On November 16, 2026, Google added three things to the Search Analytics API that until now only the Search Console interface had: a split between branded and non-branded queries, data from searches made with images, and a filter for several countries or devices in a single request. Reading reports by hand, you’ll barely notice, but in dashboards, Looker Studio reports and your own scripts you can finally drop the lists of brand-name keywords and the loops that queried the API once per country.
What Google announced
The post on the Google Search Central blog is signed by Daniel Waisberg, Nati Yosephian and Karin Gadasi from the Search Console team. Each feature already worked in the Performance report in the interface. According to the authors, they came to the API because tool makers and people who automate their reporting asked for them. Search Engine Roundtable covered the news the same day.
| Feature | How to use it in the API | Main limitation |
|---|---|---|
| Branded queries | group or filter by brandedClassification (values BRANDED and NON_BRANDED) | eligible properties only; others get a response with no rows |
| Multimodal search | set type to webMultimodal | covers Google Lens, Circle to Search on Android and image uploads to Search |
| Several countries and devices | the new filterLists field | you can’t filter the same dimension through both filters and filterLists |
The changes apply to the searchanalytics.query method, and its documentation adds details the announcement leaves out.
Branded queries: brandedClassification
The branded queries filter arrived in the Search Console interface in November 2025, and since March 2026 Google has offered it to every eligible site. Google counts a query as branded when it contains the brand name, its variants or misspellings, or when it refers to a product or service unique to the site, even without the name. An internal, AI-assisted system makes the split and recognizes the brand name across languages, so no regex is involved. Google warns that some queries may land in the wrong group and that the filter has no effect on ranking.
Until now, splitting traffic automatically meant keeping your own keyword list or a regex on the query dimension. Such a list goes stale, misses misspellings and can’t see product queries that don’t mention the brand. Now the API returns both groups itself, and you can compare them in one report.
The documentation adds conditions you need to know before building a report:
- the
brandedClassificationfilter supports only theequalsoperator, - grouping or filtering by this dimension can’t be combined with a
queryfilter in the same request, - data is available from February 21, 2026,
- it works only for top-level properties (not for path properties such as
example.com/blog/, nor for subdomains) with enough query and impression volume.
That last point matters in practice, because a property that doesn’t qualify gets an empty response instead of an error. Your script or data warehouse connector has to recognize that case: no rows doesn’t mean zero brand clicks, and the report shouldn’t break on it.
The biggest gain is an honest view of SEO results. Branded queries usually have high positions and CTR, so on a single clicks line they can hide flat non-branded traffic, which is the part that shows whether new people are finding you. Two separate series on the chart answer that better. They also show the effect of brand campaigns, because growth in brand searches no longer disappears into the rest of the traffic.
Google’s model now does the classification, so you save work but depend on someone else’s algorithm. If your brand name is an ordinary word, check the results critically.
Image search: webMultimodal
The multimodal search report reached the Search Console interface on September 24, 2026. It covers searches through Google Lens, Circle to Search on Android and image uploads to Search. The September announcement also listed the “Search this image” option in the Chrome context menu. Now you can pull the same data from the API by setting type to webMultimodal.
For properties with an active bulk data export, multimodal data is also meant to reach BigQuery. For larger sites that’s a more convenient route than the API, which returns at most 25,000 rows per request, and only the top ones, with no guarantee of completeness.
Image search used to get lost in the overall numbers, so I’d check how much traffic comes this way, especially for a shop selling products people photograph with their phones, a recipe site or a plant care guide. Empty data means there’s nothing to optimize. When the numbers grow, review image quality, alt text and the pages that show up most often in this channel, but wait before drawing conclusions about trends, since the segment has a short history.
I covered the other AI search reports Google was preparing for Search Console in my post on AI Search reports in Search Console.
filterLists: several countries and devices in one request
The filterLists field sits inside dimensionFilterGroups, next to the regular filters. A regular filter compares a dimension with one value, while filterLists takes a list of exact values for one dimension. A row matches if its value equals any of them. The documentation lists two dimensions: country (three-letter ISO 3166-1 alpha-3 codes) and device (DESKTOP, MOBILE, TABLET).
I wrote the example below myself from the schema in the documentation. It doesn’t come from Google, and I haven’t run it against the API. It asks for the branded and non-branded split of mobile traffic in Poland, Germany and the Czech Republic:
{
"startDate": "2026-10-01",
"endDate": "2026-10-31",
"type": "web",
"dimensions": ["brandedClassification"],
"dimensionFilterGroups": [
{
"groupType": "and",
"filterLists": [
{ "dimension": "country", "expressions": ["POL", "DEU", "CZE"] },
{ "dimension": "device", "expressions": ["MOBILE"] }
]
}
]
}
Before, a scope like this took several separate requests whose results you then had to combine yourself. That’s tricky, because you can’t simply add up average position and CTR. Now Google calculates them for the whole group, and the code gets simpler and stays further from the API limits.
Google gives the EMEA, LATAM and DACH market groups as examples. A single request can’t filter the same dimension through both filters and filterLists, though.
Combining the new filters
Google stresses that the filters work together. That gives you reports such as:
- Branded queries from a group of countries, for example DACH, split by desktop and mobile, show whether awareness grows where you’re entering a new market.
- Image search split by market and device shows which pages collect traffic from Lens and Circle to Search in the countries you choose.
- Non-branded traffic from phones covers visits you earn without the brand name, on the devices that bring most of the traffic for many sites.
When combining them, keep in mind that the branded filter doesn’t work with a query filter. You can’t build “branded queries containing a given phrase” in one request.
Unknowns and pitfalls
- Threshold for branded data. Google doesn’t publish the query and impression threshold. The only known fact is that a property below it gets an empty result.
- Classification accuracy. Google’s model makes the brand split and can get it wrong. Treat absolute values as approximate; trends are more reliable than single numbers.
- Data history. Branded data starts on February 21, 2026, so a full year-over-year comparison won’t be possible before February 2027.
- Share of image search. Google doesn’t say how much traffic multimodal search accounts for. Only your own data will show whether it matters for your site.
- Third-party tools. The announcement doesn’t say when the new fields will reach connectors (including the Looker Studio one) and SEO tools built on the API.
Summary
You already know this data from the interface; now you can pull it automatically. The people who gain most report across many markets and brands, or have been maintaining their own lists of branded keywords. If you use the API, check the new values in the documentation, handle empty responses and start with one simple report: branded versus non-branded traffic over the last few months. For the basics of the tool, see my guide to Google Search Console.
Frequently Asked Questions
What is new in the Search Analytics API?
Google added three options: grouping and filtering by brandedClassification (branded and non-branded queries), the webMultimodal search type and the filterLists field, which filters by several countries or devices in one request. All three already worked in the Search Console interface.
Why does the API return an empty response for the branded filter?
Branded query data is available only for eligible properties: top-level properties with enough query and impression volume. Path properties, subdomains and small sites get a response with no rows. The documentation describes this as normal API behavior, so your request is fine.
How do I get Google Lens and Circle to Search data from the API?
Set type to webMultimodal in your request. The data covers Google Lens, Circle to Search on Android and image uploads to Google Search. For properties with an active bulk data export it is also meant to reach BigQuery.
Can I combine the branded filter with a query filter?
No. Filtering or grouping by brandedClassification cannot be combined with a query filter in the same request. You can combine it with countries, devices and the search type.
Sources
- New filtering and reporting capabilities in the Search Analytics API
- Search Analytics: query
- Introducing the branded queries filter in Search Console
- Announcing web multimodal Search performance reporting in Search Console
- Google Search Analytics API Adds Branded Queries, Multimodal & Country/Devices Filters
Comments