How to Pull Branded and Non-Branded SaaS Search Data From the Search Analytics API
Google added branded queries, web multimodal data and multi-value filters to the Search Analytics API on November 16. Here is how a B2B SaaS team can build the new reports.
- Desk
- Measurement
- Format
- How-to
- Reading time
- 7 min
- Published
- Nov 17, 2026
To split branded from non-branded traffic in the Search Analytics API, group or filter by brandedClassification, using the values BRANDED or NON_BRANDED. Data starts February 21, 2026 and covers only eligible top-level properties. You cannot combine it with a query filter in the same request. Use filterLists to pull several countries or devices at once.
This guide is for B2B SaaS marketers and analysts who build their own search dashboards. It explains the three options Google added on November 16, 2026, how to write the requests, and where the data misleads. cekpr covers measurement for software companies, and branded traffic is the first number to separate when you report on SEO.
What did Google add to the Search Analytics API?
Google added three options: a branded queries classification, a web multimodal search type, and multi-value country and device filters. They bring features already in the Search Console interface to the API. Teams can now automate reports that used to need manual exports or long keyword regex lists.
The change was announced in the Search Central blog. Google said it followed requests from the ecosystem for data usable in custom reporting and automated workflows. The three options work together, so one request can cover brand growth in a region or multimodal performance on a device type.
The official query method reference holds the exact field definitions. It lists the read-only scope, webmasters.readonly, as enough for these requests. A row limit between 1 and 25,000 applies, and the default is 1,000 rows. Use startRow to page through larger results.
How does the branded queries classification work?
Google classifies each query as branded or non-branded with an internal, AI-assisted system. A branded query includes your brand name, its variations and misspellings, and queries that point to a product or service unique to your site. Everything else counts as non-branded.
That definition comes from Google's branded queries filter announcement. It also warns that some queries may be misclassified, and that the filter has no effect on ranking. The split matters for SaaS because branded queries usually rank higher and earn higher click-through rates. Non-branded queries show how new buyers find you.
In the API, set brandedClassification as a dimension to group rows, or as a filter with the equals operator. The reference says data starts on February 21, 2026. It also says the data covers only top-level properties with enough query and impression volume.
How do you write a branded versus non-branded request?
Build the report in small steps. Each step adds one thing, so you can see which change moved the numbers. Use this order for a first dashboard feed:
- Confirm the property is a top-level domain property, not a path or subdomain property.
- Send a request grouped by date and brandedClassification for the last 90 days.
- Check that both BRANDED and NON_BRANDED rows come back. If neither does, the property is not eligible.
- Calculate branded clicks as a share of total clicks for each week.
- Send a second request filtered to NON_BRANDED and grouped by page.
- Join the page rows to your page-type labels, such as product, comparison, integration and blog.
Step three deserves attention. Google says that for properties without access, the response will not include any rows. An empty response is a missing feed, not a traffic collapse. Build an alert for an empty branded result so nobody reports a false zero.
Watch one hard limit. The reference states that filtering or grouping by brandedClassification cannot be combined with a query filter in the same request. You cannot ask for branded queries that contain the word pricing. Use the page dimension instead, or pull non-branded rows and filter queries in your own warehouse.
What does a branded share calculation look like?
Here is a worked example with made-up numbers. Suppose a SaaS site earns 4,000 clicks in a week, and the API labels 2,600 of them BRANDED. Branded share is 2,600 divided by 4,000, or 65 percent. The other 1,400 clicks are non-branded.
Now suppose the next week brings 3,700 clicks, with 2,600 branded and 1,100 non-branded. Total clicks fell 7.5 percent, which looks like a mild dip. Non-branded clicks fell 21 percent, and branded clicks did not move. That is the real story, and a single total hides it.
The split also protects good news. A launch campaign can lift branded clicks by a third while content traffic slides. Without the separate lines, the total would look healthy and the content problem would stay unseen.
How do filterLists help with multi-country SaaS reporting?
The filterLists field lets one filter match any of several exact values. A row matches if its value equals any of the expressions you supply. That replaces a loop of separate requests for each country or device, and it keeps your quota and your code simpler.
Country values use three-letter ISO 3166-1 alpha-3 codes. A DACH report, for example, would list DEU, AUT and CHE in a single country filter list. Device values are DESKTOP, MOBILE and TABLET. A request cannot filter on the same dimension in both filters and filterLists, so choose one form per dimension.
Pair this with branded data to see brand growth by region. A SaaS company that launched a German-language site can compare branded clicks in the three DACH markets with branded clicks elsewhere. If branded demand rises while non-branded clicks stay flat, the brand is growing faster than the content. The reverse points to a content strength that has not yet built recognition.
What does the webMultimodal search type report?
Set the type field to webMultimodal to get performance data from Google Lens, Circle to Search on Android and image uploads in Google Search. Google says the data is also available in BigQuery for properties with an active bulk data export. It is a separate search type, so it does not blend into web results.
The interface version launched on September 24, 2026, according to the multimodal reporting announcement. That post adds that the data includes the Chrome right-click Search this image feature. For most B2B SaaS sites, volume will be small. It can still reveal which product screenshots, diagrams and logos people search with.
Treat multimodal data as a discovery tool, not a headline metric. Track it monthly and look for pages that appear unexpectedly. At cekpr, we advise keeping it in a separate tab so that a low count never distorts the main web search view.
What should a SaaS search dashboard show now?
Show the numbers that tie to decisions. The table below proposes a short set of lines, each with the request that feeds it and the question it answers.
| Dashboard line | API request | Question answered |
|---|---|---|
| Branded click share | Group by date and brandedClassification | Is brand demand growing? |
| Non-branded clicks by page type | Filter NON_BRANDED, group by page | Which content earns new buyers? |
| Regional brand growth | filterLists on country, group by date | Where is the brand spreading? |
| Multimodal impressions | type set to webMultimodal | Which visuals get found? |
Set a review rhythm before you build the charts. Check branded share and non-branded clicks weekly. Look at regional and multimodal lines monthly. Spot-check the classification quarterly. Pick one alert rule too, such as non-branded clicks moving more than 15 percent for two weeks in a row. A written rule keeps the team from reacting to every one-day swing.
Connect the dashboard to your other Search Console work. Our guide to the generative AI performance report covers the AI side of the same property. For video and social data, see how to measure SaaS video and social content with platform properties.
The branded split also supports brand audits. When AI answers appear on branded searches, branded clicks may fall even as awareness holds steady. Our walkthrough on the AI Overviews brand audit shows how to inspect those results. During a ranking update, the non-branded line is where to look first, as the November 2026 core update plan explains.
What are the limits of this data?
Three limits apply. The classification is not perfect, so spot-check the top branded and non-branded queries by hand each quarter. The data starts on February 21, 2026, so you cannot compare to earlier years. And low-volume sites may have no branded data at all.
Choose the data state with care. The reference says that omitting dataState, or setting it to final, returns only finalized data. Setting it to all adds fresh data, which can still change. Use final for monthly reports and all only for a quick look at recent days.
Availability has also changed over time. Google's update note says the branded filter reached all eligible sites on March 11, 2026. Check the earliest rows of your own property for gaps or sudden steps before you trust a trend line that starts in February.
Also keep numbers consistent across tools. A branded share from the API will not match one calculated with your own keyword rules. Pick one method, document it, and label every chart with the source. Switching methods mid-year creates trends that are only artifacts of the definition.
What is brandedClassification in the Search Analytics API?
Why does my branded queries request return no rows?
Can I filter branded queries by a keyword such as pricing?
How do I pull several countries in one Search Analytics API request?
Up next · SEO Will a Domain Move Fix a SaaS Site's Quality Problem? What Mueller Said in November 2026 join the club.
The week's best SEO and AI search posts, plus one thing to check in your own Search Console, every Thursday morning. Unsubscribe whenever.