Search Console Multimodal Search Type: What It Means (2026)

Search Console now has a Web: multimodal search type. It counts the people who found you by pointing a camera at something, through Google Lens, Circle to Search, uploaded images and Chrome’s right-click image search. Google began rolling it out globally on 24 September 2026. The catch most owners miss: this is brand new data that was never in your counts before, so a late-September lift in your totals is probably reporting, not growth.

By Cody Wise, founder of Wise Media, Calgary. Published 6 October 2026. Every factual claim below links to a Google primary source.

Summary

What is new. A Web: multimodal filter in the Search Console Performance report and in the Generative AI features report, announced 24 September 2026.
What it counts. Searches made with images: Google Lens, Circle to Search on Android, images uploaded to Google Search, and Chrome’s right-click “Search this image”.
What it is not. It is not the Image tab. Image search and multimodal search are two separate search types in the same filter.
The trap. Google’s John Mueller confirmed this data was not previously in the counts. If your clicks or impressions stepped up in late September, check whether the step is multimodal before you call it growth.
The limitation. There is no query data. Google’s position is that because multimodal searches mostly use images rather than text, text query data is not available for that traffic.
Who should care most. Businesses whose product or premises is physically visible: trades and contractors, retail and showrooms, restaurants, realtors, short-term rental operators, and anyone selling a physical object.

A founder at a home office desk reviewing Google Search Console performance reporting on a laptop — The first job is not optimisation. It is working out whether the number that just moved in your Performance report moved because your site changed or because Google’s reporting did.

What is the multimodal search type in Search Console?
Why did my Search Console numbers change in late September 2026?
Which businesses actually get camera search traffic?
What the multimodal report will not tell you
How do you optimise for multimodal search?
Common mistakes and a 30 minute baseline audit
FAQ

What is the multimodal search type in Search Console?

Multimodal is a search type filter in the Search Console Performance report that isolates traffic from searches made with images instead of typed text. Google’s documentation defines it as traffic “when users search using images (such as with a smartphone camera)”, and names four sources: Google Lens, Circle to Search on Android, uploaded images, and Chrome image search. It sits beside Web: text-based under the same Web heading.

The announcement came from the Search Central blog on 24 September 2026, posted by Harsh Kharbanda, Product Manager Lead for Google Lens, and Moshe Samet, Product Manager Lead for Search Console. Google described the rollout as global, “starting today”, and added a condition worth reading twice: “You will start seeing these metrics in your Performance report if your site is receiving traffic from these queries.” No camera traffic, no filter. An empty or missing multimodal view is a finding, not a bug.

The four entry points, and what each one actually is

Entry point	What the user physically does	Where it happens
Google Lens	Points a phone camera at an object, sign, label, plant, part or storefront, or runs Lens over a screenshot	Google app, Chrome, Android, iOS
Circle to Search	Long-presses the home button or nav bar, then circles something on screen	Supported Android devices
Image upload to Google Search	Uploads a photo or screenshot from their device into Search	Desktop and mobile web
Chrome “Search this image”	Right-clicks an image on any page and searches it	Chrome desktop

Notice what these have in common. In every case the user starts from something they can see rather than something they can name. That is the whole strategic point. Camera search captures the demand that never becomes a keyword, because the person did not know the word for the thing they were looking at.

Multimodal is not the Image tab, and this matters

This is the single most common error in the coverage so far. Google’s Performance report help page lists the search types explicitly, and image search and multimodal search are two different rows:

Search type	Google’s definition
`Web: text-based`	“Includes search traffic originating from text queries entered into the standard Google search bar.”
`Web: multimodal`	“Includes search traffic when users search using images (such as with a smartphone camera). This includes searches originating from Google Lens, Circle to Search on Android, uploaded images, and Chrome image search.”
`Image`	“Includes search traffic from results shown in the Google Images tab.”
`Video`	“Includes search traffic from results shown in the Google Videos tab.”
`News`	“Includes search traffic from results shown in the Google News tab and the news search results view.”

Image search is a destination: someone goes to the Images tab and browses results. Multimodal is an input method: someone uses a picture as the query and gets ordinary web results back. A page can earn impressions in both, neither, or only one. If you have spent years looking at your Image numbers and concluded that visual search does not matter to your business, that conclusion was built on the wrong data set.

Why did my Search Console numbers change in late September 2026?

Because Google added data that was not there before. Asked directly whether multimodal traffic had previously been folded into existing counts, Google’s John Mueller said the data was not previously in the counts and that the report contains new data. A step change in your totals dated on or near 24 September 2026 is therefore a reporting change first and a performance change only if you can prove it separately.

This is the same class of error that burns agencies and in-house marketers every time Google widens a report. The number on the dashboard went up. Nothing about the website got better. Anyone who reports that lift as a result has just set a baseline they cannot repeat, and has taught a client to expect a trend that does not exist.

How to check whether your lift is real, in four steps

Filter to Web: text-based only and pull the last 90 days. This is the closest like-for-like comparison with how your reporting looked before 24 September.
Filter to Web: multimodal only for the same window. Note the first date with non-zero data. That date is your reporting boundary.
Compare the two series across that boundary. If text-based is flat and the total rose, the rise is the new data set. If text-based also rose, you have a real change to investigate.
Re-baseline your own reporting. Any month-over-month or year-over-year comparison that crosses 24 September 2026 is now comparing two different measurement definitions unless you segment by search type.

The discipline here is the same one we wrote up when Google started reporting social and video platform performance in Search Console. Traffic, revenue, analytics and search visibility are four different signals. The moment you read them as one number, you lose the ability to diagnose anything.

What to put in the client email

Short version, and you can copy it: “Google added a new search type to Search Console on 24 September that counts searches made with a phone camera through Lens and Circle to Search. That data was never in our numbers before, so part of the September increase is new measurement rather than new traffic. I have split the two so the trend line stays honest. Here is what each one did.”

Sending that email is worth more than any optimisation in this article. It is the difference between a measurement partner and a dashboard reader.

Which businesses actually get camera search traffic?

Camera search follows physical visibility. If a customer can stand in front of your product, your premises, your signage, a part you sell or a problem you fix, and photograph it, you are in the game. If your service exists only as a concept, you are mostly not. Here is the honest read by category.

Business type	Likely camera search exposure	The query behind the photo
Trades and contractors	High	A photo of a cracked foundation, a failing shingle, a leaking valve, an unfamiliar breaker panel
Retail and showrooms	High	A photo of a product, a tag, a fixture or a competitor’s shelf
Restaurants and cafes	Medium to high	A photo of a storefront, a menu board or a dish
Real estate	Medium	A photo of a building, a for-sale sign or a streetscape
Short-term rental operators	Medium	A photo of a property exterior, a hot tub model, an appliance or a lock
Manufacturers and parts suppliers	High	A photo of an unlabelled component the buyer cannot name
Professional services, SaaS, consulting	Low	Very little. Expect the filter to stay near zero and do not force it

A pedestrian walking past a row of independent storefronts on a Canadian main street, the kind of business a passer-by might photograph with Google Lens — Camera search is a street-level behaviour. A passer-by who photographs a storefront or a shop window is now a measurable impression in your Performance report.

One Canadian nuance worth stating plainly. Google’s announcement calls the rollout global, so there is no “is this live in Canada yet” question to answer. It is. What differs by market is the base rate of Lens use, and Google has not published a country-level breakdown. The only honest statement available is that Google reported Lens being used for nearly 20 billion visual searches every month as of October 2024, and that Circle to Search was then available on more than 150 million Android devices. Treat those as the last figures Google published rather than today’s number, and do not let anyone sell you a Canadian percentage that has no source behind it.

What the multimodal report will not tell you

There is no query data. Google’s explanation is that because multimodal searches mostly use images rather than text, specific text query data is not available for that traffic. You get impressions, clicks, CTR, average position, and the page, country, device and date dimensions. You do not get to see what the picture was.

That makes multimodal a page-level signal, not a keyword signal. The workflow changes accordingly: you are looking at which URLs earn camera impressions, and reasoning backwards about what someone was photographing to land there.

Honest list of what is still unknown

Whether AI Mode and Search Live image prompts are included. Google has not said. The announcement names four sources and stops there.
How far back the data goes. Google has not published a start date for the historical series beyond the 24 September rollout.
How the top-level Web total composes the two subtypes. Verify it in your own property rather than assuming, because the arithmetic drives every report you send.
Country-level or vertical-level benchmarks. None exist from Google. Anyone quoting one is estimating.

Saying “Google has not said” out loud is not weakness. On a measurement topic it is the whole value. The pattern we flagged when Google shipped the Generative AI performance reporting that changed how local visibility gets measured is repeating here: the report arrives before the documentation does, and the people who guess loudest get quoted and then quietly corrected.

How do you optimise for multimodal search?

You cannot optimise for a query you cannot see. What you can do is make sure Google can find, index and confidently interpret every image on your site, because an image Google never indexed cannot be matched to anyone’s camera. Every step below comes straight from Google’s own image SEO documentation.

Serve images in real <img> elements. Google finds images in the src attribute of an img element and does not index CSS background images. Decorative CSS backgrounds are fine. A product shot or a project photo loaded as a CSS background is invisible.
Submit an image sitemap. Google recommends it specifically for images it might not otherwise discover, which is exactly the case for CDN-hosted or lazily loaded media.
Write alt text in layers. Google’s own worked example runs from missing, to alt="puppy", to alt="Dalmatian puppy playing fetch". Specific and contextual beats generic, and generic beats empty. Keyword stuffing is called out explicitly as the wrong move.
Use a supported format. Google Search supports BMP, GIF, JPEG, PNG, WebP, SVG and AVIF, and asks that the filename extension matches the actual file type. If you have not revisited this, our breakdown of WebP versus AVIF versus JPEG covers which to pick where.
Nominate a preferred image per page. Google names three mechanisms: the primaryImageOfPage schema.org property, an image property on the page’s main entity, and the og:image meta tag. Choose one that is relevant and representative of the page, and avoid generic stock or extreme aspect ratios.
Shoot and keep quality high. Google’s wording is that high-quality photos appeal to users more than blurry, unclear images, and sharp images raise the odds of a click from a thumbnail. For camera search this is doubly true, because matching is visual.
Keep the landing page fast. A camera search lands on a page like any other result. Core Web Vitals thresholds are unchanged: LCP within 2.5 seconds, INP of 200 milliseconds or less, CLS of 0.1 or less, assessed at the 75th percentile of page loads and segmented by mobile and desktop.

The implementation checklist

Every commercially important image is in an <img> tag, not a CSS background
Image sitemap exists and is submitted in Search Console
Alt text written at the “Dalmatian puppy playing fetch” level of specificity on product, service and project images
File names are descriptive and the extension matches the real format
One nominated primary image per key page via primaryImageOfPage or og:image
No extreme aspect ratios on primary images
LCP, INP and CLS inside Google’s thresholds at the 75th percentile on mobile
Multimodal baseline exported and dated before the series fills out

Common mistakes and a 30 minute baseline audit

Five mistakes already visible in the wild

Reporting the September step change as growth. Covered above. It is the expensive one because it is the one a client remembers.
Treating multimodal as a rename of Image search. They are separate search types with separate definitions in the same filter.
Chasing keywords for it. There are no queries. Building a “Lens keyword strategy” is selling something that does not exist.
Panicking about a zero. For a consultancy or a SaaS, near-zero multimodal is the correct and expected result. It is not a problem to fix.
Ignoring it because the volume is small today. Small and new is exactly when a baseline is cheap to establish. In six months it is an argument about what the number used to be.

Pros and cons of acting on this now

Worth doing now	Not worth doing now
Exporting a dated baseline before the series grows	Rebuilding your reporting stack around a metric with no query data
Segmenting text-based from multimodal in every recurring report	Commissioning a “visual search strategy” as a standalone engagement
Fixing image indexing, alt text and formats, which pays off across normal search and Images too	Replacing good photography with more photography before the indexing basics are fixed
Telling clients and stakeholders about the measurement change in writing	Quoting a Canadian camera-search percentage that no published source supports

The 30 minute audit, in order

Open the Performance report and check whether the multimodal filter returns data at all. Record yes or no and the date range.
Export Web: text-based and Web: multimodal separately for the last 90 days. Save both files with today’s date in the filename.
Switch the table to the Pages dimension on the multimodal filter. The URLs at the top are the ones people are photographing their way into.
Open the top three of those pages and look at the images on them as a stranger would. Are they real <img> elements? Do they have alt text that describes the thing, not the keyword?
Check the same three pages against Core Web Vitals on mobile.
Check the Generative AI features report, which Google says also carries the multimodal data now, and note whether the pages overlap.
Write the two-sentence note to whoever reads your reporting, before they notice the step change themselves and ask.

That is a half-hour of work and it protects the credibility of every report you send for the next year. If you would rather it was done properly and documented once, that is the kind of measurement hygiene built into our website growth packages, and the image and performance groundwork sits inside our website packages.

FAQ

What does “Web: multimodal” mean in Search Console?

It is a search type filter covering search traffic where the user searched with an image rather than typed text. Google names four sources: Google Lens, Circle to Search on Android, images uploaded to Google Search, and Chrome image search. It sits under the Web heading alongside Web: text-based.

Why did my Search Console impressions suddenly jump in late September 2026?

Most likely because multimodal data was added on 24 September 2026 and was not in your counts before. Filter to Web: text-based only and compare the same period. If text-based is flat, the jump is the new data set rather than new traffic.

Is multimodal search the same as Google Images search?

No. Google’s documentation defines Image as traffic from results shown in the Google Images tab, and multimodal as traffic from searches made using images as the input. One is a destination, the other is an input method, and they are reported separately.

Can I see which queries or images drove multimodal traffic?

No. Google’s stated position is that because multimodal searches mostly use images rather than text, specific text query data is not available for this traffic. You get impressions, clicks, CTR and average position, plus the page, country, device and date dimensions.

Is the multimodal report available in Canada?

Yes. Google described the rollout as global starting 24 September 2026. It appears in your Performance report only if your site is actually receiving traffic from these searches, so a missing filter usually means no camera traffic rather than no access.

My multimodal filter shows nothing. Is something broken?

Probably not. Google only populates the metrics for sites receiving that traffic. For professional services, consultancies and most B2B software, near-zero is the expected result. Check that your key images are indexable, then move on.

How do I optimise for Google Lens specifically?

Make your images findable and interpretable. Use real <img> elements rather than CSS backgrounds, submit an image sitemap, write specific descriptive alt text, use a supported format with a matching extension, nominate a preferred image with primaryImageOfPage or og:image, keep photography sharp, and keep the landing page inside Core Web Vitals thresholds.

Does this change what I should spend on SEO?

Not by itself. Every action it justifies, image indexing, alt text, formats and page speed, already paid for itself in ordinary search. Multimodal gives you a new way to measure the result, not a new budget line.

The takeaway

Google has made a previously invisible behaviour measurable. People have been finding businesses by photographing things for years, and until 24 September 2026 none of it appeared in your Performance report. That is genuinely useful. It is also, for the next few weeks, a trap, because the new data arrives as a step change in your totals and will be misread as growth by anyone who does not segment it.

Do the unglamorous version. Split the search types. Export a dated baseline. Fix the image indexing basics Google has documented for years. Tell your stakeholders what changed before they ask. That is the whole play, and it takes an afternoon.

Want this done properly, once?

We build the measurement layer, the image pipeline and the reporting that goes with it, so the numbers you send out can survive a question. If you want your Search Console segmented correctly and your image stack audited against Google’s own documentation, start with our intake form and tell us what you are looking at. We will tell you what is reporting and what is real.

Search Console’s New Multimodal Filter Counts Camera Searches. Here Is Why Your Totals Just Changed.