Search Console now counts searches made with a camera. It cannot tell you what was photographed.

For the first time, Search Console reports the searches people make by pointing a camera at something. What it leaves out matters as much as what it adds: there is no query.

On 24 September, Google announced web multimodal reporting in two places: the Performance report for Search results, and the report for Generative AI features. Google says the data covers four routes: Lens, Circle to Search on Android, image uploads to Google Search, and Chrome's right-click "Search this image". The rollout is global and started that day, so some properties will see the filter before others.

The Performance report help page now splits Web in two. "Text-based" is traffic from text typed into the standard search bar. "Multimodal" is traffic from searches made with images.

The query column is gone

This is the part that changes how you read the report. Google's dimensions documentation says it plainly: "Because multimodal searches mostly use images rather than text, specific text query data isn't available for this traffic. As a result, the queries dimension isn't available when this search type is selected."

Most Search Console habits start from the query: which words found you, and which words should have found you. With multimodal traffic that starting point does not exist. You get pages, countries, devices and dates. You do not get what the person was looking at.

So for this traffic, the page is the query. If a product page appears for multimodal searches, someone photographed something that Google decided your page answers. The page list is the only record of intent you have.

There is no before

Google's John Mueller said on Bluesky, as PPC Land reported, that "the data wasn't previously in the counts, the report has new data." Taking that at its word, there is no baseline to compare against. A line that climbs through October may only be the rollout reaching your property. Treat the first weeks as a census, not a trend.

Google's post also names exactly four routes. It does not say where Search Live sessions or images asked inside AI Mode are counted. Until it does, assume the multimodal figure is a floor.

How to read it this week

1. In the Performance report, set the search type to Web: multimodal. If the filter is missing, either the rollout has not reached your property or Google has no multimodal traffic to report for it yet.
2. Switch to the Pages tab and export it. That list is your multimodal query list.
3. For each page, write down what someone would have had to photograph to land there. Pages you expected and do not see are as informative as the ones you do.
4. Apply the same filter in the Generative AI report. That is where image-led searches meet AI answers.
5. Before comparing any Web figure across 24 September, check which search type you are looking at. If the multimodal data is new, a date range that spans the rollout is not like for like.

Google has not published how an image gets matched to a page, and we would not trust anyone who claims to know the weighting. The report gives you the outcome, not the mechanism. When the input is a picture, a page has to make clear what it is about to anything reading it: what the thing is, what it is called, and what it does. That is the same discipline behind an AI-readable website, arriving through a different door.

The camera has always been a search box. Now it shows up in the report, and the page is where you read what people were asking.