For years, one of the most useful splits in search reporting lived behind a manual process. If you wanted to know how much of your organic traffic came from people searching for your brand name versus people discovering you for the first time, you built a keyword list, wrote a regular expression, and hoped you had caught every spelling, abbreviation, and misspelling your customers use. Google has now moved that split into the Search Analytics API, and it changes what automated reporting can do.
The update, announced by Google on November 16, 2026, brings three features from the Search Console Performance report into the API: a branded query classification, multimodal search data, and filters that cover several countries or devices in a single request. All three were already available in the Performance report interface. What is new is that developers and tool makers can now build them directly into automated reporting systems instead of stitching together separate requests or maintaining long regex lists.
This matters most for teams that report on search every week. The branded versus non-branded split is one of the clearest signals of whether your brand is growing or whether you are simply riding demand that already existed. With it in the API, that split can finally live in dashboards, alerts, and client reports without a human copying numbers out of the interface.
What the Branded Classification Actually Gives You
The API now supports a dimension called brandedClassification. It returns one of two values for each query: BRANDED or NON_BRANDED. You can filter on it or group your results by it, which means a single API call can return your branded clicks and impressions alongside your non-branded ones.
There are limits worth knowing before you rebuild anything around it. The data starts on February 21, 2026. Anything before that date cannot be split, because Google was not tracking the classification yet. When a Search Console user asked Google’s John Mueller in March whether earlier data could be broken out retroactively, the answer was that there is a point when the data starts being tracked, and that point is February 21.
The dimension is also restricted to top-level properties, and it only returns rows when the property has enough queries and impressions to clear Google’s threshold. Smaller sites may get an empty table, which is Google’s way of saying the sample is too thin to classify reliably. The filter accepts only the equals operator, and you cannot combine it with a query filter in the same request.
One more detail deserves attention. Google determines whether a query is branded using an AI-assisted process, and Google itself has said that some queries may occasionally be misidentified. A product name that doubles as a common word, a brand that shares its name with a place, or a new product line that has not yet been learned by the system can all land on the wrong side of the split. If you already run a brand regex, keep it running alongside the API classification for a few weeks and compare the two. Where they disagree, investigate before you trust either one.
Multimodal Search Gets Its Own API Type
The second addition covers searches that start with an image rather than words. When someone points Google Lens at a product, uses Circle to Search on an Android phone, or uploads a photo to Google Search, that activity now has its own search type in the API: webMultimodal.
Multimodal reporting arrived in the Performance report on September 24, 2026, but the API reference did not list a separate type value for it at the time. The reference, last updated November 16, now does. That closes a gap that left multimodal traffic invisible or lumped in with ordinary web searches in automated reports.
Two caveats apply here. First, the API documentation does not say whether the regular web type already includes webMultimodal data, so do not add the two together until you have checked your API web numbers against the text and multimodal figures in the Performance report. Double counting is easy here and hard to spot later. Second, Google says multimodal data will also reach BigQuery for properties with an active bulk data export, but no date has been given. As of November 16, the bulk export table reference still lists six search types with no multimodal value. Chrome’s right-click “Search this image” feature is listed in the Performance report help page, but neither the API announcement nor the reference mentions it, so treat its coverage as unconfirmed for now.
Several Countries or Devices in One Request
The third addition is a new field called filterLists. It lets you filter results by multiple countries or multiple devices inside a single request. Google’s examples are country groupings such as EMEA, LATAM, or DACH, which are exactly the groupings regional teams actually report on.
Before this, pulling a regional view meant either running one request per country and combining the results yourself, or accepting a single global number. Both approaches cost time and introduced errors. A single request with filterLists returns the grouping directly.
The field is limited to the country and device dimensions, and it matches on exact values only. You also cannot use the same dimension in both filters and filterLists within one request, so plan your request structure before you start writing code.
Why This Matters for Your Reporting
Taken together, these three additions remove the two most common workarounds in Search Console reporting: hand-maintained brand keyword lists and per-country request stitching. Neither workaround was wrong, but both were fragile. A regex breaks the moment a brand launches a new product name. A stitched report breaks the moment someone forgets one country.
The practical effect is that a branded versus non-branded split can now sit in the same automated report as your other search metrics, updating on the same schedule, with the same definitions, every week. That makes the split useful in conversations it rarely reached before: budget discussions, brand campaign reviews, and quarterly planning. When brand search is climbing while non-brand search is flat, you are harvesting demand your brand work created. When non-brand search is climbing while brand search is flat, your discovery work is doing its job. When both are falling, you have a visibility problem, not a brand problem.
The multimodal type adds a second useful signal. Visual search is still a small share of traffic for most sites, but it is growing, and it behaves differently from text search. People searching with a photo are often further along in a buying decision. They have seen the product or something like it. If your site earns multimodal impressions, you want to see them separately, because blending them into web search hides both their volume and their conversion behavior.
What to Do This Week
Start with the branded split, because it is the feature most teams can use immediately. Ask your developer or your SEO tool vendor whether your reporting already pulls from the Search Analytics API, and whether the brandedClassification dimension is on their roadmap. If you pull the API yourself, run a test request for a recent week and compare the branded and non-branded totals against your existing brand regex. Expect small differences. Investigate large ones.
Next, check whether your property qualifies. Remember that the dimension needs a top-level property and enough query volume. If you get an empty table, that is an answer in itself: your branded volume is below Google’s threshold, and you should keep your regex approach for now.
Then look at your date ranges. Any report that spans February 21, 2026 needs a note explaining that the split only exists from that date onward. A chart that shows branded traffic starting from zero in January and jumping in February is not showing growth. It is showing the start of measurement.
Finally, add webMultimodal to your monitoring, even if the numbers are small today. Set up the request, record the baseline, and check it monthly. The sites that notice visual search early will be the ones whose images, product photos, and structured data are ready when that traffic grows.

FAQ
What is the brandedClassification dimension in the Search Console API?
It is a new dimension that labels each search query as BRANDED or NON_BRANDED, so you can filter or group your API results by whether the searcher used your brand name. It works on top-level properties and only returns data from February 21, 2026 onward.
Why does branded data start on February 21, 2026?
That is the date Google began tracking the branded classification. Data before that date was never classified, so the API cannot split it retroactively. Reports spanning that date should note the split begins there.
How accurate is Google’s branded classification?
Google uses an AI-assisted process and has said some queries may occasionally be misidentified. Run the API split alongside your existing brand regex for a few weeks and compare the two before relying on the API figures alone.
What is webMultimodal in the Search Analytics API?
It is a search type that covers searches started with an image, including Google Lens, Circle to Search on Android, and photo uploads to Google Search. It lets you measure visual search separately from ordinary text-based web search.
Can I add web and webMultimodal totals together?
Not yet without checking. The API reference does not say whether the web type already includes multimodal data, so compare your API web numbers against the Performance report before summing the two, or you risk double counting.
What does filterLists do in the Search Analytics API?
It lets you filter one request by several countries or several devices at once, so you can pull regional groupings like EMEA, LATAM, or DACH in a single call instead of running one request per country and combining the results yourself.