This checklist shows how to pull branded versus non-branded and visual-search numbers through the Search Analytics API and file them in a weekly report without a regex list. Google added those filters to the query method on November 16, 2026, after several Search Console updates over the past year. Eligibility and search type are the checks that keep that file honest.
Updated November 16, 2026
Key Takeaways
- Google added branded, multimodal, and multi-filter options to the Search Analytics API on November 16, 2026.
- Branded classification returns rows only for eligible properties.
- Set type to webMultimodal for Lens, Circle to Search, and image uploads.
- Use filterLists to query several countries or devices in one call.
What should you check before calling the Search Analytics API?
Google’s November 16 post adds branded classification, web multimodal reporting, and multi-country and multi-device filters for custom reporting and automated workflows. Branded rows come back only for eligible properties, as in Search Console. Any other property returns no rows.
Google said the console filter is only for top-level properties, not URL-path properties such as https://example.com/path or subdomain properties such as developers.google.com, and only for sites with sufficient queries and impressions. The query method is an authorized POST to searchAnalytics/query. Set a startDate and endDate of one or more days.
Rows carry clicks, impressions, ctr, and position, sorted by clicks descending, or by date ascending if date is a dimension. Days without data are omitted, and the API returns top rows, not every row.
How do you split branded and non-branded traffic?
Group or filter by brandedClassification instead of maintaining a keyword regex list. A branded query includes the brand name, misspellings, and brand-related products or services. Google’s examples are Google, Gogle, and Gmail. The branded queries announcement says an internal, AI-assisted system does the classification, covering the brand in all languages, typos, and queries that omit the name but refer to a unique product or service.
Some queries may be misidentified, and the split does not change ranking. Branded queries typically lead to higher-ranking pages and higher click-through rates. Non-branded queries show how new users find the site without intending to visit it. Read an empty response as ineligibility, not as zero brand demand.
How do you pull visual-search numbers?
Set type to webMultimodal for performance from Google Lens, Circle to Search on Android, and image uploads in Google Search. Properties with an active bulk data export can also read that multimodal data in BigQuery. Search Console’s September 24, 2026 filter also includes the Chrome right-click "Search this image" feature.
Search Engine Roundtable quoted Google’s help text on why text queries are missing:
Because multimodal searches mostly use images rather than text, specific text query data isn’t available for this traffic. As a result, the queries dimension isn’t available when this search type is selected.
John Mueller, after checking with the team, replied on Bluesky that "the data wasn’t previously in the counts, the report has new data." Older web totals are a separate series.
How do you combine markets in one weekly pull?
Use filterLists to filter several countries or devices in one request. In the query body it sits inside dimensionFilterGroups, with a dimension and an expressions array. Google describes one call for a group of countries, such as EMEA, LATAM, or DACH, instead of stitched requests.
The filters work together, so the same pull can cover brand growth in a regional hub or which pages perform best in multimodal search across markets and devices. Supported values are in the query documentation.
Schedule the steps below and keep the rows with the rest of your search measurement, including what a ChatGPT search console would tell the business.
| Need | What to set | What comes back | Limit |
|---|---|---|---|
| Branded vs non-branded | brandedClassification | Brand and non-brand performance | Eligible properties only; otherwise no rows |
| Visual search | type set to webMultimodal | Lens, Circle to Search, image uploads | No queries dimension in the Search Console view |
| Several countries or devices | filterLists | One call instead of stitched requests | Top rows only, not every row |
- Confirm the property is eligible before you treat a branded response as complete.
- Send an authorized POST to searchAnalytics/query with a startDate and an endDate.
- Group or filter by brandedClassification for the brand split.
- Set type to webMultimodal for Lens, Circle to Search, and image uploads.
- Add filterLists for the countries or devices that belong in the report.
- Store clicks, impressions, ctr, and position, and expect top rows sorted by clicks unless you group by date.
Frequently asked questions
Which properties return branded rows?
Only eligible properties. Other properties return no rows. In Search Console, Google limited the branded filter to top-level properties with enough queries and impressions, not URL-path or subdomain properties.
How do you request visual-search performance?
Set type to webMultimodal. Google says that returns performance from Google Lens, Circle to Search on Android, and image uploads in Google Search.
Are text queries available for multimodal search?
Not in the Search Console view. Google’s help text says those searches mostly use images, so specific text query data and the queries dimension are not available.
Does branded classification change rankings?
No. Google says the classification has no effect on ranking, and some queries may occasionally be misidentified.