On 24 September 2026, Google announced on the Search Central blog that Search Console now reports web multimodal Search performance, splitting the single "Web" search type that SEOs have relied on for more than a decade into two columns: Text-based and Multimodal. Search Engine Roundtable logged the staged rollout at 7:09 that morning. The number worth pausing on is not the filter itself but the volume behind it: by Google's own accounting, Lens already handles "nearly 20 billion visual searches every month." A reporting dimension twenty years old was quietly cut in half.
This did not appear out of nowhere. For the past 18 months Google's performance data has been retreating toward "you can see the volume, not the meaning." The Search Generative AI report that launched in June 2026 says so plainly in the documentation: it exposes impressions only for AI Overviews and AI Mode, with no clicks, no query strings and no position. Three months later, multimodal is missing a piece in exactly the same way. Google's explanation is that multimodal searches mostly use images rather than text, so specific text query data is not available. Look at the entry points and the logic is obvious: Circle to Search passed 300 million Android devices in July 2025, and in February 2026 Google said people had already run billions of queries a month through it. Search input moved from the box to the camera and the long-press gesture long ago. The report is just catching up.
The comparison set makes the direction clearer. Bing Webmaster Tools still has no equivalent visual-search dimension. Third-party vendors are patching the hole from the other end: Ahrefs bundles Brand Radar into every standard plan, with Lite at $129/month for 5 prompts and 150 checks and a standalone product from $199/month; Semrush folds AI Visibility straight into its $139/month SEO plan. The category is not moving toward more keywords. It is moving toward more proxy metrics for exposure, and everyone is selling you an approximation of the part you cannot see, while the native free data keeps getting thinner.
For small and medium businesses two things now happen at once: your product photography finally gets its own performance column, and you will never know what the user actually pointed the camera at. This article is about that contradiction, and how to turn multimodal traffic into a decision using free data only.
What Google Actually Split
Open the Search type filter in the Performance report and Web now expands into two children. Google's definition: Web-multimodal tracks results triggered by a query that uses an image, photo or screenshot; text-only queries are counted as Web text-based. It is not the same as the existing Image search type. Image counts your performance inside Google Images; multimodal counts traffic where an image was the input but the landing surface was an ordinary blue link.
Per Search Engine Journal's breakdown, four entry points feed the bucket: Google Lens, Circle to Search on Android, uploading an image to Google Search, and Chrome's right-click "Search this image." The rollout is still staged.
- Available: impressions, clicks, CTR, average position, plus page, country, device and date dimensions.
- Not available: query strings. Select multimodal and the queries table is empty.
- Unclear: whether the split is applied retroactively, and whether the two columns sum to the old Web total. Google has not said.
The winners are sites whose products, packaging and signage photograph cleanly, which can now see at page level which URLs were found by camera. Long-tail text publishers will mostly discover the column is close to zero. Search Console is becoming a bigger table with emptier cells.
What Three Audiences Should Do Now
Brand owners and SMB operators
- Check whether you have the filter at all: Performance report, Search type, see whether Web expands.
- Divide multimodal clicks by total clicks over the last 90 days to get your visual traffic share. Below 1%, do not move budget. Above 5%, your physical products are already being photographed.
- Pull the top 10 pages by multimodal clicks. Those are your genuinely recognisable assets; update their images and pricing first.
Marketers and SEOs
- Establish a dual baseline for text-based and multimodal so any future ranking turbulence can be attributed separately.
- With no query data, work backwards from pages: whichever product category concentrates the clicks is the category with high real-world recognisability.
- Treat filenames, alt text, surrounding copy and ImageObject structured data as your substitute for keywords. It is the only channel left for speaking to the model.
Developers and agencies
- Add the new searchType values to a daily Search Console API pull rather than exporting from the UI.
- Build clients an "image asset by multimodal performance" crosswalk. That is a billable reporting module as it stands.
- Audit whether the CDN renames images and whether an indexable full-size original exists. A product shot compressed to 400px has almost no chance in visual matching.
Measurement Tool Comparison
| Tool | Sees multimodal? | Price | Best for |
|---|---|---|---|
| Google Search Console | Yes, page level, no queries | Free | The only first-party data there is |
| Bing Webmaster Tools | No equivalent dimension | Free | Bing and Copilot exposure, excludes Lens |
| Ahrefs Brand Radar | No, tracks AI prompts | Lite from $129/mo; standalone from $199/mo | Brand mention rate inside generative answers |
| Semrush AI Visibility | No, prompt and sentiment monitoring | SEO plan from $139/mo | One-stop AI share-of-voice view |
No third-party tool can fill the multimodal query gap, because Google is not retaining that data in structured form either. Paid tools solve a different problem. Do not conflate the two and pay twice.
What Nobody Will Tell You
- A new dimension is not new traffic. These clicks were always counted inside Web; they are simply labelled now. Anyone expecting to open the filter and find undiscovered traffic will be disappointed. This is classification, not growth.
- Performance data without queries is trivially over-read. You know a page was reached via image search. You do not know whether the user photographed your product, a competitor's, or a receipt with your logo on it. Treating that signal as demand evidence and changing your product line on it is riskier than it looks.
The No-Subscription Alternative
- ☐ Export split searchType data daily via the Search Console API into Sheets or SQLite and build your own history (Google keeps only 16 months).
- ☐ Build an image-to-page-to-clicks crosswalk by crawling your own img tags. No paid crawler required.
- ☐ Give your top 50 product images descriptive filenames, alt text and an ImageObject caption.
- ☐ Remove CDN assets that exist only as thumbnails; keep one indexable original at 1,200px on the long edge or larger.
- ☐ Review the share trend monthly using a three-month moving average.
FAQ
How is multimodal different from the existing Image search type?
Image measures how your images perform inside Google Images. Multimodal measures traffic where an image was the input but the result was an ordinary web page. Both can show numbers; do not add them together.
Why is the queries table empty when I select multimodal?
Google says multimodal searches mostly use images rather than text, so there is no corresponding text query data. This is not a bug, and it is unlikely to be filled in soon.
My site has no product photography. Does this matter to me?
It depends on your physical touchpoints. Signage, menus, equipment plates and packaging can all be photographed through Lens. A pure content site can simply record a baseline.
Does this filter affect rankings?
No. It is a reporting-layer change, not a ranking signal. It does change attribution, and many sites will have to revise their estimate of how much text SEO they still have.
My Take
The mainstream reading is that Google handed SEOs a new dimension. My read is the opposite: this is the second milestone in Search Console's demotion from diagnostic tool to exposure dashboard. The first was June's generative AI report offering impressions only; now multimodal offers pages only. At this pace I expect that within 12 to 18 months, the share of Search Console traffic that carries a query string will fall below half. Not because Google is hiding data, but because a growing share of searches never produce text at all. Keyword research will not die; it will degrade into a method that explains half a world.
For an agency like ScriptWalker there is a clean productisation here: a visual asset recognisability audit. Image inventory, original-size and indexability checks, ImageObject completion, plus a dual-track Search Console API history. One project fee plus a monthly report, and the client does not have to buy another subscription starting at $129 a month, with every deliverable living on their own site.
To discuss applying this audit to your site, get in touch:
- Email：[email protected]
- 電話：0916-224-047
- LINE：@ufv9089p
Sources
- Google Search Central Blog — Announcing web multimodal Search performance reporting in Search Console
- Google Search Central — What's new in Search documentation
- Search Console Help — Generative AI performance report (Search)
- Google Blog — AI-Organized Search, Google Lens and more
- Google Blog — Bringing AI Mode to Circle to Search
- Google Blog — Circle to Search multi-object update
- Search Engine Roundtable — Performance Report Gains Multimodal Search Type Filter
- Search Engine Journal — Multimodal Filter For Image-Based Searches
- Search Engine Land — Performance reports gets multimodal filter
- Ahrefs Pricing, Semrush Pricing