Google's Search Console API adds branded and multi-country filters
Summary
The Search Analytics API can now return the branded, multimodal and multi-country breakdowns that only Search Console's Performance report could show.
Rewrite any extract that loops one request per country or device so it uses the new filterLists field instead.
Google added three of Search Console’s Performance report capabilities to the Search Analytics API on 16 November, in a Search Central post written by Daniel Waisberg, Nati Yosephian and Karin Gadasi of the Search Console team. The API now takes brandedClassification as a dimension, webMultimodal as a type, and a filterLists field that accepts several countries or devices in one request.
filterLists is the addition that changes existing code. It sits inside dimensionFilterGroups next to the familiar filters array, and each entry pairs a dimension with a list of expressions, so a DACH or EMEA rollup becomes a single query. The query reference allows it on country and device only, matches each expression exactly, and says one request cannot filter the same dimension in both filters and filterLists. The Performance report itself gained the multi-country filter in the past year, and the API is catching up.
Any extract that loops a request per market or per device is now the expensive way to get the same rows. Each pass spends quota, and the script then has to add the rows back together and recompute CTR and average position by hand, which is where country rollups usually go wrong. One filterLists entry removes both problems.
brandedClassification kills a different workaround, the regex that split branded from non-branded traffic by matching brand spellings against the query dimension. It comes with limits. Google’s post restricts the dimension to eligible properties and says other properties get a response with no rows in it. The query reference adds that data starts on 21 February 2026, only for top-level properties with enough query and impression volume, and that the dimension cannot share a request with a query filter. A script that keeps its old brand regex in filters while adding the new dimension is not a valid request.
The empty response is the harder problem. The post mentions no error and no flag for an ineligible property, and a property with no traffic in the requested range returns an empty row set as well, so a dashboard cannot tell an ineligible property from a quiet week on its own. The safer build pairs the branded call with one unfiltered request for the same dates and reports eligibility as unknown when the unfiltered call has rows and the branded one does not.
webMultimodal is a type rather than a dimension, which means Google Lens, Circle to Search on Android and image uploads come back in their own request. Any comparison against web search needs two calls and a join on date. The post also makes the multimodal data available in BigQuery for properties with an active bulk data export.
What to do
- Rewrite country and device extracts around
filterLists, and delete the per-market loop and its stitching code. - Check that nothing in the same request still filters
countryordevicethroughfilters, since the docs rule that combination out. - Replace brand regexes with
brandedClassification, drop anyqueryfilter from that request, and treat an empty response as eligibility unknown until a control query shows the property returns rows at all. - Pull
webMultimodalas its own series before building comparisons. A rebuilt dashboard inherits whatever Search Console reported at the time, and Google has undercounted AI impressions in its own data before.