How do you split branded from non-branded traffic in a Search Console export?

You split branded from non-branded traffic in a Search Console export by listing your brand terms and labeling each query that contains one as branded. Every other query is non-branded. Google's own Branded and Non-branded filter does this on eligible properties, using an internal, AI-assisted system instead of a term list you control.

• Google's Branded and Non-branded filter is determined by an internal, AI-assisted system, and some queries may occasionally be misidentified (Google, Search Central Blog, read 30 October 2026).
• The filter is available only for top level properties and not for sites with a low number of impressions (Google, Search Central Blog and Search Console Help, read 30 October 2026).
• The Queries table shows at most 1,000 rows and omits anonymized queries, and a download is limited to what the report shows and truncated to 1,000 rows (Google, Search Console Help, read 30 October 2026).
• A regular expression filter on queries uses RE2 syntax and, by default, matches part of a query without regard to case (Google, Search Console Help, read 30 October 2026).
• Penlid's branded split tool takes up to 100 brand terms of up to 80 characters and counts each query once, under its longest matching term (Penlid, branded split tool, read 30 October 2026).

What counts as a branded query in Search Console?

A branded query is a search that includes your brand name, your domain, or a product or service specific to your brand. Google's common-tasks page defines it as Search queries that include your brand name, domain, or a brand-specific product or service and says this includes common misspellings and variations.

Google also says queries in the table are exact matches (non-case-sensitive). We read that as Acme Shoes and acme shoes sharing a row while acmeshoes gets its own. A term list therefore has to cover every spelling, spacing and product name you want counted (our reasoning).

How does Search Console's own Branded and Non-branded filter work?

Search Console's Branded and Non-branded filter labels queries using, in Google's words, an internal, AI-assisted system, and the pages we read publish no term list or exact rule for it, only a description of what it covers. In the Performance report you add a Query filter and Choose either Branded or Non-branded from the filter options. Google's announcement of the filter says the label is NOT based on a regular expression method of including or excluding keywords and warns that some queries may occasionally be misidentified.

The filter is not on every property. Google says This is only available for top level properties and This filter isn’t available for sites with a low number of impressions. A note on the announcement says The branded queries filter is now available to all eligible sites. Check that your property has it first.

Google's export page says an export reflects the report with any filters or special grouping applied, so exporting once per label should give a branded file and a non-branded file (our reading; Google does not say so for this filter).

How do you split an export with a list of brand terms?

You split an export with a list of brand terms in five steps: export, write the terms, choose the matching rule, label each query once, and total each group. The free branded vs non-branded split tool does the matching and totals in your browser, and your file is not uploaded.

1. Export the Queries table. In Performance, Search results, pick the dates, open the Queries tab with Clicks and Impressions switched on, then choose Export and Download CSV.
2. Write the terms. Start with the brand name as one word, add it spaced and joined (acme shoes, acmeshoes), then add the misspellings in your own queries and the products that carry the brand. Case and accents do not matter.
3. Choose the matching rule. The default matches whole words, so acme matches buy acme online and not acmeshoes; matching inside longer words is a looser option for languages that join words.
4. Label each query once. A query counts once, under the longest term that matches it, so nothing is counted twice; a query that matches no term is non-branded.
5. Total each group. Add clicks and impressions per group and divide a group's clicks by its impressions for its CTR, which Google defines as The click count divided by the impression count. Then read the biggest queries on each side (the tool lists 15 per side): a misspelling among the non-branded is a term to add.

How does a regular expression filter compare with a term list and Google's filter?

A regular expression filter on queries lets you write the rule yourself inside Search Console, as a term list does on an export, but it matches part of a query by default. Google's filtering page says The RE2 syntax is used. and Default regex matching is not case-sensitive., and the API's includingRegex operator takes An RE2 syntax regular expression that must be matched. Google says to separate several terms with a vertical bar and surround the whole expression with parentheses, as in (acme|akme). A short brand name can match inside unrelated words, so read the rows it returns (our reasoning).

We use the term list when every number has to be explainable, and Google's filter as a cross-check where a property has it (our recommendation): a gap between the two is a lead on terms to add.

Why does a split cover only the queries in your file?

A split covers only the queries in your file because the table stops at 1,000 rows and omits anonymized queries that the chart still counts. Google says The table can display a maximum of 1,000 rows, that a download is truncated to 1,000 rows of representative examples of your data, and that anonymized (rare) results are omitted from the table. Every share is therefore a share of the table, and the tool warns when a table has exactly 1,000 rows, because the file cannot show whether it was cut.

Do not rebuild non-branded as the chart total minus the branded rows. Google says anonymized queries are included in the chart totals unless a query filter is applied, so the remainder would sweep in queries nobody can label (our reasoning). Build both groups from the same table. For more than 1,000 rows, see the guide to getting more rows out of a Search Console export; for the size of the gap, the guide to how much of your traffic sits in anonymized queries.

Why split before you read CTR or position?

You split first because branded and non-branded queries behave differently, so a blended CTR or position is a mix that fits neither group. Google says Branded queries typically lead to higher-ranking pages from your site, and in the example below branded queries are 7.0% of impressions but 46.2% of clicks. A blended CTR can also fall when non-branded impressions grow although no query's CTR changed (our reasoning). Two related questions are covered in the guides to using your own CTR-by-position curve and to impressions up but clicks down.