Branded vs Non-Branded Keywords: How to Split Them in GSC and Read the Mix

A growing traffic graph feels like proof the SEO is working. Split it in two and the story often changes: the growth is branded queries, people searching your name, earned by ads, PR and word of mouth, while clicks from people who do not know you sit exactly where they sat last year.

That split is branded versus non-branded, it takes one regex filter in Search Console, and it is the difference between reporting a number and knowing what moved it. Here is the whole workflow: build the filter, read the halves, and act on each one.

Why blended totals lie to you

A branded query names you: company, product, sub-brand, or a misspelling of any of them. A non-branded query names the problem. The two measure different things, earned by different work:

• Branded clicks are demand you already created. The searcher decided to visit before typing. SEO's only job here is not fumbling the handoff.
• Non-branded clicks are demand you competed for. Rankings, content and technical health win or lose these. This is the half your SEO program actually moves.
• The blend hides whichever half is failing. Brand growth can mask an SEO plateau for a year; a brand dip can make a genuinely improving content program look like decline.

Step 1: List every way people type your brand

The filter is only as good as the list behind it, and the list is sitting in your own query report:

1. Open Performance, sort queries by impressions, and read the top 100 rows.
2. Write down every spelling of your name you see: spaced, unspaced, misspelled, abbreviated.
3. Add product and sub-brand names. Someone searching your product already knows you.
4. Leave competitor names off the list. Ranking for them is acquisition, not navigation.

Expect the list to be longer than you guessed. Real query data is where the typos live, and the typos are branded demand too.

Step 2: Build the regex, use it both ways

In the Performance report: + New → Query → Custom (regex). Join your brand terms with pipes and you have the whole mechanism:

• "Matches regex" with the pattern is your branded view.
• "Doesn't match regex" with the same pattern is your non-branded view. One pattern, both halves, nothing to keep in sync.
• Our free GSC regex generator builds and tests the pattern against sample queries before you paste it, which is worth doing: a filter that silently matches nothing looks exactly like a site with no branded traffic.

Pattern details that save a rewrite

• GSC regex is RE2 and matches anywhere in the query by default, so acme already catches "acme pricing" and "is acme any good".
• Write patterns in lowercase. Query strings in the report are overwhelmingly lowercase; adding (?i) at the front costs nothing if you want the belt and braces.
• Beware short fragments. If your brand is a dictionary word or part of one, anchor it with word boundaries or spacing (\bacme\b) so "acmeic acid" style collisions stay out of the branded half.
• Escape what regex treats as special. A dot in your brand (acme\.io) matches any character unless escaped.

Step 3: Read the two halves separately

With the filter live, read each view for what it can actually tell you:

• Non-branded view: clicks, impressions and new queries. Impressions are demand discovering you; new queries appearing are topics Google newly trusts you for. This is the acquisition scoreboard, and the pool where striking distance keywords live.
• Branded view: CTR and losses. Branded CTR should be high; when it drops, something is intercepting your own demand, a competitor bidding on your name, a SERP feature, or a reputation problem.
• Both views: the same dates. Read them over the same range, and compare periods inside each half rather than across the blend.

Step 4: Track the mix, not just the totals

The quarterly read is where the split pays off. Not the ratio against some industry benchmark, which does not exist in any useful form, but the direction of each half on your own site:

• Non-brand growing: acquisition is working. Credit the content and rankings work, and keep feeding what worked.
• Brand growing, non-brand flat: marketing is carrying a stalled SEO program. The blended total looks fine, which is exactly why you split it.
• Both flat while you publish: effort is not converting into demand. Change the content plan before changing the pace.
• Brand declining: a problem bigger than SEO. Demand for you specifically is drying up, and no amount of on-page work fixes that.

One habit makes the quarterly read trustworthy: keep the trend against your own history, and remember GSC only holds 16 months of it. Keeping your Search Console data past that window is what makes the second-year comparison possible at all.

Step 5: Act on each half differently

The split earns its keep when it changes what you do next:

• Non-brand flat? That is an SEO problem with an SEO playbook: mine the non-branded view for rising impressions, push the almost-ranking queries, fill the content gaps the data shows.
• Branded CTR sagging? Search your own name and look at the page. Own more of it: sharpen the homepage snippet, claim the sitelinks story, and check who is advertising on your brand.
• Reporting up? Give each line its owner and its explanation. A leadership team that sees the two lines separately funds SEO for what SEO actually did.

Tips from running this split for years

• Revisit the pattern quarterly. New products, a rename, a merger: brand lists grow, and a stale filter slowly leaks branded queries into your acquisition numbers.
• Save the pattern somewhere shared. The regex in a teammate's head is a report nobody can reproduce. A shared doc, or a tool that stores the filter, ends the drift.
• Consider a third segment for competitor names. Comparison and alternatives content ranks on competitor queries, and tracking them separately shows whether that strategy earns clicks or just impressions.
• Use the split everywhere, not just in GSC. The same brand list should define brand campaigns in analytics and paid reporting, or the three teams will report three different worlds.
• Do not celebrate ratio changes alone. "70% non-brand" is trivially achieved by brand demand collapsing. Read absolute clicks per half first, ratio second.

Why your split looks wrong

• Branded view suspiciously empty? Test the regex in the generator against real queries from your report. The usual culprits are an unescaped special character or a pattern pasted with a trailing space.
• Non-brand full of your own name? The filter is missing spellings. Sort the non-branded view by impressions and harvest the branded strays into the pattern.
• The two halves do not sum to the total? They never will, by design. Anonymized queries are excluded from query-level rows and from any filtered total, so both filtered views undercount against the property total. Compare each half against itself over time.
• Brand suddenly spiking? Check the news before the filter. PR, a launch, or a viral moment moves branded search within days, and it is the one line SEO should decline credit for.

Tools that make this easier

• Free, on this site: the GSC regex generator writes and tests the brand pattern, including the brand-exclusion version, before it touches your report.
• CrawlRaven: keeps the split standing. Branded and non-branded performance tracked as separate lines on the free SEO monitoring dashboard, from your daily Search Console sync, joined to GA4 engagement and the crawl, so the quarterly mix read is waiting instead of rebuilt. The join matters here too: a flat non-brand line comes with the ranked list of opportunities that would move it.
• Honest alternative: SEO Gets ships branded-keyword tagging over GSC data and does it well; it reads the search side only. Any tool that saves query filters, or a disciplined spreadsheet, beats re-typing the regex.

Set the filter up this week, note the two baseline numbers, and put the quarterly read in the calendar. From then on, every "is SEO working?" conversation starts from the right line.