Search Console Now Reports Visual Search: What Should You Do With It?

Why did a new search type just appear in my Search Console?

Because Google added one. On Thursday, 24 September 2026, Google introduced Search Console reporting for web multimodal search, and it shows up as a new multimodal search type filter inside the performance reports. If your site gets traffic from image based queries, the numbers start appearing on their own.

That distinction is the first thing worth saying out loud, because a new report always feels like news about your own performance. It is not. This is a measurement change rather than a traffic change. Nothing about your site got better or worse on 24 September. You can simply see a slice of it now that was invisible before, and an empty slice is as informative as a full one.

The announcement came from Harsh Kharbanda, Product Manager Lead for Google Lens, and Moshe Samet, Product Manager Lead for Search Console. That pairing tells you most of what you need to know about where the data comes from.

What exactly did Google announce?

Google said it is introducing Search Console reporting for web multimodal search in two places: the report for performance on Search results, and the report for Generative AI features. To see it, you use the new multimodal search type filter in the performance reports. You can click Export to pull the data out.

The two report locations matter more than they look. Multimodal appearing in the Generative AI features report means Google is treating image led queries as part of the same family as its AI surfaces, not as a separate corner of Google Images. If you already audit that report, you now have a second dimension inside it.

Google also said the integration is rolling out globally starting that day, and that you will start seeing these metrics in your Performance report if your site is receiving traffic from these queries. So an empty multimodal filter is not a bug. It is an answer.

One thing the announcement does not say is whether you get historical data or only data from the rollout forward. I am not going to guess. Look at the earliest date the filter returns on your own property and treat that as your baseline, because that is the only honest read available right now.

What counts as a multimodal search?

Google was specific. The data reported here includes searches with Lens, Circle to Search on Android, image uploads to Google Search, and the Chrome right-click "Search this image" feature. Four named entry points, all of which start with a picture rather than a typed phrase.

Notice what those four have in common. In every case the person had something in front of them: a product on a shelf, a screenshot of a dashboard, a diagram in a slide deck, an error message on a monitor. They did not know the words for it, or they could not be bothered to type them. The image was the query.

That is a different intent shape from a keyword. Someone typing best crm for agencies has already learned the vocabulary of the category. Someone circling a screenshot of a pricing table has not. They are earlier, more confused, and more open to being told what they are looking at.

Why does this matter if you sell software and not sneakers?

Visual search gets written about as a shopping feature, and for good reason. But the entry points Google named are not shopping specific. A right-click on an image in Chrome happens as often on a B2B blog post as it does on a product listing, and screenshots are the native currency of software.

Think about how people actually encounter software. Someone posts a screenshot of your dashboard on LinkedIn. A colleague sends a cropped image of a workflow in Slack. Someone takes a photo of a conference slide that had your architecture diagram on it. Every one of those is a moment where a person has your interface in an image and no idea what it is called.

That is the gap this report lets you see. Not "are people searching for my category" but "are people pointing a camera or a cursor at something of mine and asking Google what it is". For a B2B SaaS company, that second question is a proxy for how far your visuals have travelled without your brand name attached.

I would also expect this to matter unevenly. A company whose product is mostly a set of dense screens will show up here differently from one whose product is an API. Do not decide the channel is irrelevant to you until you have looked at your own filter.

How big is visual search actually?

Google has published a number, and it is not small. On its own ads and commerce blog, Google says that every month, people use Lens for nearly 20 billion visual searches, and that 20 percent of all Lens searches are shopping-related. That is Google describing Google, which is the only source worth trusting here.

Read the second half of that sentence carefully, because most write-ups skip it. If a fifth of Lens searches are shopping, four fifths are not. That is the part of the pie that includes the screenshot of a settings panel and the photo of a whiteboard. Nobody has been able to measure their own share of it until now.

I am deliberately not going to extrapolate that number into an estimate of your traffic. Google's figure covers Lens globally across every kind of query. The only number that tells you anything about your site is the one in your own multimodal filter.

What should you check in the first week?

Start by opening the performance report, applying the multimodal search type filter, and exporting the result. Do that before you change anything, because the export is your baseline and the report is new enough that you want a copy outside Google's interface.

Then look at which pages appear, not which queries. Query data on image led searches will be thin by nature, since the person did not type a phrase. Pages are the reliable dimension. If a page you did not expect is carrying your multimodal impressions, that page has an image doing work you did not plan for.

Next, compare that page list against your ordinary web performance. A page that ranks poorly for text but shows up in multimodal is telling you something specific: the picture on it is more findable than the words around it. That is a copy problem disguised as an image win, and it is the most actionable thing this report surfaces.

Finally, write down what you found and the date. I keep a single running document per client for exactly this, because in six months nobody will remember what the first week of a new report looked like, and the comparison is the whole value. This is the same discipline I apply when I audit the AI Mode report in Search Console, and it pays off for the same reason.

Does this change how you should treat images on your site?

It changes the priority, not the practice. Everything that makes an image understandable to a machine was already worth doing before this report existed. What has changed is that you can finally see a consequence when you skip it, and a visible consequence is what moves image work out of the nice-to-have column.

The unglamorous work is still the work. Descriptive file names instead of screenshot-2026-09-24. Alt text that says what the image shows rather than repeating the heading. A caption or a nearby sentence that names the thing in the picture in the words a customer would use. Compression and correct dimensions so the image actually loads. None of that is new advice and all of it is still where the leverage is, which is why I treat image optimisation on Webflow as a foundation rather than a polish step.

Alt text deserves a specific mention because it is the field most often filled in badly by people who know it matters. An alt attribute is a description for someone who cannot see the image, and it happens to also be a description for a machine that is trying to match a photograph to a page. Those two jobs align almost perfectly, which is lucky, and it is why I push clients to write alt text for accessibility first and let the search benefit follow.

Here is my own bias, earned from six years of this: sites lose visual search not because their images are badly optimised but because their images are generic. A stock photo of a team high-fiving cannot be matched to anything, because it is not of anything. A screenshot of your actual product can.

Is this a new channel or just a new number?

For now it is a new number, and that is not a criticism. Every channel worth having started as a number that somebody bothered to write down before it was fashionable. The mistake is jumping from a fresh metric to a strategy without the months of data that would justify one.

The honest position is that no strategy is justified by a report that shipped this week. You do not have enough data to see a trend, and neither does anyone claiming otherwise. What you do have is a chance to establish a baseline before the people who wait six months, and to find out whether your own property has meaningful volume here at all.

My framework for anything like this is three questions in order. Is there volume for me specifically? If yes, is it on pages that matter commercially? If yes, can I influence it with work I would want to do anyway? A no at any step ends the exercise, and most of the time the answer is no at step one. That is a fine outcome. Knowing a channel is empty for you is worth as much as knowing it is full.

What I would resist is the reflex to build a visual search programme. The reflex is understandable, because a new report feels like a mandate. It is not. It is an instrument.

What should you do next?

Open Search Console, apply the multimodal filter, export it, and save the file with today's date. If the report is empty, note that and check again in a month. If it has data, list the pages and compare them against your text performance. That whole exercise takes under an hour and it is the entire correct response this week.

Then leave it alone. Come back in ninety days with a second export and see whether anything moved. The strongest thing you can do with a brand new metric is refuse to act on it until you have two readings, and the second strongest is to make sure you took the first one.

If you want a second pair of eyes on what your multimodal filter is actually telling you, or you want your images and alt text tightened up so there is something worth measuring, reach out. I am happy to look at a real property and tell you honestly whether there is anything there.