Search Console Adds Multimodal Search Filter

Reading time: 7 min

I was doing SEO before anyone called it SEO. The discipline hasn’t changed as much as the tooling. But every few years, the tooling changes the discipline.

On September 24, 2026, Google announced a multimodal search filter for the Search Console performance reports. A dedicated reporting category that isolates traffic from image-based queries. I’ve seen this play out before — every time search expands into a new input modality, publishers get a fresh measurement layer that eventually becomes mandatory to learn whether they want to or not. Voice search did this in 2016. Visual search through Lens has been cooking since 2017. The measurement never followed the behavior. Until now.

A New Search Type Now Sits Beside the Old Ones

Until this week, click data from Google Lens and similar visual tools was buried inside whatever bucket Search Console felt like assigning it to. You couldn’t isolate it. You couldn’t chart it. You couldn’t put it in front of a product manager and say: look, here’s the delta this visual optimization work produced.

Here’s what actually changed. The performance reports now expose a new search type labeled “multimodal.” When you apply it, the interface filters down to clicks, impressions, CTR, and average position that originate from a query launched with an image — not text. It works in both the standard Search Results report and the newer Generative AI features report, and there’s an export button if you want to pull the rows into a spreadsheet or BI tool.

No surprises. No hidden dashboards. No “contact your account manager.” Default surface.

Four Sources Feed the Multimodal Bucket

Google spelled out exactly what it counts as a multimodal query. The list is short enough to check against your own behavior as a user:

• Google Lens — the point-your-camera-at-something workflow that’s been baked into Android and the Google app for years.
• Circle to Search on Android — gesture over anything on screen, and the circle turns into a query.
• Direct image imports to Google Search — drag, drop, or upload a photo into the search bar.
• Chrome’s “Search this image” — the right-click option on any page, which most people forget exists until it doesn’t.

All four share the same trigger: the user starts with a picture, not a phrase. That’s the grouping logic. There’s nothing exotic underneath.

Where to Find the Filter — and What to Expect When You Do

Open the Performance section of Search Console. Click the Search type dropdown. Select Multimodal. That’s it.

Then brace yourself, because there’s a catch that’s going to disappoint a lot of people who were hoping for a new growth chart.

The metric only shows up if your site actually received traffic from multimodal queries. If nobody has ever pulled up your page from a Lens search or a Circle to Search gesture, the row is simply absent. Google’s documentation phrases this politely, as a “condition.” Translation: no data does not mean no filter. It means your content hasn’t been surfaced through image-first discovery. That’s a signal, not a bug.

Slow down. Think about that for a second. An empty multimodal report is not a metadata problem. It’s a visibility problem. And it’s telling you something you’d never have learned from the old interface.

The Rollout Is Global, but Slow

The deployment went live worldwide on September 24, 2026. In practice, global rollouts never hit everyone in the first week. If your property doesn’t show the option on day one, give it a few days — but check the performance-anomaly reports for whether image traffic might already be arriving under your radar from a mirrored version of the tool.

This is standard Google behavior. I’ve seen this play out before with almost every Search Console feature shipped since 2015 — the Discover report, the Core Web Vitals report, the URL Inspection tool. Global on paper, uneven in practice.

The YouTube, Instagram, and TikTok Angle That Got Buried

Nobody talks about this part.

Google published a LinkedIn post from Search Central around the same time as its blog entry. Most reports skipped it, or buried it in a footnote. The post stated clearly that multimodal data is also available for platform properties — the accounts for YouTube, Instagram, and TikTok that Search Console now tracks. Meaning: creators and brand accounts operating on social platforms can see whether their content is being served through visual search surfaces they likely never optimized for.

That’s not a small thing. Most social teams I’ve worked with have zero visibility into how their asset catalogs — thumbnails, short-form video covers, product photos posted directly to Instagram — are being picked up by Google’s crawlers. Some of those assets already earn image search traffic. Until this week, tracking it required crawling server logs, building custom regex, or guessing.

The playbook changed. Again.

Why This Is a Business Argument, Not a Dashboard Update

Google didn’t just ship a metric. It shipped a negotiating tool.

Every SEO I’ve worked with has been dragged into a version of this conversation: engineering says alt-text cleanup is a low-priority ticket. Content ops says image processing is out of scope. Design says image files at that size are already big enough. Marketing is the only one advocating for making images crawlable, indexable, and cleanly named.

Now you can walk into the next roadmap review and put actual numbers on the table. “This page generated X clicks and Y impressions from Circle to Search and Lens in the last 30 days. Those came from an audience that never typed a query.” Nobody has ever successfully argued that work needs doing with that evidence before. Because before, the evidence didn’t exist.

Google is barely hiding this. The Search Central team explicitly mentions e-commerce, media, recipes, decor, and fashion as the verticals that benefit most. Translation: sites with rich image catalogs. Sites where the images often are the product.

The Category That Was Always There in Sight

Let me show you the data angle most people are missing. If you’ve ever pulled top image URLs from an e-commerce property’s server logs, you already know that a meaningful slice of your image-displayed impressions have always come from something other than a keyboard. Lens has been serving as a shopping intent entry point for years. The purchase decisions happening from photo-first queries never showed up as “query” rows because there was no query to show.

Now there is. And now it’s chartable.

For publishers in verticals where images carry the page — recipe sites, holiday rentals, product catalogs, magazine features — the multimodal filter is likely to become a first-glance tab within 12 months. Not because Google demands it, but because the traffic numbers will make the case for you. The long tail of visual search behavior is real, it’s growing, and it’s been totally uncounted. That changes today.

How to Actually Use It

Here’s my practical playbook, and it’s intentionally lean because there’s no reason to overthink this:

• Pull the report monthly, export the rows, and stack them against your standard query data. Look for pages that show multimodal clicks but weak text-query clicks. Those are your visual-strong, text-weak assets — and they deserve unique treatment.
• Check image alt attributes, filenames, and file sizes on those pages first. If multimodal traffic is clustering on content where images have empty alt text or ugly filenames like IMG_2934.jpg, you’ve found a ten-minute win.
• Cross-reference with Core Web Vitals. Multimodal queries are more likely to land on mobile, and mobile performance is where most poorly optimized image-heavy pages collapse.
• Don’t panic-optimize the whole site. Nobody has the budget for that, and most sites don’t need it. Focus on the content clusters where image-first discovery actually fires.

I’ve built my share of image-optimization campaigns that produced nothing measurable. Not because the work was bad. Because I had no way to attribute the results. This filter fixes exactly that problem. Which means the excuse just walked out the door.

What’s Coming Next

The pattern is now set. Google is slowly exposing the inputs that were always there — voice, video, visual, AI-summarized — as distinct search types inside Search Console. Once a modality reaches enough volume to matter, the reporting arrives. The multimodal filter of September 2026 is one node in that arc.

This isn’t a take — it’s a pattern. Search has been losing its text-only character since the first camera icon appeared in the search bar. The metrics just caught up.

If you spot issues or want to compare approaches with other practitioners, Google’s help forum accepts feedback through the familiar “Send feedback” link inside Search Console. Do that. Reports sometimes improve faster than publishers assume when they say something concrete.