Build1 publisher2 min readPublished
Google Search Console pools four image-led search entry points into one multimodal filter
Google Search Console is rolling out a multimodal filter that counts searches started from an image, pooling Lens and three other entry points into one view. Teams can now size image-led discovery but cannot yet tell which of the four surfaces sent it.
The Engineer · Build desk
What happened
- The filter groups Google Lens, Circle to Search on Android, image uploads to Google Search, and Chrome's right-click "Search this image" feature.
- Google said the global rollout began on September 24, 2026, with data appearing as image-led traffic reaches each site.
- The filter works in both the standard Search results Performance report and the Generative AI features report.
Compiled by The EngineerSomething wrong?How this is made
Why it matters
- capability Web and ecommerce teams can put a size on image-initiated discovery inside the Performance report they already open, where it used to be spread across other traffic.
- constraint A rise or fall in the multimodal line cannot be pinned on Lens, Circle to Search, uploads or Chrome, so any surface-specific investment needs evidence from somewhere else.
- decision Thin data in the weeks after the rollout start reflects arrival lag as much as demand, so an early cut to visual-asset work would rest on incomplete numbers.
The filter sorts searches by how they started [6]. When a user starts a Google search with visual input, the resulting traffic falls under the multimodal search type [6]. It describes the query side [6]. A product photo that ranks for a typed query stays outside the filter. A Lens search that ends on the product page is inside it [2][6].
Four entry points share one filter value, with no separate filter for each [2][5]. Per the dev.to summary, Google's announcement does not split traffic among the four, establish that every image on a site is eligible, or guarantee more traffic [5][7]. For the pooled number to guide a decision, one of two conditions has to hold. Either the mix of entry points stays roughly stable across the periods a team compares, or the decision does not depend on which surface the user came from. A review of product photography on high-traffic pages can use the pooled count. A case for Android-specific work needs evidence from elsewhere, because Circle to Search, the Android entry point, is mixed in with three others [2][5].
The dev.to post describes image-led discovery before this change as invisible or scattered across reporting [11]. I think fixing that inside the existing Performance reports is the right design. Owners open a Performance report and apply the multimodal search type. This keeps them in the workflow they already use, inside a report they already open [9]. The filter also works in the Generative AI features report [4]. Within that report, a team can separate searches that started with an image from typed ones [1][4].
Google said the global rollout began on September 24, 2026, and that data appears as image-based traffic reaches each site [3]. An empty view in the first days is weak evidence of anything. The dev.to post argues that missing early data is not proof that images never contribute to discovery [8]. It proposes a narrow first pass, in order [10]:
1. Confirming that multimodal data is appearing for the property. 2. Identifying the pages tied to that activity in the Performance report. 3. Comparing the filtered view with broader search performance to judge its relative role. 4. Revisiting the report as the rollout matures and data accumulates.
What to watch
- Whether Google adds a per-entry-point breakdown that separates Lens, Circle to Search, uploads and Chrome image search inside the multimodal filter.
- How large the multimodal share turns out to be for ecommerce and local properties once the global rollout has matured and data has accumulated.
- Whether the multimodal search type becomes available outside the web interface, for teams that pull Search Console data into their own reporting.