Google Search Console now reports traffic from image-based searches. Select Performance, then Search type, then "Web: multimodal", and you'll see clicks, impressions, CTR and position for web results reached through Google Lens, Circle to Search, image uploads and Chrome's "Search this image", by page, country and device, but with no query data. It rolled out globally on September 24, 2026. It isn't backfilled, and the API doesn't expose it yet. For stores, product catalogues and any site with real photos of real things, this is the first honest look at a traffic source that's been invisible for years.
What does the Search Console multimodal filter actually count?
Google's announcement on the Search Central blog lists four entry points that feed the new search type:
- Google Lens, from the Google app or the camera on Android and iOS.
- Circle to Search on Android, where someone circles something on their screen.
- Image uploads to Google Search.
- Chrome's right-click "Search this image".
All four are lumped together under one label. You can't split Lens from Circle to Search, which is a shame, because they're different behaviours. Lens is usually someone pointing a phone at a physical object. Circle to Search is usually someone who saw something in a video or on social and wants to know what it is.
The filter sits in the same Search type dropdown as Image, Video and News, and it comes with a sibling called "Web: text-based", which is the classic typed query in the search bar. Google also added the same multimodal split to the Generative AI features report, so you can see image-started searches that surfaced your page in AI experiences separately.
The part that matters most: this is new data, not a relabelled slice of old data. John Mueller confirmed the multimodal traffic "wasn't previously in the counts." If total impressions on a product-heavy site jumped after September 24, check the filter before you celebrate or report it as SEO growth.
Why is there no query data?
Because nobody typed one. The search was a photo of a lamp, a screenshot of a pair of shoes, or a circled logo on a TikTok. Google does turn images into internal representations to find matches, but it doesn't hand those back to you as text queries.
So the multimodal report has four usable dimensions: pages, countries, devices and dates. The Queries tab is empty for this traffic, and it will stay that way.
That changes how you work with it. With text search you start from the query and ask "which page should win this?" With multimodal you start from the page and ask "what in this page's images is Google matching, and is the page good enough to convert the person who clicked?"
How do you read the report without queries?
Here's the routine we've set up for client properties. Once it's running it takes about fifteen minutes a month per site.
Start from the Pages tab
Set the search type to Web: multimodal, set the date range to begin on or after September 24, 2026, and open Pages. Sort by impressions. What you're looking at is a list of URLs Google considers a good answer to someone's photo.
On an e-commerce site that's usually product pages. On a service business site it can be surprising. A roofing contractor might find a blog post about a specific shingle profile getting matched, because homeowners photograph their roofs. A landscaping company might see a plant identification page. Those pages are telling you what people point their cameras at in your market.
Compare each page against text-based search
Switch to Web: text-based for the same dates and put the two views side by side. Three patterns show up:
- High multimodal, low text-based impressions. The images are ranking and the copy isn't. This page's photos are doing more work than its words, so improve the copy and the title.
- High text-based, near-zero multimodal. The page ranks for words, but its images don't get matched. Often these are stock photos or generic lifestyle shots that look like a thousand other pages.
- Multimodal impressions with a much lower CTR than text-based. The page gets shown for a photo but the result doesn't look like the thing the person photographed. Usually a mismatch between the image and the title.
Split by device
Open the Devices tab while the multimodal filter is on. Lens and Circle to Search are mobile-first behaviours, so expect mobile to dominate. If a page with meaningful multimodal clicks loads slowly or has a broken mobile layout, you're paying for that traffic in bounces. Fix that before you touch any images.
Can you get multimodal data through the API or Looker Studio?
Not yet. At launch the Search Analytics API reference doesn't list a multimodal search type, and Looker Studio's Search Console connector sits on that API. If your client reporting runs off a dashboard, the dashboard won't show this traffic.
Until Google adds API support, the only way to keep history is to click Export in the Performance report. Search Console keeps 16 months of data, so nothing is lost this month, but people forget. We save a monthly multimodal export per client into the same folder as the rest of the reporting data, and we flag it separately in reports so nobody mistakes it for text-search growth.
When API support lands, we'll move it into the automated pipeline with everything else. If you're building dashboards clients actually open, our post on marketing dashboards clients read covers what we put on page one.
Why doesn't GA4 show Google Lens traffic separately?
Because a click from a Lens result lands on your site the same way a click from a typed search does. In GA4 it shows up as google / organic. There's no referrer flag, no special parameter, nothing to filter on.
The workaround is page-level. Take the pages with multimodal clicks from Search Console, then look at those same landing pages in GA4 for the same dates:
- Organic sessions and engaged sessions.
- Conversion rate compared with the site's organic average.
- Add-to-cart or form starts, if you track them.
It's not clean attribution, and it isn't meant to be. It answers a simpler question: do people who arrive from a photo convert on this page, or bounce? If you're already separating AI assistant traffic in GA4, the setup in our guide to tracking AI assistant traffic in GA4 shows how to build custom channel groups. Multimodal traffic just won't fit into one yet, because there's nothing in the hit that marks it.
What should you change to get more multimodal traffic?
Lens works by matching what's in the photo to images and pages Google has indexed. That tells you most of what to do.
Use original photos of the actual thing
Stock photography is the biggest reason good pages get zero multimodal impressions. If your product shot is the manufacturer's image, the same file is on every reseller's site, and Google has no reason to pick yours. An original photo, taken from a real angle in real light, is the single biggest lever you have. For service businesses that means photos of your actual work, your trucks, your installs, not a smiling model in a hard hat.
Put the image next to words that name it
Google reads the text around an image to understand it. A photo of a pendant lamp in a gallery with no caption gives it very little to work with. The same photo beside a heading that names the lamp's style, finish and size is much easier to match. Google's image SEO best practices cover this: descriptive filenames, useful alt text and relevant nearby text.
Write alt text for the thing, not for the page
Bad alt text: "best pendant lights 2026". Good alt text: "brushed steel dome pendant light, 12-inch shade, hanging over a kitchen island". The second one describes what's in the photo, which is the only thing a visual search cares about.
Mark up product pages
Product structured data with an image property, a real name and a price connects the photo to a purchasable item. If you sell physical products and your pages don't have it, that's the first technical fix. Our technical SEO audit checklist covers the rest of what we check.
Make sure the images are crawlable
It sounds obvious, but we still find images blocked in robots.txt, served only through JavaScript that never renders, or lazy-loaded in a way Googlebot never triggers. Run the URL Inspection tool on a key product page and check that the rendered HTML includes your image tags with real src values.
Which businesses should care about this?
Ask yourself one question: do your customers ever photograph the thing you sell or fix?
- Retail and e-commerce: yes, constantly. Furniture, lighting, fashion, auto parts, tools, plants. This report is built for you.
- Home services: more than you'd think. People photograph a broken part, a pest, a water stain, a plant, a roof. A page that identifies the problem and explains the fix can win that click.
- Restaurants and local retail: partly. Someone pointing Lens at a storefront often gets a Maps result rather than a web result, and that isn't in this report. Your Google Business Profile photos still matter more there. See our local SEO guide for Dallas businesses.
- B2B software and pure services: probably not much. If the multimodal filter shows nothing after a month, that's a fine answer. Don't invent work.
Does this connect to AI search visibility?
Yes, and that's why we're watching it closely. Circle to Search and Lens increasingly hand people an AI Overview rather than ten blue links, and Google put the multimodal split into the Generative AI features report on day one. Visual search and AI answers are converging, and the pages that get cited are the ones with clear, specific, original content, in the photos as much as in the words.
If you've been working on getting cited in AI answers, the same discipline applies. Our breakdown of the AI search citation gap explains why generic content gets skipped. Generic images get skipped for the same reason.
What we're telling clients this week
Don't rebuild anything yet. The filter is five days old, and a week of data is noise. Here's the order we're working in:
- Check the filter on every property now so you know whether there's any multimodal traffic at all.
- Add a note to reporting: any lift in total impressions after September 24 may be new multimodal data, not growth.
- Start the monthly export.
- After 30 days, list pages with multimodal impressions and a CTR well below their text-based CTR. Fix those first.
- Replace stock images on your top ten product or service pages with originals.
The sites that win this aren't the ones with the cleverest SEO. They're the ones whose photos actually show the product.
Get a Free Automation Audit
If you want someone to go through your Search Console, GA4 and reporting setup and tell you where the gaps are, including whether multimodal traffic is hiding in your numbers, we'll do it free. We'll also show you which parts of your monthly reporting can run on their own instead of being rebuilt by hand. Get a Free Automation Audit and we'll come back with a written list of fixes, ranked by impact.