Visual search SEO used to have an irritating measurement problem. Someone could point Google Lens at a plant, circle a chair in a photo, or upload an image to find a product, then land on your site. You could see the visit, but you could not cleanly separate that image-led search from the rest of your Google traffic.
Key Takeaways
- Search Console can now separate traffic from Lens, Circle to Search, image uploads, and Chrome’s Search this image feature.
- Impressions and clicks show that image-led search happened. Analytics and customer data show whether the visit led anywhere useful.
- Better reporting does not create a new ranking recipe. Images still need crawlable markup, clear page context, sensible alt text, and a file size that will not drag down the page.
Google changed that on September 24, 2026. Search Console now has a web multimodal search filter, which is a rather clunky name for a useful bit of information: it shows when people found your site by searching with an image.
Search Console finally labels image-led searches
Google’s announcement about web multimodal Search performance says the filter covers Google Lens, Circle to Search on Android, image uploads to Google Search, and Chrome’s right-click Search this image feature.
The data appears in the Performance report for Search results and in Google’s report for generative AI features. Google is rolling it out globally, but a site will only see numbers once it receives reportable traffic from those searches.
I would not panic over an empty report. It may simply mean the site has not received enough multimodal search activity yet. It does not prove that Google cannot find the site’s images, and it certainly does not call for rewriting every alt tag before lunch.
Search Console lets you export the filtered data, too. Once it is out of the dashboard, you can compare image-led search activity with what people did after they arrived on the site.
A click is where Search Console stops
Before this update, a person could photograph an object, circle something on a screen, or start with an image in Chrome. The resulting website visit was folded into broader search reporting. Site owners had no clean way to isolate that behavior.
Now they do. The filter can show which pages earn impressions and clicks when a search begins with an image instead of a typed phrase.
A garden site might discover that plant-identification pages pull in Lens traffic. A furniture seller might find that one product category attracts people who began with a photo. Different sites will find different uses for the filter. It shows which pages already attract image-led searches.
The report still stops at the click. An impression means Google surfaced the content in a reportable multimodal search. A click means the person continued to the site. Neither one tells you whether that person bought something, subscribed, filled out a form, or left ten seconds later.
I discussed the same problem in How Do You Measure AI Search Visibility?. Search Console can show visibility and traffic. Analytics can show what happened on the page. Sales, lead, or customer data must tell you whether the visit was worth anything.
If impressions rise while clicks stay flat, several explanations are possible. The image may be appearing for more searches but failing to earn a visit. The preview may answer the question without a click. The result may be relevant enough to appear but not relevant enough to choose. The report will not settle that argument for you.
Google still has to find the image
The filter gives us better reporting, but the basic technical work has not changed.
Google’s image SEO documentation says crawlers can find an image in the src attribute of an img element, including an img inside a picture element. CSS background images do not work the same way.
Beautiful sites can hide this problem. A designer may treat the most important photograph as part of the page background because it looks good there. If the image carries information people might search for, it should also exist in crawlable markup with a dependable fallback source.
Large media libraries may benefit from an image sitemap, especially when the files live on a content delivery network or are hard to reach through ordinary page links. An image sitemap can help Google discover those files. It cannot rescue a thin page or guarantee a place in image results.
Start with access. If Google cannot reliably reach the image or the page blocks indexing, polishing the filename will not fix the larger problem.
A good image needs a page that explains it
Images do not float free of their pages. Google uses nearby text, headings, captions, the page title, and other signals to work out what an image shows and when it may be useful.
A clear product photo on a vague page forces both the search engine and the reader to guess. The same photo beside a precise product name, useful description, dimensions, and relevant details has much better context.
AI search retrieval depends on the same clear topic ownership and direct answers. Visual search gives people another door into that page. It does not remove the need for words.
Google also accepts signals that identify a preferred image, including primaryImageOfPage, an image property attached to the page’s main entity, and og:image. Those hints can help Google choose a preview image, though Google still makes the final selection.
Alt text is for people first
Alt text helps a person who cannot see the image understand its relevant content or purpose. Google also uses it, along with the page and computer vision, to understand the image.
Good alt text describes what matters in context. A product photo should identify the item and the detail the photo adds. A chart needs a useful summary, with the full data explained in nearby text. A decorative flourish may need an empty alt attribute because it adds no information.
Packing the target keyword into every alt attribute makes the page worse for assistive-technology users. Google warns that repeated keyword stuffing may be treated as spam. The multimodal filter is not an invitation to turn every image description into an SEO slogan.
Filenames offer a small clue. Replacing IMG00023.jpg with a short, descriptive name is sensible when you have the chance. It will not make up for a confusing page, weak image, or missing alt text.
Sharp images should not punish the page
Visual search needs images with enough detail to be useful. A grainy or badly cropped photo gives the searcher little reason to choose it.
The giant original file straight from a camera creates a different problem. It can slow the page, particularly on a phone. Responsive images and current compression formats let a site send an appropriate file size to each device without turning the picture into mush.
There is no single perfect compression setting for every site. A close-up used for plant identification needs more detail than a small author headshot. The practical goal is a clear image that loads without making the visitor wait for it.
Generative search did not add a secret image trick
Google’s guidance for generative AI features says those experiences use the core Search index and ranking systems. Google recommends crawlable content, sound technical structure, useful original material, and relevant images or video. It does not require a special AI schema or a separate writing format.
The multimodal filter gives us a clearer look at image-led search. Google’s familiar ranking foundations still apply.
I would also resist making visual pages for every query you can imagine. An image should do something the prose cannot do as well, such as identify an object or show a useful comparison. If it is only there because a content plan said the page needed a picture, it is probably not doing much for the reader or the search engine.
Capture a baseline before changing everything
Take a snapshot of the current multimodal data before a large image overhaul. Note which pages receive activity and what visitors do after they arrive.
Inspect the pages that already receive activity and look for a problem you can name. The main image may be trapped in a CSS background, stranded far from its explanation, or loading at full camera size on every phone.
Fix the problem you can name, record the change, and compare the next reporting period with the baseline. Search performance moves for many reasons, so one good week does not prove that one edit caused the improvement.
Visual search is no longer buried inside the rest of the Search Console data. At least now we can stop guessing whether the traffic exists and ask the harder question: did it help the business?
FAQ
What does Search Console’s multimodal search filter measure?
It covers web searches made with Lens, Circle to Search on Android, image uploads to Google Search, and Chrome’s right-click Search this image feature.
Why is the multimodal report empty?
Google only shows the metrics after a site receives reportable traffic from multimodal searches. An empty report does not prove that the site’s images are inaccessible or unindexed.
Does visual search SEO require special schema?
No special schema is required for Google’s generative search features. Relevant structured data and image properties can still help Google understand the page and identify a preferred image.