People search by photographing products, circling objects, or uploading screenshots. The new Google Search Console multimodal search report shows which of your pages appear when an image is part of a web search.
Announced September 24, 2026, the report can guide visual SEO work when its page-level data is interpreted carefully.
What is the new Web: multimodal filter?
The Google Search Central announcement adds Web: multimodal as a search type in two Search Console reports: Performance for Search results and Performance for Generative AI features. It covers web searches that use an image. Google lists Lens, Circle to Search on Android, image uploads to Google Search, and Chrome’s “Search this image” as included examples.
This is a view of web search results produced after an image was used in the search, not an inventory of images hosted on your website. That distinction changes the entire analysis. A clicked product URL tells you someone reached that page after an image-led search. It does not reveal the photograph they submitted or prove that your hero image matched it.
Rollout began globally on September 24. Check the dates present in your property before comparing periods; Google has not given every site a universal historical start date.
If you need the basics first, use Scale Xpert’s Search Console setup and reporting guide. The new filter is useful only when you already know which verified property and URLs you are reviewing.
Where does the answer come from when someone searches with an image?
There are three distinct things in an image-led journey: the user’s input image, the results Google retrieves, and the pages linked in the response. They are related, but they are not the same object. Google Search Console’s new report mainly helps you inspect the last of those at the URL level.
For a visual search, Google can interpret objects in a picture and retrieve results. Google describes AI Mode identifying multiple items, searching for them, and assembling an answer with web links. That account does not reveal the steps behind a specific report row.
For AI Overviews and AI Mode, Google’s site-owner guidance describes related searches across subtopics and data sources, sometimes called query fan-out, that can produce supporting website links. The links can vary between AI experiences. Google does not publish a formula telling you which photograph or page will be selected for a given visual prompt.
How to find your multimodal pages in Search Console
Open your verified property, then select Performance > Search results. Choose Search type > Web: multimodal, set a date range with available data, and open Pages. You can inspect clicks, impressions, CTR, and average position under the Search results report definitions. Google also supports exporting the selected view for analysis.
Start with the Pages tab because the Queries dimension is unavailable for multimodal traffic. Google’s dimensions documentation says specific text query data is not available, since many of these searches begin with images. You can still examine pages, dates, devices, and countries that the interface provides. Do not fill the missing query column with a guessed keyword.
Filter to one important URL, then keep the page and date range fixed while viewing Web: text-based. You can compare the page’s performance for two types of web search. Use the same device and country filters when they matter. This is a comparison of measured page visibility and clicks, not an explanation of what the user photographed.
Check the canonical if your CMS creates duplicates: Google usually assigns page data to it. Chart and Pages table totals can differ because they use property versus page aggregation.
Four numbers that must stay in separate columns
Keep each measurement labelled. An AI impression is not a visitor, and an Image search click is not necessarily a Lens click.
| View | What it can tell you | What it cannot establish |
|---|---|---|
| Search results > Web: multimodal | Page-level web impressions and clicks after an image was used in the search | The submitted image, exact query, or individual Lens-versus-Circle share |
| Search results > Web: text-based | Page-level web impressions and clicks from text-led searches | The cause of a difference between the two groups |
| Search results > Image | Performance of Google’s separate Image search type | That every visit came from an image-led Web: multimodal result |
| Generative AI > Web: multimodal | Link impressions in AI Overviews and AI Mode for image-led searches | AI clicks, the specific source photograph, or a breakdown of each AI feature |
Google’s Generative AI report documentation specifies impressions for AI Overviews and AI Mode. An impression means a link to your site was shown in a generative AI feature. It does not prove a visit, identify the image the person uploaded, or show which visual element contributed to the answer.
For context, compare our AI Performance Report guide and coverage of Google Images and visual discovery. Keep metrics in separate columns: AI impressions are included in broader Web reporting, so adding them to Web totals as independent visits would mislead readers.
An image-in, page-out audit for SEO teams
Build a small worksheet before editing images. Export the multimodal Pages view for a complete comparison period. Record the property, dates, device and country filters, final canonical URL, impressions, clicks, and any data freshness caveat. Add a column for the URL’s business goal: purchase, lead, signup, support resolution, or another action your site measures.
Inventory each page’s images and the question it answers. This is an audit inventory, not a list of images Google matched to users’ uploads. The filter cannot identify which photograph caused an impression.
Use the following blank template for every URL you investigate:
| Field | Enter your property’s evidence |
|---|---|
| Date range / country / device / property | ___ |
| Canonical landing page and business goal | ___ |
| Web: multimodal clicks and impressions | ___ |
| Web: text-based clicks and impressions, same filters | ___ |
| Separate Image search clicks, if useful | ___ |
| Separate Generative AI: multimodal impressions, if available | ___ |
| Visible visual assets and page context | ___ |
| Verified issue, proposed change, and implementation date | ___ |
| Outcome and plausible alternative explanations | ___ |
Prioritize pages whose visual task and business outcome are clear, such as a product detail or screenshot-based troubleshooting page. These are examples for choosing what to audit, not predictions of future rankings.
Improve visual discoverability without promising a citation
Start with technical access. Google’s image SEO best practices say crawlers can discover images through an <img> element with a src attribute, including inside <picture>, whereas CSS background images are not indexed as images. Check whether important editorial or product images are embedded in discoverable markup and can be fetched. For responsive formats, keep a usable src fallback.
Review the whole landing page: surrounding text, captions, filenames, and alt text give images context. Describe what each image shows, with relevant product details nearby when helpful. Avoid stuffing keywords into alt text.
Create useful assets: original product angles, labelled process photos, or screenshots of relevant interface states. Explain them in readable text. Google advises using relevant, high-quality images alongside text for generative AI search. Neither an original picture nor a caption guarantees an AI citation.
Confirm the URL is indexable and useful on mobile; our on-page SEO checklist covers wider page checks.
What actually improves the chance of being linked in an AI answer?
For a page to be eligible as a supporting link in AI Overviews or AI Mode, Google states that it must be indexed and eligible for a Search snippet. That is a baseline requirement, not a promise of inclusion. If a page is blocked, miscanonicalized, or provides too little accessible explanation, fix that concrete problem first.
Give readers verifiable evidence. A product comparison needs labelled photos and current specifications; troubleshooting needs a screenshot, tested conditions, and steps in text. Record when and where original photos were taken where it matters. These examples improve usefulness, not a secret selection formula.
Google’s generative AI guidance does not require AI-specific schema, llms.txt, tiny content chunks, or rewrites solely for AI. Use structured data for documented Search features where it accurately describes the page. A backlink exchange cannot guarantee AI attribution.
A realistic measurement cadence
Establish a baseline from complete days, save the export, and record implementation dates alongside other changes. After Google processes the pages, repeat the same page, country, device, and comparable-period checks. Report absolute counts alongside rates and do not promise a fixed indexing deadline.
Check your analytics for the landing page’s actual outcomes. A click to a Discord invitation can be measured as an outbound action, but it should not be relabeled a completed community join without a separate confirmation. Search Console’s multimodal filter cannot on its own attribute a lead or membership to Lens rather than another qualifying image-led entry point.
Share your audit method and limitations in the Scale Xpert Discord community so others can check your observations.
Frequently asked questions
Can Search Console show the photo someone uploaded to Google Lens?
No. The new filter shows performance for web results reached through searches using images. It does not provide the user’s uploaded image or a page-by-page record of the visual input.
Why is the Queries tab missing from Web: multimodal?
Google says specific text queries are not available for this search type because many searches rely primarily on images. Use pages, countries, devices, and dates for the available analysis rather than guessing hidden terms.
Is Web: multimodal the same as Google Images traffic?
No. Google lists Web: multimodal as web results where an image was part of the search and Image as a separate Search type. Compare them as distinct labelled views.
Can the report separate Lens and Circle to Search?
The announced filter groups qualifying image-led web searches, including both of those entry points. Google has not described separate Lens and Circle totals in this view.
Does an AI multimodal impression prove Google cited my photograph?
No. It shows that a link to your page appeared in a supported AI feature during an image-led search. The report does not identify which image or source passage contributed to the generated response.
Where do I see clicks from these searches?
Use Performance > Search results > Web: multimodal for web search clicks. The separate Generative AI performance report currently documents impressions, not click metrics.
What if my property has no multimodal data?
Check the selected property, date range, and whether the rollout has reached your view. The feature appears where qualifying activity is recorded; an empty report alone cannot diagnose image quality or technical eligibility.
Can better alt text or schema guarantee an AI citation?
No. Useful alt text helps describe an image, and appropriate structured data can support specific Search features. Google does not offer a markup or wording trick that guarantees an AI supporting link.
Conclusion
Google Search Console’s Web: multimodal filter makes one part of visual discovery measurable: which pages receive web visibility and clicks when people search using an image. It does not show the input image, the exact text query, or which asset Google relied on. The separate AI report adds link impressions, not a proof of image citation or visits.
Use the filter to find pages worth inspecting, verify the problems on those pages, make useful visual and textual improvements, and compare the same labelled metrics after Google processes the changes. That method gives Scale Xpert readers an honest way to act on a new report and a stronger body of evidence to cite.
