performance.

Measurement follows the campaign goal. Conversion tracking must be set up correctly for these numbers to mean anything, and attribution windows shape how credit is assigned between ads and other channels.

Consider a team that works with data and reports for a mid-sized retail business that sells home goods online. The team is asked to evaluate whether a Discovery campaign adds value beyond the search and shopping campaigns already running. They begin by reviewing the conversion tracking to confirm that purchases, add-to-cart events and newsletter sign-ups are recorded consistently. They then set up a campaign with several image sets, two short video cuts and a range of headlines, and they label each asset so they can read performance by creative element later.

After several weeks, the team compares Discovery results with the search baseline. They find that the Discovery campaign produces many impressions and a moderate number of conversions at a lower cost per click, but it also overlaps with existing search traffic from people who saw an ad earlier and searched later. To handle this, they build a simple report that separates view-assisted conversions from direct ones and shares it with the marketing lead, who uses it to set a realistic budget split rather than judging the channel on last-click results alone.

Discovery campaigns have real limits. Attribution is also contested. Ads seen in a feed may influence a purchase that is credited to another channel, so reported numbers can overstate or understate the format's contribution depending on the model used. Privacy rules and changes to tracking signals further complicate audience building and measurement.

The main takeaway for data and reporting teams is that Discovery should be evaluated with a clear test design, consistent conversion tracking and transparent assumptions about attribution. Future development will likely depend on how well advertisers can interpret automated results and on how platforms share the signals needed for honest measurement.