Updated
Branded keywords are searches that include your brand, site or product name, such as “blacktop finder” or “blacktopfinder reviews”. Non-branded keywords are everything else, such as “asphalt paving austin”. To split them in Google Search Console, use the Performance report’s branded queries filter if your site has it, or a Custom (regex) query filter that matches your brand and its misspellings, with “Doesn’t match regex” for the non-branded side.
Why the split matters for reporting
Branded and non-branded searches behave so differently that mixing them hides what is going on:
- Branded searches come from people who already know you. They follow your offline marketing, word of mouth and returning users, not your SEO.
- Non-branded searches are new people finding you. That is the part SEO work grows, and the number to report on.
- Click rates differ a lot. Brand searches are clicked at much higher rates. Mixed in, they make your average CTR look healthy and distort any CTR benchmark; see organic CTR by position.
- They move for different reasons. A brand campaign can lift total clicks while non-brand traffic falls. Report them separately and you see both.
Search Console’s branded queries filter
Google added a branded queries filter to the Performance report on 20 November 2025, and on 11 March 2026 extended it to all eligible sites (Google’s announcement). What it does, from that announcement:
- It splits query data into Branded and Non-branded in the Search results Performance report, for web, image, video and news.
- Classification is done by an AI-assisted system, not keyword matching. It covers your brand name in all languages and typos, and can include queries without your brand name that refer to a unique product or service of yours.
- It is available only for top-level properties (not URL-path or subdomain properties), and only when the site has enough queries and impressions.
- Google says some queries may be misidentified. The filter doesn’t affect rankings.
- Search Console Insights also gained a card with branded vs non-branded clicks.
If your property has it, use it: it catches misspellings and languages a regex would miss. You can’t see or edit its rules, so check the branded list once for surprises. If you don’t see it, or you want rules you control, use a regex filter.
How to filter branded keywords with regex in Search Console
- Open Performance > Search results.
- Click + Add filter, choose Query, then choose Custom (regex).
- Leave Matches regex selected for branded searches, enter your expression and apply.
- For non-branded searches, edit the filter and choose Doesn’t match regex with the same expression.
- Compare dates or export the result. Save the expression somewhere so every report uses the same one.
Search Console uses RE2 syntax, matches anywhere in the query unless you use ^ or $, and is not case-sensitive by default (Google’s filter help).
Regex examples
For a brand called Blacktop Finder, at blacktopfinder.com:
blacktop ?finder|black top finder|blacktopfinder\.comCatches “blacktop finder”, “blacktopfinder”, “black top finder” and the domain. Add misspellings people actually type; your Queries report shows them:
bla(ck|k)\s?top\s?fi(nd|n)er|blacktopfinderCatches “blaktop finder”, “black top finer” and the rest. For a short or common-word brand, match whole words so you don’t catch unrelated searches:
\bacme\b|acme ?directoryWithout \b, “acme” would also match “acmesupply”. For a brand that is an everyday phrase, add the qualifiers people use with it:
(blue door) (directory|reviews|login|app)|bluedoordirectoryRE2 limits to know
- No lookaheads.
(?!...)and(?=...)aren’t supported in RE2 (RE2 syntax). To exclude brand searches, use Doesn’t match regex instead of a negative lookahead. - No backreferences like
\1. \bworks, but only for ASCII word boundaries, so accented brand names need care.- Case-sensitive matching needs
(?-i)at the start; you rarely want it for queries. - The filter only sees named queries. Anonymized queries (searched by few people) are left out of query-filtered totals, so branded plus non-branded won’t add up to the report’s total (Google on anonymized queries).
Which method to use
| Search Console branded filter | Your own regex filter | |
|---|---|---|
| Who decides what is branded | Google’s AI-assisted system | You, with an expression you can read and edit |
| Misspellings and other languages | Included automatically | Only the ones you list |
| Product names without the brand | Can be included | Only if you add them |
| Properties | Top-level properties with enough volume | Any property |
| Same rule over time and across tools | Can change as Google’s system changes | Yes, as long as you keep the expression |
| Works in the API and exports | In the Performance report | Yes: the Search Analytics API accepts regex filters on query |
Many teams use both: Google’s filter for a quick look, and a fixed regex for reports that must stay comparable month to month.
How to find all your branded keywords
- Start with the obvious forms: the name, the name without spaces, the domain, the name plus “reviews”, “login”, “app” or a city.
- Search the Queries report for fragments. Use a Query filter with “Queries containing” and part of your name (“blackt”, “finder”) to find misspellings and run-together versions you didn’t think of.
- Sort by position. Searches where you rank 1 with a very high click rate and that you didn’t expect are often brand variants.
- Include product and feature names that only you use. Leave out generic words that happen to be in your name.
- Check the non-branded list for leaks. Sort the non-branded side by clicks; anything branded near the top means the expression needs another variant.
- Recheck every few months. New misspellings and product names appear.
Common mistakes
- Matching a generic word. If your brand is “Pave Pros”, the regex
pavematches every paving search. Match the full name. - Forgetting the run-together form. People type “blacktopfinder” as often as “blacktop finder”;
?covers both. - Changing the expression mid-year. Then this year’s non-branded number isn’t comparable with last year’s. Keep a dated note of each change.
- Reading branded plus non-branded as the total. Anonymized queries are in neither.
What counts as branded on a directory
A directory has two kinds of name searches, and only one is your brand:
- Your brand: “blacktop finder”, “blacktopfinder austin”. Branded.
- A listed business’s name: “joe’s plumbing austin”, which lands on
/listing/joes-plumbing. Not your brand. These people searched for the business and found your page, which is SEO value your listings create, and what you show businesses when they renew.
Keep business-name searches in non-branded, or better, report them as their own group. A regex can’t do that reliably for thousands of business names, which is where a query classifier helps. Also watch for category words in your brand: if your directory is called “Austin Plumbers”, the search “austin plumbers” is mostly non-branded demand, not people looking for you. Match the exact name with its domain and qualifiers instead.
Using the split in reports
- Report non-branded clicks and impressions as the SEO number, with the period-over-period and year-over-year change.
- Watch branded impressions as a rough measure of how many people know your name.
- Build CTR benchmarks and striking distance lists from non-branded searches only.
- When traffic drops, check which side dropped first; see content decay. A branded drop points to demand or a competitor on your name, a non-branded drop to rankings.
For dashboards built on this, see the Search Console dashboard guide, and the quick wins guide for where the split fits in the order of work.
How DirectoryLens splits them
Add your brand terms under Site settings > Brand terms; DirectoryLens suggests some from your domain, and you can add up to 100, misspellings included. A search counts as branded when it contains a term, ignoring case, and also with spaces removed, so “blacktopfinder” matches “blacktop finder”.
Searches and pages > Brand vs other then shows visits from brand searches and from other searches, by day and in total, each with its share of visits and the change against the previous period, plus the top searches on each side. Visits from anonymized searches are shown separately, because Google doesn’t say what they were. Brand terms also mark brand searches as their own kind in the Click rate check, so they get their own click-rate curve once there is enough data, and on directories business-name searches are sorted into their own kind in Search intent.