Search Console adds multimodal search reporting
Released on , Google has added a multimodal search type to the Search Console Performance and Generative AI reports, separating out traffic from searches that start with an image rather than typed text. Read the Google Search Central announcement here.
Not every search starts with typed words. People point Google Lens at a pair of trainers on the bus, circle a lamp in a social post with Circle to Search, or right-click a product photo in Chrome to find out where it is sold. Each of those is a search, and each can send someone to a website. Until this update there was no dedicated view of that behaviour in Search Console. The new filter gives it one by treating image-led web searches as their own search type.
The timing matters for two reasons. The filter is rolling out globally from 24 September, so UK properties are included from the start. It also lands in two places at once: the standard Performance report for Search results, and the Generative AI performance report, which Google finished rolling out to all websites on 31 August. For any business whose products, premises or work get photographed, the question "how often do people find us by pointing a camera at something?" now has a first-party answer, even if it is a partial one.
How it works
The change sits in the search type filter, which is one of two filters (the other is date) that are always present at the top of the Performance report and cannot be removed. Web search is now split into two options. According to the Performance report documentation, "Web: text-based" covers traditional traffic from queries typed into the standard Google search bar, while "Web: multimodal" covers web search results where an image was used as part of the search. Google's announcement names four sources that feed the multimodal figure: Google Lens, Circle to Search on Android, images uploaded directly to Google Search, and the "Search this image" option in Chrome's right-click menu.
It is worth separating this from the existing Image search type, because the two are easy to confuse. The Search Console API reference defines the image type as results from the Image tab in Google Search, which describes where a result appeared. Multimodal describes how the search started. A shopper who photographs a coat and clicks through to your product page from the resulting web results is a multimodal search, even though nobody visited Google Images.
The biggest structural difference is the missing query data. Because multimodal searches mostly use images rather than text, Google states in its dimensions documentation that specific text query data is not available for this traffic, and the Queries dimension is unavailable when the multimodal search type is selected. That is the only dimension Google documents as missing, so pages, countries, devices and dates remain available for slicing the data.
In the Generative AI performance report, the same text-based and multimodal split applies to impressions from AI Overviews and AI Mode. That report counts impressions only, meaning how many times links to your site were shown in a generative AI feature, and it draws on the Web search type of the main Performance report. With the multimodal filter applied, it shows how often your pages were linked from an AI Overview or AI Mode answer to a search that began with an image. Both reports keep their Export button, so the filtered view can be downloaded for analysis elsewhere.
Examples and use cases
A UK footwear retailer is the clearest case. Someone photographs a pair of trainers with Lens and taps through to a stockist. Filter the Performance report to Web: multimodal, open the Pages tab, and you can see which product pages collect that traffic. With no query data, the landing page becomes the best available evidence of what was photographed. If a discontinued colourway keeps drawing image-led clicks, that says something about demand as well as about search.
Homeware and interiors brands are likely to see a different pattern. Their products often appear in other people's pictures: a lamp in a styled room shot, a chair in a social post. Circle to Search on Android turns those moments into searches. The report will not tell you which post or photo started the search, but a product page that performs noticeably better in multimodal than in text-based search is being found by sight rather than by name. That usually means shoppers do not know what the item is called, which is useful input for product titles and category naming.
Informational content can benefit too. A garden centre's plant care guides, a restaurant's pages about its dishes, or a venue's pages about its building can all surface when someone photographs the real thing. For this kind of content it is worth checking the Generative AI report as well, to see whether image-led searches are surfacing your pages inside AI Overviews or AI Mode rather than only in standard results.
For agencies, comparison mode is the practical tool. Search Console lets you compare search types, limited to two at a time, so setting Web: text-based against Web: multimodal for the same date range shows how much of a client's web visibility starts with a camera. That is a single line that can sit in a monthly report without any extra tooling.
What this means for UK site owners
For UK retailers, the multimodal figure is a direct read on whether product imagery is earning search visibility, rather than something inferred from a vague rise in product page traffic. Use the Countries dimension to isolate searches from the United Kingdom, since the rollout is global and a brand that ships abroad will see image-led searches from other markets mixed in.
What to act on is less new than the report itself. Google's image SEO documentation is the closest thing to official guidance on making images understandable. It asks for images embedded with HTML image elements, because Google does not index CSS background images. It recommends placing images near relevant text on pages that match the image's subject, using short descriptive filenames and alt text, and submitting an image sitemap for images Google might not otherwise discover. It also favours sharp, high-quality photos while warning that images are often the largest contributor to page weight. None of this is a Lens ranking formula, and the report does not reveal how Lens matches an image to a page. It is the hygiene that gives an image-led search something to match against.
If you already track AI visibility through the Generative AI report, which we covered when Search Console first added generative AI performance reports in June, add the multimodal filter to the same routine rather than treating it as a separate project. For catalogue-heavy sites, reading these numbers against product imagery and page structure is part of ecommerce SEO work, and the same approach applies to any site where visitors are likely to photograph what they are looking for.
Service businesses should set expectations accordingly. An accountant or a solicitor is unlikely to be found by someone pointing a camera at something, and a near-empty multimodal view for that kind of site is not a problem to solve. The report is most useful where the product or place is visual.
Practical considerations
The missing Queries dimension is the main limitation. You will know which page an image-led search reached, but not what was photographed, so any conclusion about intent is an inference from the landing page. Google also says you will only see these metrics if your site is receiving traffic from these searches, and the rollout only began on 24 September, so an empty view in the first days may simply mean the data has not arrived yet. Comparing more than two search types at once is not possible, which limits how much a single view can show.
Two things Google has not said are worth keeping in mind. The announcement does not state how far back the multimodal data runs, or whether this traffic was previously counted inside the Web totals. Check the date range available on your own property before building any trend claim across the launch date, and treat the first few weeks as a baseline. Separately, as of 25 September the Search Console API reference lists discover, googleNews, news, image, video and web as its type values, with no multimodal option. Any dashboard that pulls Search Console data through the API cannot split this traffic out yet, so use the Export button in the interface until the reference changes.
FAQ
What counts as a multimodal search in Search Console?
Google defines it as web search results where an image was used as part of the search. The announcement lists four sources: Google Lens, Circle to Search on Android, images uploaded directly to Google Search, and the "Search this image" option in Chrome's right-click menu. Searches typed into the standard search bar fall under Web: text-based instead.
Where do I find the multimodal data?
Open the Performance report for Search results, click the search type filter at the top of the report and choose Web: multimodal. The same filter is available in the Generative AI performance report, where it applies to impressions from AI Overviews and AI Mode. Both reports can be exported with the Export button.
Why can't I see search queries for multimodal traffic?
Because multimodal searches mostly use images rather than text, Google says specific text query data is not available for this traffic, so the Queries dimension is unavailable when the multimodal search type is selected. The Pages dimension is the best substitute: the landing page is the clearest evidence of what people were searching for.
Is multimodal the same as the Image search type?
No. The Image search type covers results shown in the Image tab of Google Search, so it describes where your result appeared. Multimodal describes how the search started, with an image as the input, and it covers web results. Someone who photographs a product with Lens and clicks through to your product page counts as multimodal, not Image.
Why is the multimodal filter showing no data for my site?
There are two likely reasons. The feature is rolling out globally from 24 September 2026, so it may not have reached every property yet. Google also says the metrics only appear if your site is receiving traffic from these searches, which many service businesses will not. An empty view on a site whose products or premises are rarely photographed is normal.
Can I pull multimodal data through the Search Console API?
Not as documented at the time of writing. On 25 September 2026 the Search Analytics API reference listed discover, googleNews, news, image, video and web as its type values, with no multimodal option. Until that changes, dashboards built on the API cannot separate this traffic, so export it from the Search Console interface instead.
