What Google’s Multimodal Report Revealed Across My Small Business Sites

What Google’s Multimodal Report Revealed Across My Small Business Sites

On 24 September 2026, a blue banner appeared in Search Console offering something I’d been waiting for: a way to separate image-based search traffic from text queries. I immediately checked four sites I manage—Adcraft, The Content Agency, Made by Krystian, and Dubai Boxing Coach—to see which pages were being discovered through visual search rather than keywords. The reporting shows which URLs receive multimodal traffic, but it stops there: you can’t see the actual visual queries that brought visitors to those pages. That limitation matters less than you’d expect, because knowing which content surfaces through image search changes how you think about screenshots, portfolio pieces, and even product documentation. The question is no longer just whether your site ranks for the right terms, but whether someone searching with an image finds you when they don’t yet have the words.

The New Visibility: Image-Based Discovery Is Finally Easier to Measure

On 24 September 2026, I noticed a blue banner in Search Console that gave me a new way to isolate pages receiving traffic from multimodal search. I checked four very different sites I work on, Adcraft, The Content Agency, Made by Krystian and Dubai Boxing Coach, to see where visual discovery might surface across a software product, video led production business, creative portfolio and local service.

In practical terms, multimodal discovery begins without a conventional typed keyword. Someone can search with a phone camera, upload an image, circle an item on an Android screen or use Chrome’s “Search this image” function. Search Console now separates standard text based Web traffic from Web results where an image formed part of the search, using a multimodal search type filter in Performance reports, as Google announced. A restaurant menu, a decorator’s before and after gallery, a product catalogue, a videographer’s showreel and a studio’s location photography can all answer a visual prompt more effectively than a keyword targeted landing page alone. For independent creatives in particular, people often cannot name the style they want. They may start with an image of a lighting setup or an aesthetic reference, then work backwards towards a supplier.

The commercial stakes are not theoretical: Google says Lens handles more than 20 billion visual searches each month, and 1 in 4 Lens visual searches has commercial intent, according to Think with Google. For founders, agencies and SEO teams, a page level view means visual led visits are no longer buried inside total organic performance and mistaken for a ranking change in typed search. I can now ask a more useful question: which pages are being reached when the starting point is visual? A Made by Krystian portfolio image, for example, may lead a visitor to the relevant project page even if the visitor never formulated a conventional search phrase for that design approach.

This is not a separate image optimisation channel detached from SEO. The page still has to be relevant and accessible, while the visual asset has to communicate something useful. Google’s image guidance is clear that crawlable HTML <img> elements matter; images used only as CSS backgrounds are not indexed in the same way, as Google explains.

My initial view is cautiously positive. Search is becoming more visual, and the question is no longer only whether a site appears when someone searches for the business by name or service. It is whether it can appear when someone sees something that reminds them of it. But the new visibility is only valuable if we remain precise about its limits: it identifies the pages receiving multimodal traffic, not the exact visual queries that produced it.

What I Could Learn From the Pages Receiving Multimodal Traffic

I started with the landing page list, not a theory about which products, photographs or hero images ought to be winning. The useful question was simply: which URLs are Google actually sending this traffic to? In Search Console, I applied the multimodal search type filter in the Performance report, then reviewed the Pages tab and exported the list where I needed to sort and annotate it. Google describes this bucket as web results reached after an image formed part of the search, including image uploads and visual search actions in supported products (Google Search Central).

Classify the destinations before interpreting them

For a small site, I would label every receiving URL by business purpose:

• Product pages: individual items, packages or plans.
• Service pages: a specific treatment, installation, repair or professional offering.
• Gallery and portfolio pages: completed projects, before and after work or visual case studies.
• Articles: explanatory pages that answer a visual or practical question.
• Category pages: collections of products, styles, materials or services. If a kitchen installer sees portfolio projects and material category pages repeatedly appearing, that is a different signal from a retailer whose individual product pages dominate. Neither outcome is automatically better; it tells you where to look next.

I compared this approach across SaaS, creative services, content production and local businesses because the usual ecommerce only framing is too narrow. A SaaS landing page with a well presented interface may appear beside a photographer’s project page or a local tradesperson’s finished work gallery. Product screenshots are often treated as secondary page decoration, but I now regard them as potential discovery assets in their own right.

Put the list beside ordinary organic landing pages

Next, I compared the multimodal recipients with the pages already receiving ordinary web search visits. A page that performs in both views may be benefiting from an established topic and gaining another route into the same destination. A page that is prominent only in multimodal reporting is more interesting: it may be reaching people whose starting point is visual rather than keyword led.

I would not call either pattern a win without context. Impressions show how often URLs appeared in this reporting bucket, rather than proving that the content became stronger overall; clicks still combine query mix, ranking and result appeal (SEO Vendor). For a broader baseline, I also check the Image search type: it reports Google Images clicks, impressions, CTR and average position, which helps separate established image visibility from this newer reporting view (Advanced Web Ranking).

Read recurring page traits, not isolated URLs

A bespoke furniture maker, for example, might find that pages showing a completed oak dining table in situ, with dimensions, material details and a concise project description, recur more often than broad “our furniture” pages.

The crucial restraint is attribution. A product page receiving multimodal clicks shows that the destination was clicked after an image based search; it does not identify the particular image a searcher used (Search Engine Journal). I therefore treat the page as evidence of discovery at URL level, not as proof that its first gallery image produced every visit.

Finally, I would wait for recurring patterns across multiple pages and reporting periods before redesigning templates or shifting photography budgets. A handful of visits can be useful for investigation, but it is not enough to declare a visual strategy successful, or to decide that one image format, product type or page layout has won.

The Limitation That Changes How I Interpret the Report

Google withholds the Queries tab for this search type because multimodal searches primarily use images rather than text, so specific text query data is unavailable Search Engine Journal. A single URL can contain several images, answer several customer needs and be visually relevant in more than one way. The page is the measurable unit; the user’s visual interpretation is not.

Take a completed café fit out project page. It may show banquette seating, painted joinery, pendant lights, terrazzo flooring and the finished room as a whole. A user might reach it because they photographed a similar colour treatment, wanted ideas for the overall interior style, searched visually from a particular component, or began with a related product category. The multimodal report records the destination page, but it does not identify which of those routes applied.

What the data supports, and what it does not

I can conclude that the café project page participates in Google’s measured web multimodal search appearance. I can compare its clicks, impressions and CTR with its ordinary text based web performance, then look for a pattern across similar project pages.

The report is a scoped Search Console bucket rather than a full account of every image, video or mixed format discovery route a page may earn SEO Vendor. - Optimising only the most visually striking asset on the page while neglecting the page copy, image set, project details and related service or product context.

Both moves confuse a useful signal with a confirmed query match. Alt text should still describe the image accurately and serve its accessibility purpose; it should not become a speculative bid for an unseen phrase.

The API makes the boundary clearer

I also tested the Search Console API. At present, it does not support multimodal as a separate search type, even though the interface exposes the view.

For small teams, that limits automated reporting. I would not build a dashboard that silently treats API web data as multimodal performance, nor would I expect a scheduled extraction to provide the missing query layer. The interface is currently the place to inspect the distinct reporting slice.

I see this as an attribution problem, not a reason to disregard the report. The landing page list remains useful precisely when I triangulate it: compare the URL against text based search performance, inspect the full on page image and content context, and check whether user behaviour supports the commercial purpose of that page. It does not turn an unknown visual prompt into a keyword.

My Practical Workflow for Turning Page Signals Into Decisions

I keep this deliberately lightweight. The report gives me a shortlist for investigation, not a verdict on why a page appeared or what a visitor searched with.

1. Export the receiving URLs and classify their job.
2. Audit each page as a landing experience, not as an isolated image.
3. Compare it with a similar page that received no multimodal traffic.
4. Check downstream behaviour in site analytics.
5. Write one hypothesis, make one substantial change, and log the result.

1. Build a page level working sheet

For every URL receiving multimodal traffic, I record its business purpose, primary offer, image types, visible subject matter and intended customer action.

I then sort by practical importance, not just impressions. A service page for a high value kitchen installation deserves earlier investigation than a low intent blog post, even if the article has more visibility. Page level reporting is useful here because it shows which product and category pages earn image impressions, and which comparable pages earn none despite containing imagery, without pretending to offer complete attribution (Advanced Web Ranking).

2. Audit the whole destination

I review the page from the perspective of someone who arrived because something visual caught their attention. Does the lead image show the same subject named in the H1? Does the surrounding copy explain the material, service, style or product shown? Are dimensions, availability, specifications, location or process details present where the visitor would expect them? Is the call to action the logical next move?

For example, if a joinery firm’s walnut shelving project page receives multimodal traffic, I would check whether the page makes clear that it is bespoke walnut shelving, shows enough useful angles, explains the installation, and offers a relevant route to request a quote. An impressive image paired with vague “quality craftsmanship” copy is a weak landing experience.

3. Use nonreceiving pages as the control group

Next, I compare each priority URL with a close equivalent that has not appeared in the report. I look for visible differences: original photography versus stock or generic imagery; whether the subject dominates the page; image quality; clarity of the page’s topic; and the depth of supporting information.

This comparison helps me avoid a common mistake: treating every successful page trait as causal. If all the receiving pages happen to have larger images, that is a lead to test, not proof that image size created the outcome. I check whether the apparent pattern holds over a wider date range and agrees with other Search Console views before acting on it (SEO Vendor).

4. Check whether the visit is commercially useful

I join the landing page list to my existing analytics and examine what visitors do next: view another relevant page, submit an enquiry, add an item to a basket, begin checkout, download a specification or take another meaningful action.

I do not claim that multimodal reporting proves conversion causation. It can identify an entry context and a destination; it cannot isolate the visitor’s motivation from every other factor on the page.

5. Test one decision at a time

For each priority page, I write a single working hypothesis: “A clearer project description may help visually led visitors assess this service,” or “Replacing an ambiguous room shot with a product focused lead image may make the offer easier to understand.” For ecommerce pages, I may also test Product markup alongside a category or product page update; Google notes that product structured data can make details such as price, availability and shipping information eligible to appear in Google Images and Google Lens (Google Search Central).

Finally, I keep a decision log with three labels: measured fact, working hypothesis and test in progress.

The On-Page Foundations Multimodal Reporting Still Cannot Replace

Better measurement does not make the landing page basics optional. An image should show the product, service, project or subject that the page actually promises, not fill a blank space in a template. A dramatic interior shot may earn attention, but it is a weak commercial asset if the page cannot establish what was supplied, who it is for or how a prospect can engage the business.

I treat alt text first as accessible text. It should describe what is in the image in the context of that page: “Bespoke walnut media unit installed in a family living room” is useful when the page is about fitted furniture. A pasted string of hoped for terms is not. Alt text gives screen readers and search systems context about the image and its relationship to surrounding copy; Yoast’s guidance is to make it descriptive, relevant and information rich rather than repetitive Yoast.

Give each image corroborating context

The filename, caption where it genuinely adds information, nearby heading and body copy should all support the same straightforward reading of the visual. If the photograph depicts a kitchen renovation, the page should not surround it with vague claims about “quality craftsmanship.” State the service, the material or scope where relevant, and the customer problem it solved.

A service provider’s project gallery is where this often breaks down. A gallery of before and after bathroom work needs more than attractive tiled room photographs. Each image or group needs enough context to identify the service, such as a full bathroom refurbishment, the visible outcome, and the type of need it addresses, such as improving an inaccessible or dated family bathroom.

A beautiful image without meaningful context can impress a visitor while communicating very little about its purpose to a search engine. Use standard HTML <img> elements for meaningful site images, too: Google says it can find images through an <img> element’s src attribute, while CSS background images are not indexed Google Search Central.

Original evidence beats generic decoration

It provides clearer evidence of what a customer can expect than generic stock imagery. The image still needs a technically accessible implementation, relevant textual context and a page that fulfils its promise.

There is also a practical trade off. Rich photography can be central to a convincing page, but oversized or poorly handled files make that page slower and less usable. I would rather publish a considered set of strong, properly delivered images than let a gallery delay the moment when a visitor can read key details, compare options or contact the business.

For ecommerce, keep the visual and commercial data aligned with Product markup as noted in the workflow section.

Getting discovered through an image is only the opening move. The postclick page must answer the visitor’s likely next question, what is this, is it right for me, what does it involve, and what happens next, and make the appropriate action obvious: buy, book, request a quote or view related work.

How I Would Report Multimodal Performance Without Overclaiming

I would frame the report as a page level discovery signal, not a list of visual keyword wins. Google’s multimodal filter covers web searches in which an image formed part of the search, including Lens, Circle to Search, image uploads and Chrome’s “Search this image” feature, but it does not separate those entry points for reporting purposes Google Search Central.

The language matters. I use “pages associated with multimodal discovery”, not “this image ranked for” a query or “customers found us through walnut shelving searches.” The report has no query dimension for multimodal traffic, so naming the visual query would turn an informed inference into an unsupported claim Search Engine Journal.

My client facing reporting view

I would compare like with like: equivalent date ranges, the same property and the same reporting filters. Then I would group the receiving URLs by what they are there to do. A product page with clear purchasing intent deserves a different reading from a portfolio image, a service page or an editorial guide. If multimodal clicks rise on product pages that also generate add to basket activity, that is more commercially interesting than a rise concentrated in a gallery that receives brief visits and no onward movement. Discovery is not revenue. A page can gain visibility because its visual subject is relevant, while still failing to produce value because the offer is vague, the price is unclear, proof is thin or the route to enquire or buy is awkward. I would report those as separate findings: “multimodal discovery increased” and “commercial actions did not yet move.” They are not contradictory.

Pair the signal with the rest of the landing page evidence

For each page type, I would place multimodal clicks and impressions beside broader organic performance, engagement and the relevant action: enquiries for a service page, specification downloads for editorial content, or baskets and checkout starts for products. This prevents the new filter from becoming a vanity metric.

For example, a kitchen studio might see more multimodal activity on a bespoke island portfolio page while its product led appliance pages remain flat. I would describe that as an opportunity to examine whether the portfolio page better explains the service, shows project scope and directs a ready visitor towards a consultation, not evidence that a particular kitchen photograph “won” visual search.

I would also keep a change log beside the reporting dates: major photography replacements, content rewrites, template releases, internal link changes, pricing updates and tracking changes. A rise or fall may reflect query mix, classification or a broader traffic pattern rather than one optimisation, so the report is most useful for directional checks rather than claims about total visual search reach SEO Vendor.

For a creative agency, this creates a more credible client narrative: the work may be increasing the number of commercially relevant pages associated with multimodal discovery, but incomplete query and origin data means the next step is to validate landing page quality and business outcomes, not promise a named ranking, a Lens specific result or revenue that the data cannot isolate.

What I Still Want Google to Reveal About Visual Search Behaviour

The missing query dimension is understandable: an image led search does not reliably translate into a clean text string, and I am not asking Google to expose a searcher’s uploaded photograph or anything that could compromise privacy. Google’s own explanation is that specific text query data is unavailable because multimodal searches mostly use images rather than text, so the Queries tab disappears under this filter (Google Search Central). But there is useful ground between raw visual inputs and a page level click or impression. I would like aggregated insight into the likely visual intent behind a visit: was the searcher identifying an object, seeking a visually similar product, comparing a material, looking for design inspiration, or trying to find a local supplier? More importantly, I want to know the relationship between the landing page and the visual trigger. Did Google surface the page because of its primary image, a secondary gallery image, a product detail, a style reference in the copy, or a user supplied visual input that happened to resemble something on the page? If a furniture retailer’s category page earns multimodal traffic, the current report cannot tell me whether visitors arrived after searching a close up of fluted oak cabinetry, a whole room kitchen reference, or a particular cabinet handle. Those are materially different intents, and treating them as one “visual search opportunity” is how teams end up rebuilding a category template around a guess. Future reporting could offer privacy preserving, aggregated intent categories or page image relationship labels, primary product, secondary detail, room scene, style match, exact object match, without disclosing the underlying image or person. It would also help separate a page that wins because its hero product is clear from one that is being associated with a useful supporting image buried halfway down the page.

Until then, the absence of exact visual queries makes disciplined experimentation more important, not less. On my small business sites, I improve clarity and usefulness where the report points, then watch whether the relevant page level signal persists alongside meaningful visitor behaviour. I do not infer that a rise means people wanted a particular colour, style or product variant just because I can imagine a plausible visual prompt. My decision rule is simple: use multimodal traffic to prioritise page reviews and thoughtful tests, but do not base major content, design or inventory decisions on guessed visual queries alone.

Sources

1. Announcing web multimodal Search performance reporting in Search Console (developers.google.com)
2. How well do you know Google Search? (business.google.com)
3. Google image SEO best practices (developers.google.com)
4. Agency SEO October 17, 2026 by Elisa Murphy Google Web Multimodal Search Reporting: How to Read It (seovendor.co)
5. Measuring Visual Search (advancedwebranking.com)
6. Google Search Console Adds Multimodal Filter For Image-Based Searches (searchenginejournal.com)
7. Introduction to Product structured data (developers.google.com)
8. Image SEO: How to optimize alt and title text (yoast.com)