As of November 16, 2026 · data from 22 sites, October 17 – November 13, 2026
New Search Console features often land in the web interface first and in the API later. This time is no different. Google announced the branded queries filter in November 2025, added Lens data to the interface in September 2026 and multi-country selection in October 2026. All three reached the API on November 16, 2026. If you build reports in spreadsheets, Looker Studio or your own scripts, that matters more than a new button. Only API data can feed a recurring client report or an automated alert.
This is not a revolution. The API didn’t get new data, just access to what has been visible in the interface for months. Still, it’s another small step toward better analytics: fewer regular expressions, less manual summing of countries and the first Lens numbers outside the interface. I tested the new fields the same day on 22 sites, my own and the ones I can access in Search Console. A few things turned up that the announcement doesn’t mention.
What Google announced
When and where: November 16, 2026, Google Search Central blog, a post by the Search Console team.
Branded queries: a split into branded and non-branded queries (the brandedClassification field), for eligible sites only.
Lens: multimodal search, meaning Google Lens, Circle to Search on Android and images uploaded to Google Search (the webMultimodal type), also in the BigQuery export.
Countries and devices: a filter for several countries or devices in a single request (the filterLists field).
Why, according to Google: to automate reporting, pull groups of countries such as DACH or LATAM in one request and separate branded traffic without keyword lists in regular expressions.
What my data showed
Branded queries cover all traffic. Query rows showed fewer than half of all clicks, so a regex could only classify that part. The new field splits everything, including queries Search Console doesn’t show.
Google can get it wrong. The same query with my name is branded one day and non-branded the next, and a generic phrase at an online cake shop ends up as branded.
Lens brings visibility, not traffic. Average positions between 1.5 and 2.3, with an overall CTR of 0.7%.
Combined countries add up. One request with a list of countries returns the same result as separate requests added together.
Search Analytics API, October 17 – November 13, 2026: 8 of my sites and 14 sites run by clients and friends (blogs, shops, service businesses, media). 83,060 clicks and 4.32 million impressions in web search.
Before and after
| Feature | In the API until November 15, 2026 | In the API from November 16, 2026 | In the Search Console interface |
|---|---|---|---|
| Branded and non-branded | Only a regular expression on visible queries. Anonymized queries were left out. | The brandedClassification field (BRANDED, NON_BRANDED) as a dimension or filter. All traffic, data from February 21, 2026. | Filter announced on November 20, 2025, available to all eligible sites since March 11, 2026. |
| Lens, Circle to Search, image search | Not available in the API. | type: webMultimodal, also in the BigQuery export. On my sites, data starts September 10, 2026. | A separate search type (multimodal search) since September 24, 2026. |
| Several countries at once | A separate request per country, adding up and manually calculating position weighted by impressions. | filterLists with a list of countries in one request. Totals and position are calculated for you. | Multi-country selection in the filter since October 6, 2026. |
| Several devices at once | Separate requests or grouping by device and adding up. | filterLists with a list of devices (DESKTOP, MOBILE, TABLET). | According to Google’s help page, the filter also combines several devices. |
Search Analytics API before and after November 16, 2026. Dates from Google’s announcements and the Search Console help center. I checked when the data starts with API requests.
Branded queries: all traffic, not just the visible part
Until now, a branded split in API-based reports worked like this: you pull queries, filter them with a regular expression containing the brand name and add them up. The catch is that Search Console doesn’t show every query. Some are hidden for privacy reasons. In my test, the query was known for only 45% of clicks, ranging from 8% to 57% depending on the site. On podrez.pl, it was only one click in four.
The new field doesn’t have that problem. BRANDED plus NON_BRANDED matched the site’s total traffic down to the click on each of the 12 sites that got the split. For the first time, you can calculate the branded share of all traffic through the API, not just of its visible part.
Not every site qualifies. The split only works for properties that cover a whole site, not a URL path or a subdomain. About the traffic threshold, Google only says it takes a “sufficient volume of queries and impressions.” In my test, the smallest site with the split had 29,212 impressions in 28 days and the largest without it had 23,733. The threshold lies somewhere in between, but that’s my observation, not a number from Google. A site below the threshold gets an empty response with no error message, so it’s easy to conclude there’s simply no branded traffic.
My name is a brand one day and not the next
Google warns that the classification is “dynamic and contextual” and some queries may be misidentified. On podrez.pl, you can see exactly that. The query “ewelina podrez-siama,” my first and last name spelled correctly, with a hyphen, had 195 impressions over 28 days. Google classified 145 of them as non-branded and 50 as branded.
When I broke it down by day, the classification turned out to switch. Each day, all impressions went to one group, but the group kept changing: branded on October 21–22, non-branded on October 23–28, branded again on October 29–30, then back and forth, and non-branded every day since November 5. The version without the hyphen (“ewelina podrez siama”) and just “ewelina podrez” were always branded during that time.
It gets more interesting on my food blog, ms-fox.pl. There, the same query is branded almost every time: in 180 of 186 impressions (97%), compared with 50 of 195 (26%) on podrez.pl. I write about how Google recognizes people and brands as entities in Knowledge Graph and Knowledge Panel: How Google Decides Who You Are.
Where Google gets it wrong and where it beats regex
Misses. The same thing happens with generic phrases. At an online cake shop, “torty na zamówienie” (custom cakes), a phrase the whole industry competes for, has 490 impressions as branded and 384 as non-branded. As a result, the average position of “branded” queries comes out at 7.2, worse than non-branded (5.7). On my Bornholm travel guide, bornholm-online.pl, some impressions of questions about ferries from Poland, such as “prom na bornholm z polski” (ferry to Bornholm from Poland), also land in the branded group. The takeaway: before you send a report to a client, check what ended up as branded. Just pull the brandedClassification and query dimensions together.
Hits. At a language school with branches in several countries, Google classifies as branded the name of the exam the school runs, its abbreviation and the school’s name in Polish, Hungarian, Serbian and German. A long, prompt-style question comparing this school with another one is branded too. A regex wouldn’t catch that. On ms-fox.pl, a regular expression with the blog’s name found 25 branded clicks, while Google found 231. Among visible queries, Google also caught typos such as “mrs fox” and “ma fox,” and it found the rest of the clicks in queries you can’t see.
The differences between branded and non-branded queries are what you’d expect. Across the 12 sites, branded CTR ranged from 3.6% to 75% and non-branded CTR from 0.6% to 3.3%. The branded share of clicks was under 5% on 9 of the 12 sites and between 18% and 30% on the other three.
Lens and Circle to Search: plenty of impressions, few clicks
The webMultimodal type covers Google Lens, Circle to Search on Android and images uploaded to Google Search. On my sites, the first data is from September 10, 2026, two weeks before Google announced the feature in the interface.
Across the 22 sites, that’s 15,806 impressions and 105 clicks (0.7% CTR). For comparison, it equals 0.4% of web search impressions. On sites with at least 40 such impressions, the average position is between 1.5 and 2.3, and yet hardly anyone clicks. Between 74% and 96% of these impressions come from phones.
podrez.pl is the exception. Here, 5% of impressions come from Lens and Circle to Search (under 1% on the other sites with more than 5,000 impressions), and 91% of them go to a single post, my guide to AI Mode in Google (in Polish). Until October 26, the post had at most 12 such impressions a day, often none. On October 27, I added screenshots of the AI Mode button and screen. Since November 1, it has had between 138 and 231 a day.
My hypothesis: people take a photo or a screenshot of the button or an error message and search for how to turn on AI Mode themselves, or why they don’t have it. The API doesn’t show Lens queries, but regular searches that showed the same post point the same way.
In the two weeks before I added the screenshots (October 13–26), not one of them asked how to turn AI Mode on. In the two weeks after (October 31 – November 13), such queries had 59 impressions, while “what is it” queries had only 3. People also typed in error messages word for word, such as the Polish message saying that Web & App Activity is turned off, and even the full text of the AI Mode start screen with no spaces, as if copied from a photo. One in four Lens impressions for this post comes from a desktop computer, so at least some of these are screenshots, not phone photos.
What else do people photograph? On ms-fox.pl, spaghetti squash and mulled wine. On my Madeira guide, madera-online.pl, poncha and Monte Palace. At the cake shop, a Halloween pumpkin. At the language school, Krampus from Austrian folklore. In short: food, places and objects, and very seasonal.
Several countries in one request
I checked whether filterLists returns the same thing as separate requests added up. On madeira-planer.de, Germany, Austria and Switzerland in one request come to 65 clicks, 1,824 impressions and a position of 16.07. Three separate requests add up to exactly the same, and the position is an average weighted by impressions. Countries use three-letter codes (deu, aut, che). Here’s a request that combines countries with the branded split and devices:
{
"startDate": "2026-10-17",
"endDate": "2026-11-13",
"dimensions": ["brandedClassification", "device"],
"dimensionFilterGroups": [{
"filterLists": [
{"dimension": "country", "expressions": ["deu", "aut", "che"]}
]
}]
}
What works and what doesn’t
I tested every row with an API request on November 16, 2026.
| What you want to do | Works? | What the API returns |
|---|---|---|
| Split all traffic into branded and non-branded | yes | Both groups add up to the site’s total traffic, anonymized queries included. |
| See which queries count as branded | yes | The brandedClassification and query dimensions together: every visible query gets a label. |
| Combine the branded split with a query filter | no | Error 400: branded classification can’t be combined with a query filter. |
| Exclude branded queries (“not equals”) | no | Error 400: only the equals operator. Instead of excluding, choose NON_BRANDED. |
| Split image search traffic by brand | yes | Works for the image type. |
| Split Discover or Lens by brand | no | Error 400 for the discover and webMultimodal types. |
| Check branded data before February 21, 2026 | no | Empty response, no error. |
| Get the branded split for a small site | no | Empty response, no error. In my test, the threshold fell between about 24,000 and 29,000 impressions in 28 days. |
| See Lens pages, countries, devices and days | yes | The webMultimodal type groups by page, country, device, date and search appearance. |
| See what people searched for in Lens | no | Error 400: webMultimodal can’t be grouped by query. |
| Pull several countries or devices in one request | yes | filterLists: the result equals separate requests added up, position weighted by impressions. |
Filter a country in both filters and filterLists | no | Error 400: one dimension can use only one kind of filter. |
| Combine countries, branded split and devices | yes | Works, see the example above. |
The new Search Analytics API fields in practice. API requests from November 16, 2026 on the sites in my test.
What to do with it
- 1
Switch branded reports from regex to
brandedClassification, but once a month check which queries Google counted as branded. Especially if the site’s name contains a generic phrase. - 2
If there’s no split, check the site’s traffic first, not your request. An empty response doesn’t mean nobody searches for the brand.
- 3
Treat Lens as visibility, not traffic. Check which pages show up there. If they’re products, dishes or places, use your own clear photos. If they’re how-to guides, screenshots of the real interface help.
- 4
Pull regional reports (DACH, the Nordics, Latin America) with a single
filterListsrequest. Fewer requests mean fewer chances to get the position math wrong.
To sum up: a small step, but in the right direction. Google isn’t changing how it counts traffic, it’s shortening the path from data to report. It’s worth using, as long as you remember that Google’s AI, not you, decides what counts as your brand.
Earlier, Search Console also got a report on generative AI. I wrote about what it shows and what it doesn’t in Search Console Generative AI Report: What It Shows and What It Doesn’t.
Frequently asked questions
What’s new in the Search Analytics API as of November 16, 2026?
Three things: a branded and non-branded query split (brandedClassification), data from Google Lens and Circle to Search (type: webMultimodal), and a filter for several countries or devices in one request (filterLists). Before that, these features were only available in the Search Console interface.
Why doesn’t the API return a branded split for my site?
The split only works for properties that cover a whole site and have enough traffic, and only from February 21, 2026. In my test, 12 of 22 sites got it, and the threshold fell between about 24,000 and 29,000 impressions in 28 days. Below the threshold, the API returns an empty response with no error.
Does the branded split include anonymized queries?
Yes. Branded plus non-branded matched total traffic on all 12 sites that got the split. Query rows covered only 45% of clicks in total.
Can I see which photos led people to my site through Lens?
No. webMultimodal data can’t be grouped by query. You can see which pages appear, in which country, on which device and on which day.
Google, Search Analytics: query, API documentation, accessed November 16, 2026.
Google Search Central, Introducing the branded queries filter in Search Console, November 20, 2025, updated March 11, 2026.
Google Search Central, Announcing web multimodal Search performance reporting in Search Console, September 24, 2026.
Search Console Help, Performance report (Search results): Advanced filtering and comparison, accessed November 16, 2026.
Search Engine Roundtable, Google Search Console Performance Report Multi-Select Country Filters, October 6, 2026.
My data: Search Analytics API, 22 sites, October 17 – November 13, 2026, pulled November 16, 2026.
More on this topic: Measuring AI visibility


