How do you filter question and brand queries in Search Console?
This free Search Console regex builder writes the pattern for the Performance report's Custom (regex) filter, so you can pull out question queries, long conversational queries, brand queries, comparison, price and review searches, or the pages in one folder. It also tests the pattern on queries you paste. It runs in your browser and nothing you paste is sent anywhere.
Part 1
Build and test a pattern
Choose what you want to find, adjust it, and optionally test it on a few queries.
1. Which queries or pages do you want to filter?
Pick a pattern, fill in the details it asks for, and press Build. Each build uses one of your free tool runs.
2. How do I apply the pattern in Search Console?
Add a Custom (regex) filter on Queries or Pages in the Performance report.
- Open the Performance report for your property and click + Add filter.
- Choose Queries or Pages, as the builder tells you, then Custom (regex).
- Paste the pattern. Keep Matches regex, or choose Doesn't match regex to see everything else, then click Apply.
Matching is case-insensitive by default. To make it case-sensitive, start the pattern with (?-i). Use ^ to anchor to the start of the text and $ to the end; without them a pattern can match anywhere.
Confirmed Google's advanced filtering article, read 1 November 2026. Multiple values are separated with a vertical bar and wrapped in parentheses, which is how this builder writes them
Part 2
Recipes, a workflow and common mistakes
Tested patterns for the questions marketers ask of their data, a cheat sheet for editing them, and a way to turn a filtered report into a decision.
3. Which pattern answers which question?
Start from the question you are trying to answer, copy the pattern, and test it on a handful of your own queries before you trust the filtered report.
Each pattern below was run against the same 16 invented queries, listed under the table, so you can see what it catches and what it misses.
| You want to… | Pattern | Matched in the sample | Then |
|---|---|---|---|
| See question-shaped queries | (^(who|what|when|where|why|how|which|can|could|do|does|did|is|are|was|were|should|will|would)\b|\?$) | 5 of 16, such as "what is answer engine optimization" | List the topics and check each has a page that answers it in the first paragraph |
| See long, conversational queries | ^(\S+ ){7}\S+ (8 or more words) | 1 of 16, the "who should i hire…" query | Read them in full; they show how buyers phrase a real problem |
| Find comparison searches | \b(vs|versus|compare|compared|comparison|alternative|alternatives|difference between)\b | 2 of 16, such as "semrush vs ahrefs for ai search" | Check whether you have a page that compares the options fairly |
| Find "best of" searches | \b(best|top|leading|recommended)\b | 1 of 16 | Note which list pages rank for them; they are often not your own pages |
| Find price and cost searches | \b(price|prices|pricing|cost|costs|cheap|cheapest|how much)\b | 2 of 16, such as "seo audit cost" | Decide whether your pricing page answers the question or avoids it |
| Find proof searches | \b(review|reviews|rating|ratings|testimonial|testimonials|case study|case studies)\b | 2 of 16, such as "seo consultant case study" | Check that proof pages exist and are easy to find |
| Find provider-evaluation searches | \b(hire|agency|agencies|consultant|consultants|expert|experts|freelancer|company|companies)\b | 4 of 16 | These are closest to buying; prioritize them in your reporting |
| Find requirements and how-to searches | \b(requirements|requirement|checklist|steps|guide|tutorial|template)\b | 1 of 16, "llms.txt checklist" | Check for a downloadable or step-by-step page |
| Isolate brand queries | (northwind signs|northwind sign|northwindsigns) | 2 of 16 | Apply with "Doesn't match regex" to see the non-brand queries (14 of 16) |
The 16 sample queries used for the test
- how to check robots.txt for gptbot
- what is answer engine optimization
- robots txt checker
- schema markup validator
- semrush vs ahrefs for ai search
- best seo consultant for ecommerce
- ai seo agency pricing
- seo audit cost
- northwind signs reviews
- northwind signs
- can i block gptbot but allow oai-searchbot?
- who should i hire to track ai citations for a saas company in the united states
- geo vs aeo difference between
- llms.txt checklist
- seo consultant case study
- is llms.txt needed
Groups overlap on purpose. "Who should i hire to track ai citations…" is a question, a long query and a provider-evaluation search at once, and "can i block gptbot…?" is a question that also ends in a question mark. Do not add the group totals together.
Confirmed Match counts are the tool's own output on the 16 invented queries above. Hypothesis The "Then" column is this article's suggestion, not a Google recommendation
4. What about filtering pages?
Choose Pages instead of Queries and match on the address: a folder, a few path parts, parameters or a file type.
| To find | Pattern |
|---|---|
| Everything in one folder | ^https://example\.com/blog/ |
| Two folders at once | /blog/|/guides/ |
| Addresses with parameters | \? |
| PDF files | \.pdf$ |
| Two language folders | ^https://example\.com/(en|fr)/ |
| Only the home page | ^https://example\.com/$ |
Replace example.com with your own address. The dot is escaped with a backslash because an unescaped dot means "any character".
Confirmed Each pattern was checked against sample addresses with this tool; use "Pages in one folder" above to build the first one
5. What do the symbols mean?
A few symbols cover nearly everything you need, and they behave as follows in Search Console's RE2 filter.
| Symbol | Means | Example |
|---|---|---|
| | Or | New York|San Francisco matches either phrase |
( ) | Group the alternatives | (en|fr) |
^ and $ | Start and end of the text | ^example matches "example.com" but not "an example" |
[ ] | Any one of these characters | c[aie]t matches cat and cit |
*, +, ? | Zero or more, one or more, zero or one | fo+d matches food and fod but not fd; fo*d matches fd too |
\b | A word boundary | \bcost\b matches "seo cost" but not "costume" |
\S | Any character that is not a space | ^(\S+ ){4}\S+ matches queries of five or more words |
\ | Treat the next symbol as plain text | \.pdf$ matches addresses ending in .pdf |
(?-i) | Make matching case-sensitive, placed first | (?-i)AAA matches "AAA" but not "aaa" |
Confirmed Matching rules and several examples are from Google's advanced filtering article, read 1 November 2026; the filter does not support lookahead, lookbehind or backreferences
6. How do you turn a filter into a decision?
Filter, read the actual queries, compare with the rest of the report, and write down one change to test, so the regex leads to an action and not just a number.
- Choose one question. For example, "Which question-shaped queries already bring us impressions?"
- Apply the pattern to Queries in the Performance report for a period of at least a few weeks, so the numbers are not dominated by chance.
- Read the queries, not just the totals. Sort by impressions, open the top ten, and note which have a page that answers them and which do not.
- Compare with the rest. Apply the same pattern with "Doesn't match regex" and compare click-through rate and average position. A group with many impressions and few clicks may mean the snippet or the page does not answer the question.
- Choose one change. Rewrite the opening of one page so it answers the question directly, or add a page for a question you have none for.
- Check again later. Re-apply the same pattern after a few weeks and compare, keeping the dates in your notes.
These filters sort queries by wording. They do not say whether a query came from an AI feature, so for that question use Search Console's own Generative AI reports, which another article explains. Question-shaped and long queries are worth reading because people often phrase problems that way when they ask an assistant, but that is a reason to look, not evidence of where a query came from.
Hypothesis The workflow and the link between conversational phrasing and AI search are this article's suggestions; neither is a measured result
7. Which mistakes cause the most trouble?
The usual causes are a pattern that matches inside longer words, forgotten escapes, and reading a group total as if it measured intent.
- Matching inside other words. The pattern
costalso matches "costume". Wrap short words in\bboundaries, as the built-in groups do. - An unescaped dot or question mark.
.means any character and?alone is an error. Write\.and\?to match the symbols themselves. - Forgetting the anchors. Without
^and$, a pattern matches anywhere in the text, so^matters when you mean "starts with". - Expecting case to matter. Matching ignores case by default, so add
(?-i)at the start if you need it exact. - Copying a pattern from a tool that uses another flavour. Lookahead and backreferences work in some regex engines but not in RE2. This tester warns about them.
- Treating a group as intent. A query containing "cost" may be a buyer or a student. Read the queries before you decide what a group means.
Part 3
Understand the groups
What each pattern looks for, and where it stops being reliable.
8. What does each query group look for?
Each group is a list of words or a shape, so it finds queries that contain them, not queries that mean the same thing.
| Group | What it matches |
|---|---|
| Question queries | Queries starting with who, what, when, where, why, how, which, can, could, do, does, did, is, are, was, were, should, will or would, and optionally those ending in a question mark |
| Long queries | Queries with at least the number of words you choose; the default is seven |
| Comparison | vs, versus, compare, compared, comparison, alternative, alternatives, difference between |
| Best or top lists | best, top, leading, recommended |
| Price and cost | price, prices, pricing, cost, costs, cheap, cheapest, how much |
| Reviews and proof | review, reviews, rating, ratings, testimonial, testimonials, case study, case studies |
| Provider evaluation | hire, agency, agencies, consultant, consultants, expert, experts, freelancer, company, companies |
| Requirements and how-to | requirements, requirement, checklist, steps, guide, tutorial, template |
Hypothesis These word lists are my own starting grouping for sorting buyer questions, not a Google classification. Read the matched queries before you trust a group, and edit the words using the "my terms" option
9. Questions about regex filters in Search Console
Short answers on what regex is used, how to see non-brand queries and why a pattern may behave unexpectedly.
What regex flavour does Search Console use?
RE2. It does not support lookahead, lookbehind or backreferences, and the tester warns you if a pattern uses them.
How do I see only non-brand queries?
Build a brand pattern with your name and its misspellings, then apply it with Doesn't match regex instead of Matches regex.
Why does my pattern match more than I expected?
Without the start and end anchors, a pattern matches anywhere in the text. A short term can appear inside longer words, so test it on real queries first.
Is the test the same as Search Console's?
Not exactly. The test uses your browser's regular expressions, which agree with RE2 for the patterns built here. Check the filtered report before you act on it.
Where do I find the Custom (regex) option?
In the Performance report, click + Add filter, choose Queries or Pages, then Custom (regex). Menus can change, so if a label differs, look for the Custom (regex) option under that filter.
Can I save a filter in Search Console?
This page does not rely on saved filters. Copy the pattern from the builder and keep it in your own notes, with the date range you used, so you can reapply it.
Why does a pattern that tests fine here show different results in the report?
The test uses your browser's regular expressions on the text you pasted, while Search Console applies RE2 to the queries it reports and may hide some queries for privacy. Differences in the query list, not the pattern, are the usual cause.
How many words make a query "long"?
There is no official threshold. The builder defaults to seven or more words as a way to surface conversational phrasing; lower it to see more queries or raise it to see fewer.
10. What can this builder not do?
It writes patterns and tests them on text you paste. It does not read your Search Console data.
- It cannot tell what a searcher meant. A word list finds words, so a query can match two groups or none.
- It does not know what queries your site receives; paste an export to test against real ones.
- Search Console shows only the queries it reports, so a filter cannot reveal anonymised queries.
Next: record whether AI answers name your site → see all the free tools →