
As of November 2026, the Search Analytics API answers three questions it previously left to the Performance report: how much of a site's search traffic is branded, how much arrives through multimodal search, and how several countries or devices performed within one request. Google announced the addition on 16 November 2026. For a team already pulling Search Console data, the decision this week is whether to keep a hand-built brand list and a second request for Lens traffic, or move both into the calls that already run, and then settle what to verify before the numbers reach a client report. Anyone who has maintained a brand regex for years, the daily work of a seo consultant, will recognise that this is a reporting-definition question before it is a code question.
Google's Search Console team added three new options to the Search Analytics API, announced on the Search Central blog. All three already existed in the Performance report; the release exposes them to custom reporting and automated workflows rather than adding new measurement.
| What you want | Request field | What the reference allows |
|---|---|---|
| Branded and non-branded rows | brandedClassification, as a dimension or a filter, with values BRANDED or NON_BRANDED | equals only; cannot be combined with a query filter in the same request; only for eligible top-level properties |
| Multimodal search traffic | type: webMultimodal | Its own type value, covering Google Lens, Circle to Search on Android and image uploads; also promised in BigQuery for properties with an active bulk data export |
| Several countries or devices in one call | filterLists on country or device | Exact-match lists only; the same dimension cannot appear in both filters and filterLists in one request |
Two details matter more than the field names. First, branded classification is gated: data is available starting 21 February 2026, and only for top-level properties with sufficient query and impression volume. An ineligible property does not receive an error; it receives no rows, which in a dashboard reads exactly like zero branded traffic. Second, filterLists takes explicit values: three-letter country codes, or the device values DESKTOP, MOBILE and TABLET. A grouping such as EMEA or DACH is therefore a list you define and maintain, not a region the API already knows.

The Performance report is a fixed view over the same data the API serves. Its dimensions are chosen for it: queries, pages, countries, devices, search appearance, dates. The API lets a caller choose the dimensions, filters and date range, which is why the same three features had to be re-released rather than simply inherited. Google's stated reason was demand from the ecosystem: teams were stitching separate requests together, or maintaining keyword regex lists, to produce a brand split the report could already show.
The classification itself belongs to Google, not to the caller. The reference fixes the two values and allows only exact matching, so there is no parameter for supplying your own definition of a brand query. That design choice is the one a reporting lead has to accept or work around: the platform decides which queries are branded, and the API reports that decision.
A second, quieter gap is the window. The Performance report opens on the past three months, while an API request carries its own start and end dates. Two figures that look like the same metric can differ by window alone, before any question of branding arises. That gap is the first thing a seo consultant checks when a client asks why a dashboard moved.
The question most teams will ask this week is whether the three fields can feed a single dashboard. They can, if the calls are sequenced and the limits are written down.
brandedClassification alongside date gives one row per classification per day, which is what a chart needs. Leave the query filter out of that request; the reference refuses the combination.type: webMultimodal request returns Lens, Circle to Search and image-upload traffic. The reference does not state whether ordinary web rows already include it, so adding the two totals is unsafe until that is settled one way or the other.filterLists, one dimension in one place. Country and device lists belong in filterLists; putting the same dimension in filters as well is rejected. Define your regional groupings once, in code, and reuse them.startRow to page, and expect summed page rows to fall short of property totals on a large site.The limits are not footnotes; each one produces a specific wrong number.

The change touches four people in a typical marketing team, and each of them differently.
An in-house marketing manager running a single Domain property faces the eligibility question first. A smaller Singapore site may fall below the query and impression volume the branded classification requires, and the empty response will look like a data outage unless someone has tested it in advance.
An analyst inside a seo marketing agency mostly gains throughput. Country and device lists collapse several per-market requests into one, which shortens refresh cycles for multi-client dashboards. The catch is that eligibility varies by property, so the dashboard needs a distinct null state: "not eligible" rather than zero.
An SME selling into Singapore, Malaysia and Hong Kong gains a per-market branded share it could not compute from one request before. Grouping by country and classification together produces that view, provided the market lists are maintained deliberately rather than improvised per report.
An ecommerce or retail team gains visibility of image-led discovery: traffic from Lens and screenshot searches arrives as its own row. It remains a separate layer, not a bonus added to organic web clicks, because the overlap question is still open.
Two related reads for a Singapore team sit alongside this one: AI search behavior in Singapore: what it means for marketing covers where query behaviour is heading, and What Google's Q3 results mean for your SEM budget covers the commercial side of the same platform shifts.
This is the checklist to run this week, before anyone changes a client dashboard.
brandedClassification and record whether rows come back. Write the result into the property register as eligible or not.
Topkee's SEO services and reporting pages describe a structure the three fields happen to fit. Brand terms are listed separately from terms already ranking and from new non-brand targets, rather than blended into one traffic figure. AI answers, ads, maps and organic results each get their own layer, and tracking conditions are fixed by engine, region, language and device before any comparison is made. Those are groupings, not new metrics, and the knowledge base records no integration of these API fields; what it records is why the grouping matters in the first place.
The same notes hold the caution worth repeating. Search Console's generative-AI performance report supplements impression data, and its official scope is AI Overviews and AI Mode impressions only, so it cannot be extended to full prompts, internal fan-out or standalone conversions. Average position does not replace a fixed-condition rank. A multimodal row, read with the same discipline, is another layer to keep apart rather than a better number. Topkee's own SEO API documentation goes further and leaves the organic counting rule of the position field, and its handling of missing values, at engineering confirmation, and the same standard applies to branded totals before anyone writes them into a contract.
Three questions are open, and each decides how a report should be built, which is a fair briefing for a seo marketing agency starting a client account this month.
The first is whether web totals already contain multimodal traffic. Google's announcement and the reference are both silent, and the answer changes whether the two rows can be added. Until it is published, test against the report's own split between text-based and multimodal web results.
The second is how far back multimodal data goes. No start date appears in either document, and the BigQuery availability is promised without a date, so a year-on-year multimodal comparison is premature.
The third is the classification itself. Since the platform assigns the brand label, a shift in branded share could be a change in classification rather than a change in customer behaviour. A monthly three-line check covers it: rows returned against rows expected, branded share against your own list, and multimodal total against the report's multimodal figure. That is the check a seo marketing agency should put in front of a client before the trend line is discussed.

Article Author