Build1 publisher2 min readPublished
Google Search Console folds Lens and Circle to Search visits into one Multimodal filter
Google Search Console's new Multimodal filter, rolling out since September 24, 2026, counts four image-search entry points as one traffic channel. Site owners can now count visual-search visits directly, but proving that image work pays off still takes a page-level test.
The Engineer · Build desk
Drafted by a language model from the sources cited here and checked against its claim ledger before publication. How we use AISend a correction
What happened
- The filter groups four entry points: Google Lens, Android's Circle to Search, images uploaded to Google Search, and the image search Chrome offers from its right-click menu.
- It sits under the Web Search type in Search Console's Performance reports and separates visual-path activity from the broader web-search data.
- According to Google, the multimodal data appears in both the Search performance report and the Generative AI features report.
- The rollout reaches properties gradually, and the write-up advises site owners to check each of their own properties for the filter.
Compiled by The EngineerSomething wrong?How this is made
Why it matters
- capability Teams can check in Search Console's own data whether image-based search sends them any visits before they commit budget to image optimization.
- constraint A product team that cares about one surface, such as Circle to Search on Android, cannot credit a traffic change to that surface from this report alone.
- decision Image work can now be ranked page by page, starting with the pages that already appear in the filtered data and their content formats.
I like where Google put it. Any comparison runs against the report a team already reads, for the same period, with the filter switched on and then off [5][9].
The grouping is coarser than the feature list suggests. All four entry points sit behind one filter. The dev.to write-up describes it as a way to assess "the combined multimodal channel" without evaluating each interaction separately [8]. The write-up does not describe a per-surface split, and it does not say whether data from before the September 24, 2026 rollout is available for comparison [4].
The count also needs careful reading. It measures visits that already come in through image-based search [1]. A page that never shows up in the filtered data is also a result, if an unwelcome one for whoever bought the stock photos. The write-up says the reporting "does not turn those interactions into a separate optimization channel with guaranteed outcomes" [7]. A site can do well in typed search and draw little multimodal traffic. Pages with useful visual material can show the reverse [10].
I treat a page's multimodal number the way I treat a vendor benchmark. It describes one site's traffic over one period. Before it can justify image work on other pages, filtered traffic has to rise on pages where the images changed and stay flat on comparable pages where they did not. Google says the data can be exported to external tools [6]. The write-up suggests joining it with internal content, conversion or editorial data [9].
I'd run the test in this order:
1. Export the filtered view and the unfiltered Web Search view for the same period [9]. 2. List the pages that appear in the filtered data, along with their content formats [9]. 3. On half of those pages, replace images that misrepresent the page and update alt text and image structured data [9]. Leave the other half alone. 4. Re-export after several weeks. Compare each group's filtered traffic against its own unfiltered baseline.
The held-back half matters because the filtered count moves when any of the four entry points moves [8]. Without it, a rise in Lens use for unrelated reasons would look like the payoff from new alt text.
What to watch
- Whether Google adds a per-surface breakdown inside the Multimodal filter, so Lens, Circle to Search, uploads and Chrome image search can be measured separately.
- Whether the Multimodal filter becomes available through the Search Console API, so the page-level comparison can be automated without manual exports.
- Whether multimodal data appears for dates before the September 24, 2026 rollout; without it, any before-and-after test starts from the rollout date.