Search Console Now Counts Image Searches: How to Read the Multimodal Filter

Someone points a phone at a pair of trainers on a train, or circles a lamp in a friend's Instagram story, and Google shows them where to buy it. Until last week, a site that won that visit had no clean way to see it in Search Console. From 24 September 2026 there is one: a filter that separates searches made with an image from searches typed as text. It is useful, it has a large blind spot, and its first weeks overlap with a Google spam update.

What Google added

Google announced the change on its Search Central blog on 24 September, in a post by Harsh Kharbanda, Product Manager Lead for Google Lens, and Moshe Samet, Product Manager Lead for Search Console. The rollout is global and started that day. Google wrote: "Today, we are excited to introduce Search Console reporting for web multimodal search, in the report for performance on Search results as well as in the report for Generative AI features."

The data covers four ways of searching with a picture: Google Lens, Circle to Search on Android, image uploads to Google Search, and Chrome's right-click "Search this image". In the Performance report, the Web search type now splits into two options. Google's help page defines "Web: multimodal" as results "where an image was used as part of the search", and "Web: text-based" as traffic from text typed into the standard search bar. Asked on Bluesky whether this traffic used to sit inside the Web totals, Google's John Mueller checked with the team and replied that "the data wasn't previously in the counts, the report has new data." So this is traffic you have not seen in the report before, not a relabelling of old numbers.

What the report shows, and what it does not

The missing query is the big one. Google's help page on Performance report dimensions explains: "Because multimodal searches mostly use images rather than text, specific text query data isn't available for this traffic." The queries dimension is not available when you select the multimodal type. So you learn where image searchers landed, not what they photographed. Your analysis has to happen at page level: which product, listing or menu page is being matched, in which country, on which device.

The history question is only partly settled. Mueller's answer means earlier Web figures never contained this traffic, but we found no Google statement on how far back the new multimodal data goes. Until there is one, do not compare a multimodal line against a period before the filter existed.

How to read it this week

1. Open Performance, then Search results, and set the search type to Web: multimodal. Check the Generative AI features report the same way.
2. Go to the Pages tab and export it. Sort by impressions, then by clicks, and look for pages with many impressions and few clicks.
3. Compare the same pages under Web: text-based. A page that does well in image searches and poorly in typed ones tells you the picture is doing work the copy is not.
4. Split by country. For businesses serving Cyprus, the UK, Greece and the Gulf, this shows where visual discovery is happening rather than assuming it matches your text traffic.
5. Split by device. Expect mobile to dominate, since most of these entry points are phone-first, but confirm it for your own site before planning around it.

Who should look first

Any business whose customers see the product before they know its name. Retail and fashion are the obvious cases: people photograph things they want and do not know what to call. Hospitality is worth checking, because people search from a photo of a dish, a hotel pool or a view. In real estate, a building photographed from the street, or a screenshot of a listing, is a plausible starting point for a search, and the page report will tell you whether it is happening for your listings.

Some businesses will find very little here, and that is a result too. A professional services firm with stock photography should not expect Lens traffic.

What to change in image SEO

Without queries, the levers are the image and the page around it. Google's image SEO guidance is the right starting point, and parts of it matter more now that this traffic can be measured.

Original imagery

If your product photo is the manufacturer's image, the same picture sits on every reseller's site, and nothing in it ties a match to your page rather than theirs. Your own photography gives Google something only you publish. Google also says high-quality photos appeal to users more than blurry, unclear ones. For multi-location businesses, photograph the real premises and rooms rather than reusing brand-library shots.

Product images and structured data

Google's product structured data documentation says that with it, product information "can appear in richer ways in Google Search results (including Google Images and Google Lens)." Check that every product page marks up its image, price and availability, and that the image URL in the markup is the same image shown on the page. Where a gallery exists, make sure the images are standard img elements Google can crawl, and use an image sitemap for images Google might not otherwise find.

Alt text, filenames and placement

Google asks for alt text that is useful, information-rich and in context of the page, and warns against keyword stuffing. Describe what is in the picture: colour, material, model, setting. Google says it takes information about an image's subject from the page around it, including captions, so place images near the text that explains them. Filenames give Google only very light clues, but a descriptive name still beats a camera default. Good alt text also helps people using screen readers, which is reason enough on its own.

Keep this separate from the AI Overviews work. If you are already tracking Search Console's AI Overviews report, the multimodal type in the Generative AI features report is a new slice to add to that baseline, not a replacement for it.