If someone points a phone camera at a product, a plant, or a screenshot and Google sends them to your website, that visit used to be buried in your regular web traffic. You had no way to tell it apart from a typed search. As of late September 2026, you do.
Google Search Console multimodal search reports split web traffic into two groups: searches that started with typed words and searches that started with an image. The new filter covers Google Lens, Circle to Search on Android, images uploaded to Google Search, and Chrome's right-click "Search this image" option. It shows which pages people reach this way, from which countries and devices, and how often they click.
That is the good news. The catch is what it leaves out. There is no query data, so you can see that a page was reached through a camera search but not what the camera was pointed at. The data is not yet in the Search Console API either, so automated dashboards cannot pull it.
This guide explains what Google actually shipped, how to open the report, what the numbers can and cannot tell you, and how to decide whether image-led search deserves your attention.
What Google changed on September 24, 2026
Google announced the change on the Search Central blog on September 24, 2026, under the names of Harsh Kharbanda, product manager lead for Google Lens, and Moshe Samet, product manager lead for Search Console. The post says reporting for web multimodal search now appears in two places: the Performance report for Search results and the newer report for Generative AI features. Rollout began globally the same day, so some properties saw the option before others.
The Search Console update for 2026 does not add a separate report. It adds a choice inside the existing Search type filter. Where you used to pick "Web," you can now pick "Web: text-based" or "Web: multimodal." Google's Performance report help page defines the multimodal option as web search results where an image was used as part of the search. Image, Video, News, Discover, and Google News search types are unchanged.
In the Search results report, the multimodal filter shows clicks, impressions, click-through rate, and position. In the Generative AI features report, which Google introduced in June 2026 and later extended to all sites, it shows impressions from AI Overviews and AI Mode only.
One point of confusion in early coverage is worth settling: this is not Google Images traffic. The Image search type still tracks people browsing the Images tab. Multimodal tracks the input, not the result. A person who took a photo and then clicked a normal blue link lands in the multimodal bucket.
What counts as a multimodal search
"Multimodal" is a technical word for a search that uses more than one kind of input, usually a picture plus some text. For this report, Google's announcement lists four entry points:
|
Entry point |
What the person does |
|---|---|
|
Google Lens |
Opens the camera in the Google app, Chrome, or the Lens icon and searches with a photo |
|
Circle to Search |
On a supported Android phone, holds the home button or gesture bar and circles something on screen |
|
Image upload to Google Search |
Drags or uploads a picture into the search box on desktop or mobile |
|
Chrome "Search this image" |
Right-clicks an image on any web page and picks the Google search option |
All four land in one bucket. Google does not break out Google Lens traffic from Circle to Search analytics or either of those from uploads, and nothing in the help documentation suggests that split is coming. If you need to know which entry point sent a visitor, this report will not tell you.
Google also says the metrics appear only for sites that receive traffic from these searches. An empty multimodal view is not a bug. It usually means people are not yet reaching your site this way, or your property has not received the rollout.
How to open the multimodal report
- Sign in to Search Console and choose your property.
- In the left menu, open Performance and then Search results.
- Click the Search type filter near the top of the report. By default it reads "Web."
- Choose Web: multimodal and apply the filter. If the option is missing, your property either has no multimodal traffic yet or has not received the rollout.
- Set the date range. Several early reports from site owners say data begins around September 10, 2026, though Google has not published a start date, so check your own chart for where the line begins.
- Switch to the Pages tab under the chart. This is the most useful view because the Queries tab is unavailable for this search type.
- Use the Compare option in the Search type filter to put text-based and multimodal side by side for the same pages.
- Click Export to download the table. The announcement points to the Export button as the way to get data out.
For the Generative AI features report, repeat the same filter choice. That report shows impressions only, so treat it as a visibility signal rather than a traffic count.
The two limits that shape everything else
No query data. Google's dimensions help page states that because multimodal searches mostly use images rather than text, specific text query data is not available, and the queries dimension is turned off when this search type is selected. You are left with pages, countries, devices, dates, and search appearance. A product page with 40 multimodal clicks tells you it was reached through an image search. It does not tell you whether the image was your product, a competitor's product, or a photo of something similar.
No API or bulk export yet. As of early October 2026, Google's Search Analytics API reference lists web, image, video, news, discover, and googleNews as the valid search types and has no multimodal value. Site owners who tested the BigQuery bulk export report the same gap. If your reporting runs through Looker Studio or a data warehouse, the multimodal split only exists in the Search Console interface and its manual export for now.
Two smaller open questions remain. Google has not said whether "Web" at the top level now equals text-based plus multimodal combined, or whether historical web totals were restated. If you track week-over-week web clicks, compare a full day's totals across the three views once your property has the option.
The Search Console API reference can change without an announcement, so check it before building any automation.
What it means for SEO
Nothing about your rankings changed on September 24. The report measures traffic that already existed; Google's announcement frames it as insight into how content surfaces when people search with images. The value is in what you decide to do once you can see it.
Multimodal search performance reporting makes a different kind of page visible. Text search rewards pages that answer a typed question. Image search rewards pages that show a recognizable thing clearly: a product, a landmark, a plant, a piece of furniture, a dish, a logo, an error message on a screen. If your site has pages like that, this is the first time you can measure whether they earn their keep.
A hypothetical worked example
This example is invented to show the method, not a real site's data.
Suppose a home-and-garden blog opens the Pages tab with the multimodal filter and sees 1,200 clicks over 28 days. Three pages account for most of it: a guide to identifying common houseplant pests, a page on reading the date codes stamped on propane tanks, and a review of a specific brand of cordless drill. The text-based view shows the pest guide gets 9,000 clicks, the propane page gets 300, and the drill review gets 4,500.
The propane page is the one to study. It earns roughly as many clicks from camera searches as from typed ones, which suggests people are photographing the stamp on a tank and Google is matching it to the page's photos. That page deserves sharper, better-lit images of each stamp format and descriptive alt text, because the images are doing the ranking work. The drill review, by contrast, is a text-first page that happens to pick up some camera traffic, and improving it is a normal SEO task.
A simple decision framework
Use the ratio of multimodal clicks to text-based clicks for each page to sort your pages into three groups:
|
Pattern |
What it suggests |
What to do |
|---|---|---|
|
Multimodal clicks near or above text clicks |
Google is matching your images, not your words |
Improve photo quality, add images of variants, write literal alt text, keep images on the page that owns the topic |
|
Multimodal clicks under 10 percent of text clicks on a visual topic |
Your images may be weak, duplicated from a supplier, or blocked |
Check that images are crawlable, original, and large enough; test with Google's image guidelines |
|
Multimodal clicks on a page with no meaningful images |
Likely a screenshot or right-click search on a diagram or error text |
Leave it; the traffic is a byproduct, not a target |
The thresholds are starting points, not rules. A site selling parts will skew toward camera searches; a news site will barely register.
The practical ceiling is that you cannot see the query, so you cannot write toward it. The fix is to work from the page instead: ask what someone would photograph that led them here, then make sure that thing appears in a clear image on the page.
Common mistakes when reading the new data
- Treating the launch date as a traffic change. The filter started showing data in late September; the traffic was there before. Do not read a line that starts from zero as growth.
- Confusing multimodal with Google Images. They are separate search types. A page can be strong in one and absent from the other.
- Assuming clicks came from your own product photos. Someone may have photographed a competitor's item and Google matched it to your comparison page. The report cannot tell you which.
- Attributing revenue to it. Search Console reports clicks, impressions, CTR, and position. Connecting a camera search to a sale needs your own analytics, and even then the match is approximate.
- Building dashboards on it too early. Until the API or bulk export carries the multimodal type, anything automated will quietly show the old combined web numbers.
- Trimming images to speed up pages. Compression is fine; removing the one clear image of the thing people photograph is not.
If you are also working on local visibility, the image habits above overlap with the advice in our guide to local SEO for small business, where storefront and product photos play a similar role.
The next step
Open the Performance report this week, apply the multimodal filter, and export the Pages table with today's date in the file name. Even if the numbers are small, that export is your baseline. Repeat it monthly. When Google adds the search type to the API, or when a site change moves the numbers, you will have something to compare against instead of a guess.
