Search Console Multimodal Search Report: How to Read Google Lens and Image Search Data

What is the Search Console multimodal search report?

The Search Console multimodal search report is a new search type filter that separates clicks and impressions from searches people start with an image instead of text. It covers Google Lens, Circle to Search on Android, image uploads to Google Search and the Chrome right-click "Search this image" option.

Google announced the feature on September 24, 2026. As a result, the "Web" search type in your Performance report now splits into two options: text-based and multimodal. In this guide I explain what the filter measures, what it leaves out and how to read the numbers without jumping to the wrong conclusions. If you need a refresher on the basics first, start with my Google Search Console guide.

When and how did Google announce it?

The announcement went live on Thursday, September 24, 2026, on the Google Search Central Blog. Harsh Kharbanda, Product Manager Lead for Google Lens, and Moshe Samet, Product Manager Lead for Search Console, signed the post. So this is not a rumour or a limited test; it is an official global rollout.

The post gives three clear facts. First, the data appears in the Search results performance report and in the generative AI performance report. Second, the rollout started globally on the day of the announcement. Finally, you only see the metrics if your site actually receives traffic from these searches.

In other words, an empty chart does not automatically mean something broke. Your site may simply not appear in image-led searches yet, or the rollout may not have reached your property. Keeping that distinction in mind saves you from a false diagnosis.

The announcement at a glance

• Date: September 24, 2026, global rollout.
• Location: the search type filter in the Performance reports.
• Scope: Lens, Circle to Search, image uploads and Chrome image search.
• Export: the "Export" button lets you download the data for other tools.

Which search surfaces count as multimodal?

The Performance report help page defines the type as traffic from searches where users search with images, for example with a smartphone camera. Four surfaces feed it.

1. Google Lens: someone points a camera at a product, a sign or an object.
2. Circle to Search: someone circles part of the screen on an Android phone.
3. Image upload: someone uploads a photo straight into Google Search.
4. Chrome image search: someone right-clicks an image and picks "Search this image".

All four share one trait: the search starts with a picture, not a phrase. Therefore the intent looks different too. Most of the time the user asks "what is this, where can I buy it, is there something similar?" For online stores, fashion, furniture, spare parts and travel sites, that intent often carries a clear purchase signal.

On the other hand, the filter does not include searches inside the Google Images tab. That traffic still lives in the separate "Image" search type. The table below makes the difference explicit.

How does text-based web differ from multimodal web?

The search type filter now offers five options, and two of them sit under "Web". To avoid confusion I put them all in one table. The definitions come from Google's help documentation.

The key point: multimodal web traffic captures moments where the user starts with an image but ends up on a web result that shows your page. In practice, your page answers a visual question. That is why this data can carry more commercial intent than traffic from the Images tab.

How do you find Search Console multimodal search data?

Getting to the data takes only a few clicks. However, the default view keeps the old behaviour, so you need to pick the filter yourself. The steps below follow the English interface.

1. Open the property in Search Console.
2. Go to Performance, then open the Search results report.
3. Click the "Search type" filter at the top.
4. Choose the multimodal option under "Web" and apply it.
5. Focus the date range on the period after September 24, 2026.
6. Check the Pages, Countries and Devices tabs in turn.

You can also use comparison mode. That way you see text-based and multimodal traffic side by side for the same period. The comparison quickly shows which pages perform unusually well in image-led search.

One more tip: after you set the filter, bookmark the URL in your browser. Then you return to the same view each month in one click, and you can share the exact settings with colleagues.

The same filter also exists in the generative AI performance report. That report behaves a little differently, so I cover it in its own section.

Which metrics and dimensions does the report show?

With the multimodal filter active in the Search results report, you get the familiar metrics: clicks, impressions, click-through rate and average position. The definitions stay the same; only the data set gets narrower. For example, average position now reflects your topmost result in the web results that follow an image-led search.

For dimensions, pages, countries, devices and dates will do most of the work. The device split matters most. Circle to Search runs only on Android and people mostly use Lens on a phone, so expect a high mobile share. Chrome image search, by contrast, can bring in some desktop traffic.

• Clicks: visits from image-led searches.
• Impressions: how often your page appeared in those web results.
• CTR: how well visual intent turns into a click.
• Average position: your average rank in those results.

When you compare CTR across page groups, my CTR calculator helps you check segments by hand. It also makes differences between exported page groups easier to spot.

Why is there no query data?

Above all, the query tab causes confusion. According to press coverage of Google's help documentation, multimodal searches mostly use images rather than text, so specific text query data is not available for this traffic. Consequently, you cannot use the queries dimension once you select this search type.

That limitation makes sense, though. When someone photographs a pair of shoes, nobody types a word. Google interprets the image, but it does not report that interpretation as a query. So your analysis has to start from pages.

How do you estimate intent then? My method is simple. I look at which images sit on the pages that receive multimodal traffic. Then I switch to the text-based queries for the same pages. Put the two lists next to each other and you can usually infer which product or object drove the visual search. It is not an exact match; still, it gives you enough of a signal to act on.

To group text queries faster, try my search terms n-gram analyzer.

What does the filter show in the generative AI report?

In June 2026 Google introduced the generative AI performance reports and opened them to all sites worldwide as of August 31, 2026. The Search version tracks impressions inside AI Overviews and AI Mode. With the new announcement, the same report also gained the multimodal search type filter.

There is one important difference. The generative AI report currently shows impressions only, not clicks. So when you pick the multimodal filter there, you see how often your page appeared in AI answers to image-led searches. You cannot read how many visits those appearances produced.

This data shows how closely visual search and AI answers now overlap. For instance, a user photographs a plant, AI Overviews suggests how to care for it and cites your page as a source. Even without a click, that impression builds brand visibility. I cover the wider picture in my article on zero-click searches.

Why shouldn't you mix it up with Image search traffic?

For years many site owners judged image SEO by the "Image" search type. The new filter does not change that data; it adds a new layer on top. Adding the two together does not double count, because the sources differ. But if you read them as the same thing, you will draw the wrong conclusions.

In the Images tab, a user types words and browses visual results. With multimodal web search, however, a user provides an image and gets web results back. In the first case your image is the shop window; in the second your page is the answer.

• Low Image traffic, high multimodal traffic: your page copy answers visual intent well.
• High Image traffic, low multimodal traffic: your images attract attention, but the page context stays weak.
• Both low: you need to work on image assets and page context together.

These three scenarios clarify which team should prioritise which task.

Which sites benefit most from this data?

That said, the data does not carry the same weight for every site. It matters most in sectors where people see an object and ask "what is this?" In my experience, sites built around physical products or identifiable objects stand out here.

• Ecommerce: clothing, shoes, accessories, furniture, home decor, electronics.
• Spare parts and technical products: users photograph a part to find the model.
• Travel and venues: users ask what a building or a view is.
• Food and recipes: users want the recipe behind a photo of a dish.
• Nature, garden and hobbies: identification searches for plants, insects or stones.

Service businesses can benefit too, but less directly. For example, an architecture practice might catch traffic from a photo of a facade detail. On the other hand, abstract services such as consulting will probably see small volumes. So I suggest you match the time you spend on this report to your sector.

How should you read the data in the first weeks?

The rollout began on September 24, 2026, which means you only have a short time series so far. The most common mistake I see right now is drawing big conclusions from a few days of data. With a short series, even the weekday versus weekend pattern can distort the trend.

At this stage, look for answers to three questions. First, does your site get this traffic at all? Second, which pages collect it? Third, how does it compare with the text-based performance of the same pages? For trend analysis, wait for at least four to six weeks of data.

Also keep your expectations for historical data modest. Google's announcement makes no promise about how far back the data goes. Therefore it is safer to mark the rollout date as the starting point in your own reports.

The principles from my guide on how to read a digital marketing report apply here as well.

What should you ask in a page-level analysis?

Because there is no query data, the Pages tab becomes the heart of the report. I use a short checklist for every page that receives multimodal traffic. It also standardises the repeat work when my team and I review client sites.

1. Does the main image resemble a photo a user might actually take?
2. Can users read the object's name clearly in the text around the image?
3. Does the page answer "what is this?" above the fold?
4. If it is a product page, can users see price, stock and variants?
5. Does the page load quickly on mobile?
6. Is its CTR below the site's multimodal average?

Specifically, pages with high impressions but low CTR offer the fastest wins. They already appear in visual search; they just fail to convince users on the result. Working on the title, description and page context is the shortest route here. To preview titles and descriptions, use my Google SERP preview tool.

How do you export and analyse Search Console multimodal search data?

Google specifically points to the "Export" button in the announcement. You can download the data to Google Sheets, Excel or CSV and work with it in your own tools. However, values that show as "~" or "-" in the report turn into zeros in the download, so account for that in your calculations.

Here is the simple workflow I recommend:

1. Export the page lists for text-based and multimodal traffic for the same date range.
2. Join the two lists on the page URL.
3. For each page, calculate the multimodal share of total web impressions.
4. Tag pages with a high share as "visual intent pages".
5. Prioritise image quality and page context work on those tagged pages.

You can also connect this table to your analytics. For example, compare the conversion rate of the same pages in GA4 to estimate how much visual intent traffic contributes to sales. For methods to separate AI traffic, see my article on AI traffic in GA4.

How do you match it with GA4?

Do not expect a separate source label in GA4 for visits from image-led searches. In most cases these visits sit inside general Google organic traffic. So you need to match the two tools at page level rather than at source level.

The practical method works like this. Take the list of pages with multimodal traffic from Search Console. Then pull organic sessions, engagement and conversions for the same landing pages from GA4. Once you join both tables on the landing page, you see the sales performance of pages with a high visual intent share side by side.

This method does not give you exact attribution; instead, it shows direction. For instance, if a product page with a high multimodal share also converts above average, your image investment is likely paying off. If not, review the page context and product details.

• Join key: the page URL.
• Date range: identical in both tools.
• Comparison: page groups with high and low multimodal share.

How do you prepare your images for Lens searches?

The report only measures; image quality and page context drive the results. Google's Google Images best practices set the basic framework: place images near relevant text, write descriptive alt text, use high-quality images and keep image files crawlable.

For visual search, those recommendations translate into something concrete. The user's camera sees a real-world object. So studio shots alone may not be enough; photos from different angles and in real use settings also help. Moreover, a clean, simple composition makes the object easier to recognise.

On the technical side, file size and loading speed also matter. I cover compression, formats and lazy loading in detail in my guide on how to optimize images for speed and SEO, so I will not repeat it here. For a quick resize, use my image resizer.

What changes for product pages?

For online stores, this report also makes previously invisible demand visible. When someone photographs a bag they like on the street, your product page might answer that search. Now you can measure how often that moment happens.

To answer visual intent on a product page, a few basics stand out. The copy should state distinguishing details such as product name, brand, model, colour and material. In addition, product structured data presents price and availability in a machine-readable way. My schema markup guide is a good starting point for that.

For the overall structure of a product page, see my ecommerce product page guide. If you want to build your store's growth plan together, my team and I run this analysis for you as part of our ecommerce consulting.

What can small businesses do with this report?

For a small business, the first benefit is simple: you can see whether demand exists at all. A local furniture workshop, for example, can now measure whether its product photos get any response in visual search. Even with low numbers, a single page that earns regular visits from this channel sends a valuable signal.

Also, you do not need a big budget. Clear, well-lit product photos from different angles, even when shot on a phone, often make a good start. Also, stating the product name, size and material next to each image helps Google match the picture with the right context.

My advice for small businesses: open the report once a month and check only three things. Look at total multimodal clicks, the page with the most traffic and the mobile share. Those three numbers tell you where to put your photo and content effort.

Which mistakes should you avoid?

Whenever a new report arrives, the first reaction often comes too fast. I list the following mistakes separately because I see them often.

• Reading an empty report as a penalty: no data may simply mean no traffic from these searches.
• Looking for queries: this type has no query dimension, so think in pages.
• Adding it to Image traffic: merging two intents into one metric distorts the reading.
• Calling trends too early: do not decide based on a few days.
• Expecting clicks in the AI report: that report counts impressions only for now.
• Relying on stock photos: images that appear on many other sites do not set you apart.

One more mistake: looking only at the page with the most traffic. The long tail may hold dozens of product pages with few impressions each but real volume in total. When you review them in groups, you spot category-level image problems or opportunities much more easily.

Which decisions does Search Console multimodal search data support?

A report is only as useful as the decisions it drives. In my view, Search Console multimodal search data helps with four concrete decisions. Making them with real data means your image budget rests on evidence rather than guesswork.

1. Photo shoot priority: which product group needs new photography first.
2. Content refresh order: which pages carry visual intent but have weak copy.
3. Mobile improvements: which high mobile share pages need speed and layout fixes first.
4. Category expansion: which groups draw visual demand while the site offers few products.

For example, a home textiles store that sees its curtain pages leading in multimodal traffic can increase photography and content investment in that category. Above all, that decision now rests on a real number. In short, the report moves image work from "make it look nice" to "make it bring traffic".

Do not base these decisions on a single month. A signal that points the same way across at least two periods gives you much firmer ground.

Which targets make sense?

In practice, setting targets for a new channel calls for realism. On most sites, multimodal traffic will start as a small slice of total web traffic. So for the first quarter, I recommend share and coverage targets instead of absolute click targets.

1. Coverage target: increase the number of pages with multimodal impressions.
2. Efficiency target: bring CTR on those pages closer to your own site average.
3. Business target: track the conversion rate of visual intent pages.

Do not benchmark these targets against an outside industry average; no reliable, official benchmark exists yet. Use your own site's baseline as the reference and measure the same way each month. That way you see progress consistently.

Who should see the Search Console multimodal search report?

Moreover, this data matters beyond the SEO team. The people who shoot product photos, category managers and content writers all benefit directly. For example, the photo team can see for the first time, with real numbers, which shooting style shows up more in visual search.

When you share the report, share the context too. State up front that there is no query data, that the AI report has no clicks and that the data starts on September 24, 2026. That way everyone reads from the same frame and nobody forms the wrong expectations.

After that, a short monthly summary usually does the job: total multimodal clicks, the top five pages, the device split and the change versus last month. Those four lines give you solid ground to discuss the return on image investment.

How does this data change your SEO strategy?

I think the real importance of this announcement goes beyond measurement. Google now treats image-led search as a performance channel of its own. That confirms visual assets are not just a supporting element in SEO; they are a direct traffic source.

In practice, I suggest three changes. First, add a "visual intent" column to your content plan and mark which pages describe an object. Next, steer your image production budget towards those pages. Finally, add multimodal traffic as its own line in your monthly report.

This approach also fits AI search. Because visual search and AI answers increasingly meet in the same experience, I recommend you treat image SEO as part of the same plan as classic SEO and AI visibility work, not as an isolated island.

How do my team and I use this data?

On the sites we work on, my team and I added this report to our existing SEO audit process. Step one: we open the filter and check whether any data exists. If it does, we export the page list, join it with text-based traffic and pull out the visual intent pages.

Next, we review image quality, alt text, the copy around each image, structured data and mobile speed on those pages. For prioritisation we focus on pages with high impressions but low CTR, because they deliver the fastest return. Finally, we track multimodal traffic as a separate line in the monthly report.

If you want to run this analysis on your own site together, our SEO consulting includes visual search performance in the measurement plan. Put simply, the goal is not to follow one more report; it is to turn the demand that report reveals into sales.