Marketing Strategy

AI Mode queries in Search Console: ten regex filters for GEO

Search Console counts AI Mode searches without a label. Ten regex filters pull out the long, prompt-like questions, so you can answer them on the page that already ranks.

· 20 min read

Robot with a magnifying glass searching web pages. Illustration: Midjourney / Sebastian Hesse
Illustration: Midjourney / Sebastian Hesse

One year after AI Mode launched in the U.S., Google reported that it “has surpassed a billion monthly active users globally and AI Mode queries have more than doubled every quarter since launch.”[1] Google says people use it to “ask nuanced questions”, for example when “comparing options”, and get a response “with links to supporting websites”[2]. For a B2B company, a page linked in one of these responses can be a buyer’s first point of contact.

Search Console does not list these questions separately. It includes AI Mode in the standard web search figures, without a label[3]. A regex filter is the quickest way to pull the likely AI Mode queries out of the list. A filter for queries with eight words or more is a good place to start, because many of them read like prompts typed into a chatbot.

The aim is to find the questions your site already appears for and answer them on the page that already ranks. Anyone who clicks through then finds the answer, and Google’s AI features have a clear passage they could link to. This guide explains that 8-word regex, adds nine more filters for B2B sites, shows how to remove noise and combine patterns, and sets out what to change on the page.

Key points
  • AI Mode has no label in Search Console. Since June 2025, AI Mode has been counted in the standard web search figures[2, 3]. The Generative AI performance report has no query dimension[4], so the wording of queries is only available in the Performance report, where regex filters work[5].
  • Length is a usable signal. Google says the average AI Mode search in the U.S. is three times as long as a traditional search query[1]. A regex for queries with eight words or more pulls out the likely candidates.
  • Twelve patterns, ready to copy. Ten filters find long queries, questions, commands, comparisons, recommendations, pricing, follow-ups, how-tos, integrations and roles. Two exclusions remove tool prompts and brand queries.
  • One Query filter at a time. The interface takes one filter per dimension. To combine patterns, merge them into one: | for OR, .* for AND.
  • Answer on the page that already ranks. Pick queries with many impressions and few clicks, and add a short, direct answer under a heading phrased like the question. Google warns against building a separate page for every query variation[6].

How AI Mode queries appear in Search Console without a label

In AI Mode, people ask detailed questions and can follow up with more questions[7]. Google’s AI then writes a response with links to supporting websites[2]. AI Overviews are a separate feature: a short AI summary of a complicated topic or question[2].

In June 2025, Google confirmed: “AI Mode is now counting towards totals in Search Console.”[3] Google’s documentation says that AI features such as AI Overviews and AI Mode are reported in the Performance report, within the “Web” search type[2]. You cannot isolate AI Mode as a separate segment[3]. AI Mode searches appear in the same query list as classic searches.

What sets them apart is length. In May 2026, Google wrote about AI Mode in the U.S.: “While we’re seeing growth in both short and long queries in AI Mode, the average AI Mode search is triple the length of a traditional Search query.”[1] A filter on query length therefore gives a usable first signal.

AI Mode also produces fragments. When a user asks a follow-up question in AI Mode, Search Console counts it as a new query[7]. This is why short replies such as “yes, pricing” can turn up in the query list. When this was raised publicly in 2026, Google’s John Mueller replied: “Search Console includes information on AI Overviews and AI Mode in the general performance report.”[8]

Since 3 June 2026, Search Console has had a separate Generative AI performance report. Google first rolled it out to a subset of websites, then to all websites worldwide on 31 August 2026[9]. The report shows impressions by page, country, device and date. It has no query dimension[4] and no click data[10]. It shows which pages appear in generative AI features, but not the questions that led there.

The wording of the questions is therefore only available in the standard Performance report, and a regex filter is the quickest way to pull out the long ones.

One caveat: length is only an indicator. Not every long query comes from AI Mode, and short queries are growing in AI Mode too[1]. Treat the filter results as a list of likely chatbot-style questions.

How to set a regex filter in Search Console

  1. Open Search Console and choose your property.
  2. Go to Performance → Search results.
  3. Click + Add filter in the filter row and choose Queries[5].
  4. Choose Custom (regex), keep Matches regex, paste the pattern and click Apply[5].
  5. Set the date range. The report shows the past three months by default[11]. For a smaller B2B site, choose 6 to 16 months to get enough rows. Search Console keeps 16 months of data[12].
  6. Sort the table by impressions.

Path to the regex filter in Google Search Console. Graphic: Claude / Sebastian Hesse

The regex filter sits in the Performance report under Search results. It works for Queries and Pages.

A few rules for regex in Search Console:

  • Two dimensions only. Regex works for Queries and Pages[5]. You can filter both at the same time, for example long queries on blog pages only[5].
  • One filter per dimension. The interface lets you add one Query filter and one Page filter. You cannot add a second Query filter. Google’s help page does not mention this limit; it comes from the interface itself. “How to combine regex patterns in one Search Console filter” below shows how to work around it.
  • RE2 syntax. Regex comes in several flavours. Search Console uses RE2[5], which accepts roughly a subset of what other flavours such as PCRE accept[13]. For this guide, one gap matters: RE2 has no lookahead or lookbehind[13]. In other flavours, a lookahead lets a single pattern express “contains X but not Y”. In Search Console, that needs a separate Doesn’t match filter. So “long but not a tool prompt” cannot be written as one pattern (see “How to combine regex patterns in one Search Console filter” below).
  • Partial match. A pattern can match anywhere in the query unless you anchor it with ^ (start) or $ (end)[5].
  • Not case-sensitive. “CRM” and “crm” match the same pattern[5].
  • Parentheses around alternatives. To match one of several values, separate them with | and wrap the whole group in parentheses[5]. ^(how|what) checks both words at the start. ^how|what only anchors “how”.
  • Exclusions. Choose Doesn’t match regex to see everything except the matches[5].

Test a new pattern on a few sample queries in an RE2-compatible regex tester before you paste it. If a filter returns no rows, check the pattern first.

How the 8-word regex finds long, prompt-like queries

The pattern for queries with eight words or more:

^(?:\S+\s+){7,}\S+$

Read it from left to right:

  • ^ the query starts here.
  • \S+ one “word”: a run of characters without a space.
  • \s+ followed by one or more spaces.
  • (?: … ){7,} this word-plus-space block appears at least seven times.
  • \S+ one last word.
  • $ the query ends here.

Seven words with a space after them plus one final word: the query has at least eight words. To change the threshold, change the number. {5,} finds queries with six words or more, {9,} with ten or more.

A “word” here is anything between spaces. “mid-sized” counts as one word. So does a search operator such as -site:example.com.

Because Search Console matches partially, a shorter form gives the same result: (\S+\s+){7}\S. It looks for eight words anywhere in the query. The longer version is easier to read, so the rest of this guide uses it.

The 8-word regex broken down into its parts. Graphic: Claude / Sebastian Hesse

Seven words with a space after them plus one final word: the pattern finds queries with eight words or more. The example query is fictional.

Ten Search Console regex filters for AI Mode queries

Fictional example: “JohnDoeInc” is an invented vendor of maintenance-planning software for mid-sized manufacturers, with its site at example.com. All queries and figures below are invented for illustration.

Each filter goes into Queries → Custom (regex) → Matches regex, unless stated otherwise.

1. Long, prompt-like queries

  • Pattern: ^(?:\S+\s+){7,}\S+$
  • Catches: queries with eight words or more.
  • False positives: prompts from AI visibility tools, pasted error messages, classic long-tail searches.
  • Fictional example: “we run 3 shifts and 60 machines which maintenance software fits a mid-sized plant”

2. Questions

  • Pattern: ^(how|what|why|when|where|who|which|can|could|do|does|is|are|should|will|would)\b
  • Catches: queries that start with a question word. \b marks the end of the word, so “whatsapp” and “isolation” do not match.
  • False positives: few. “is” also catches queries such as “is a cmms worth it”, which you usually want.
  • Fictional example: “how does condition based maintenance work”

\b only works with ASCII letters[13]. For German queries, this version works as long as the question words do not start or end with an umlaut: ^(wie|was|warum|wieso|wann|wo|welche[rsmn]?|wer|kann|ist|sind|soll|gibt es)\b. A simple \? finds queries that contain a question mark. Few people type one, so expect a short list.

3. Chatbot-style commands

  • Pattern: ^(explain|compare|list|give me|show me|write|summarize|draft|create|help me|act as|tell me)\b
  • Catches: queries phrased as instructions to a chatbot.
  • False positives: “list of cmms vendors” (a classic search), “write for us” (guest post searches).
  • Fictional example: “compare cmms tools for a mid-sized food plant”

4. Comparisons and alternatives

  • Pattern: \b(vs\.?|versus|alternatives?|compared? to|comparison|which is better)\b
  • Catches: comparisons with competitors, tools or approaches.
  • False positives: few. Leave out a plain or; it is so common that it matches most long queries.
  • Fictional example: “cmms vs spreadsheet for maintenance planning”

5. Recommendations

  • Pattern: \b(best|top \d+|top-rated|recommend\w*|which (one|is better))\b
  • Catches: requests for a recommendation or a shortlist.
  • False positives: “best practice” queries. top \d+ needs a number, so “top of funnel” does not match.
  • Fictional example: “best maintenance planning software for small manufacturers”

6. Pricing

  • Pattern: \b(costs?|prices?|pricing|budget|how much|fees?|rates)\b
  • Catches: questions about cost, budget and fees, singular and plural.
  • False positives: “rates” in other senses, such as “failure rates”.
  • Fictional example: “how much does a cmms cost per user”

7. Follow-up fragments

  • Pattern: ^(yes|ok|okay|sure|and|what about|how about|tell me more)\b
  • Catches: conversation fragments, such as “yes, pricing” or “what about …”. In AI Mode, a follow-up counts as a new query[7].
  • False positives: “and” and “ok” at the start of normal searches. \b keeps out “android”.
  • Fictional example: “what about integration with sap”

8. How-to and process

  • Pattern: ^how (to|do|does|can|should)\b|\b(step by step|checklist|template|framework)\b
  • Catches: people who want to do a task themselves and are looking for steps or a template.
  • False positives: “template” also catches download searches.
  • Fictional example: “how to build a preventive maintenance checklist”

9. Integrations

  • Pattern: \b(integrat\w*|connect\w*|sync\w*|api|plugin|works? with)\b
  • Catches: questions about how your product fits into an existing tool stack.
  • False positives: developer searches such as “api documentation”.
  • Fictional example: “maintenance software that works with sap”

10. Role or company size in the query

  • Pattern: \b(for|as) (a |an |the )?(cmo|ceo|founder|head of \w+|marketing (lead|manager|team)|small (team|business)|startup)\b
  • Catches: queries in which people describe who they are or how large their team is.
  • False positives: few. Replace the roles with those of your own buyers.
  • Fictional example: “maintenance software for a small team of technicians”

The patterns for filters 2 to 10 are starting points. Patterns 8 to 10 are my own suggestions for B2B sites. Add the words your buyers use.

All twelve patterns, ready to copy

Click a pattern once to select all of it, or use the Copy button. Then paste it into Custom (regex) and pick the setting shown in the left column. Search Console takes one Query filter at a time, so use the two exclusions on their own or see “How to combine regex patterns in one Search Console filter” below.

Filter and setting Pattern
1. Long, prompt-like queries
Queries · Matches
^(?:\S+\s+){7,}\S+$
2. Questions
Queries · Matches
^(how|what|why|when|where|who|which|can|could|do|does|is|are|should|will|would)\b
3. Chatbot-style commands
Queries · Matches
^(explain|compare|list|give me|show me|write|summarize|draft|create|help me|act as|tell me)\b
4. Comparisons and alternatives
Queries · Matches
\b(vs\.?|versus|alternatives?|compared? to|comparison|which is better)\b
5. Recommendations
Queries · Matches
\b(best|top \d+|top-rated|recommend\w*|which (one|is better))\b
6. Pricing
Queries · Matches
\b(costs?|prices?|pricing|budget|how much|fees?|rates)\b
7. Follow-up fragments
Queries · Matches
^(yes|ok|okay|sure|and|what about|how about|tell me more)\b
8. How-to and process
Queries · Matches
^how (to|do|does|can|should)\b|\b(step by step|checklist|template|framework)\b
9. Integrations
Queries · Matches
\b(integrat\w*|connect\w*|sync\w*|api|plugin|works? with)\b
10. Role or company size
Queries · Matches
\b(for|as) (a |an |the )?(cmo|ceo|founder|head of \w+|marketing (lead|manager|team)|small (team|business)|startup)\b
11. Exclude tool prompts
Queries · Doesn’t match
my location is|output (only|rules)|do not invent|evaluate the .+ on |-site:
12. Exclude your brand (fictional example)
Queries · Doesn’t match
john\s?doe\s?inc|jondoeinc|johndoeinc\.com

How to remove tool prompts and brand queries

Tool prompts

Some long queries do not come from people. AI visibility tools send scheduled prompts, and agents log their full instructions. Writing in Search Engine Journal, S. Mohanadasan classified 16 months of his own Search Console data. He found prompts ending in “. my location is usa.”, prompts in the form “evaluate the [company] on [facet]” and complete machine instructions such as “do not invent results or urls”[14]. On his site, 124 of 1,127 queries were such tracker probes[14]. His figures cover his site only.

I also see queries with several -site: operators. To me they look like the output of monitoring or scraping tools. This is my own observation; I have no source for it.

To exclude these signatures, choose Doesn’t match regex and paste:

my location is|output (only|rules)|do not invent|evaluate the .+ on |-site:

Add the signatures you find in your own data. This exclusion uses up the single Query filter, so it cannot be applied on top of the long-query filter. “How to combine regex patterns in one Search Console filter” below shows the alternatives.

Your own brand

Queries with your brand name show you what existing contacts ask. To find new questions, exclude them. Google’s own example lists the brand name with its misspellings: willow tree|wilow tree|willowtree|willowtee[15]. For the fictional JohnDoeInc, choose Doesn’t match regex and paste:

john\s?doe\s?inc|jondoeinc|johndoeinc\.com

Search Console also has a built-in filter for branded and non-branded queries. Google announced it on 20 November 2025 and made it available to all eligible sites on 11 March 2026[16]. An AI-assisted system classifies the queries. It includes typos of the brand name and queries about the site’s unique products or services[16]. The filter has two limits. It is only available for top-level (domain-level) properties and only for sites with sufficient query and impression volume[16]. Sites with few impressions do not get it[17]. Many smaller B2B sites therefore still need the regex exclusion.

How to combine regex patterns in one Search Console filter

Search Console takes one Query filter at a time, and RE2 has no lookahead[13]. Any combination of two patterns therefore has to fit into a single pattern. Four rules cover most cases:

  • OR: join with |. A|B finds queries that match A or B. If both patterns start with ^(…)\b, merge the word lists into one group.
  • AND, two word groups: connect with .*. .* stands for any characters in between[13]. A.*B only finds A when it comes before B. For both orders, add the reverse: A.*B|B.*A.
  • AND, start word plus length: append (?:\s+\S+){7,}. A question word followed by at least seven more words gives questions with eight words or more.
  • NOT inside a Matches filter: not possible. Use Doesn’t match regex for exclusions. An exclusion uses up the Query filter on its own.

Six combinations from the twelve patterns, ready to copy. The numbers refer to the table above.

Combination and setting Pattern
Questions or commands
2 OR 3 · Queries · Matches
Fictional example: compare cmms tools for a mid-sized food plant
^(how|what|why|when|where|who|which|can|could|do|does|is|are|should|will|would|explain|compare|list|give me|show me|write|summarize|draft|create|help me|act as|tell me)\b
Long questions
1 AND 2 · Queries · Matches
Fictional example: which maintenance software fits a mid-sized plant with 3 shifts and 60 machines
^(how|what|why|when|where|who|which|can|could|do|does|is|are|should|will|would)\b(?:\s+\S+){7,}
Buying signals: comparisons, recommendations or pricing
4 OR 5 OR 6 · Queries · Matches
Fictional example: best maintenance planning software for small manufacturers
\b(vs\.?|versus|alternatives?|compared? to|comparison|which is better|best|top \d+|top-rated|recommend\w*|which one|costs?|prices?|pricing|budget|how much|fees?|rates)\b
Pricing for small teams, in either order
6 AND 10 · Queries · Matches
Fictional example: maintenance software for a small business how much does it cost
\b(costs?|prices?|pricing|how much)\b.*\b(for|as) (a |an |the )?(small (team|business)|startup|founder)\b|\b(for|as) (a |an |the )?(small (team|business)|startup|founder)\b.*\b(costs?|prices?|pricing|how much)\b
Comparisons that mention integrations, in either order
4 AND 9 · Queries · Matches
Fictional example: cmms alternatives that integrate with sap
\b(vs\.?|versus|alternatives?|compared? to)\b.*\b(integrat\w*|api|works? with)\b|\b(integrat\w*|api|works? with)\b.*\b(vs\.?|versus|alternatives?|compared? to)\b
One exclusion for tool prompts and brand
11 OR 12 · Queries · Doesn’t match
Fictional example: removes “johndoeinc login” and “evaluate the johndoeinc on pricing. my location is usa.”
my location is|output (only|rules)|do not invent|evaluate the .+ on |-site:|john\s?doe\s?inc|jondoeinc|johndoeinc\.com

What does not fit in one filter

“Long, but not a tool prompt” needs a Matches and a Doesn’t match condition on Queries at the same time. Three ways around it:

  1. Add a page filter. It limits the list to one section of the site, for example the blog: ^https?://(www\.)?example\.com/blog/. To hide the home pages, use Doesn’t match regex with ^https?://(www\.)?example\.com/(de/)?(\?.*)?$.
  2. Clean up the export. Run the long-query filter, export the table and delete the noise rows in the spreadsheet.
  3. Use the API. The Search Console API works differently. A request can contain several filter groups, and a row is returned only if all groups match[18]. Each filter can use includingRegex or excludingRegex with RE2 syntax[18]. You need a script or a tool that uses the API for this.

How to turn AI Mode queries into answers on your page

Each query in the list is a question someone asked Google, and a page from your site was already in the results[7]. Google describes what people do in AI Mode: they “can ask nuanced questions that might have previously taken multiple searches — from exploring a new concept, to comparing options, and beyond — and get a comprehensive AI-powered response with links to supporting websites.”[2] Some of these are buying questions. For the fictional JohnDoeInc, one of them is “how much does a cmms cost per user for a plant with 40 technicians”.

The aim is a page that answers such a question directly. Anyone who clicks through finds the answer there, and Google’s AI features have a clear passage they could link to as a supporting website. Google does not promise that they will.

There are five steps: pick the queries worth answering, find the page that already ranks, add a short answer there, keep related questions on one page and check the effect later.

1. Pick the queries with many impressions and few clicks

Google itself describes this use of the Performance report: finding “queries with high impressions and low CTR” as opportunities to improve the page title and description[17]. The same list shows where the page content could answer more. Sort by impressions and look for queries that appear often but get few clicks.

Fictional example (invented figures, last 12 months):

Query (fictional) Impressions Clicks Page that ranks
how much does a cmms cost per user for a plant with 40 technicians 310 2 /pricing/
what is the difference between preventive and predictive maintenance for small plants 240 0 /blog/preventive-maintenance/
we run 3 shifts and 60 machines which maintenance software fits a mid-sized plant 95 1 /product/

In my experience, zero clicks are common for queries like these. One possible reason is that the person gets the answer in the AI response itself. The aim is to be one of the linked sources. If someone does click a link in AI Mode, Search Console counts it as a click[7].

2. Find the page that already ranks

Filter the report by the exact query and check the Pages tab. Edit that page; it already ranks for the topic.

3. Add a short, direct answer

Add two to three sentences that answer the question. Put them under a heading phrased like the question. Google says people “generally appreciate it when web pages are organized by paragraphs and sections, along with headings that provide a clear structure to navigate content.”[6] In an opinion piece for Search Engine Journal, Jarboe recommends placing the main definition, key metric or conclusion in the first sentence of each core section[19]. Google does not mention a ranking effect for this. Do it for the reader who asked.

Fictional example: on the JohnDoeInc pricing page, a new heading reads “What does maintenance software cost for a plant with 40 technicians?”. Below it: “JohnDoeInc charges per technician, from €29 per month. A plant with 40 technicians pays about €1,160 per month. The number of machines and sites is not limited.” (Fictional prices.)

You do not need special markup or extra files. Google says: “You don’t need to create new machine readable files, AI text files, or markup to appear in these features.”[2] Its guide on generative AI search adds: “There’s no requirement to break your content into tiny pieces for AI to better understand it.”[6]

4. Keep it on one page

Several long queries often ask the same thing in different words. Answer them in one place. Google warns: “While it might be tempting to create separate content for every possible variation of how people might search (for example, by focusing on other queries that people have asked, or fan-out queries), doing so primarily to manipulate rankings or generative AI responses in Google Search violates Google’s scaled content abuse spam policy.”[6]

5. Note the date and check again later

Write down the date of the change and the page. Search Console data usually lags by two to three days[20]. My habit is to check the same query and page again after four to six weeks. Compare impressions, clicks and position for the weeks before and after the change. Google does not specify a time frame.

From high-impression queries to an updated page and a recheck. Graphic: Claude / Sebastian Hesse

Start with queries that appear often and get few clicks. Update the page that already ranks, then check again. The four-to-six-week interval is a working habit.

What the regex method cannot show

  • It is an estimate. Search Console does not mark which queries come from AI Mode[3]. Length and phrasing are only indicators.
  • Rare queries are missing. Search Console omits some queries to protect user privacy. They are included in chart totals, unless a query filter is applied[17]. Conversation fragments are rare almost by definition[14]. On Mohanadasan’s site, 57.7% of web impressions carried no query string[14]. Treat your filtered list as a lower bound.
  • Filtered totals shrink. Filtering by query or URL can change the report totals because of truncation and anonymized queries[5].
  • 1,000 rows. The table shows at most 1,000 rows[20]. On larger sites, narrow the list with a page filter.
  • Small sites need longer date ranges. Three months is the default[11]. Choose 6 to 16 months if the list stays short.
  • The opt-out setting. Under Settings → Search generative AI, a site can control its inclusion in AI Overviews, AI Mode and Discover. By default, sites are included[21]. A site must be included to be eligible for these features[6]. Check the setting once.

What to do this week

  1. Open Performance → Search results, set the range to 6 to 16 months and add the long-query filter ^(?:\S+\s+){7,}\S+$.
  2. Export the list. Delete tool prompts and brand queries.
  3. Run filters 2 to 10 from the toolkit. Note which types appear most often on your site.
  4. Pick five queries with many impressions and few clicks.
  5. For each one, find the page that ranks and add a two- to three-sentence answer under a heading phrased like the question.
  6. Note the date. Check the same queries and pages again in four to six weeks.

Sources

  1. Mohan, S. (2026) ‘How AI Mode is changing the way people search in the U.S.’. Google The Keyword, 19 May 2026. Available at: https://blog.google/products-and-platforms/products/search/ai-mode-us-insights/ (Accessed: 19 November 2026).
  2. Google Search Central (2025) ‘AI features and your website’. Google Search Central Documentation, last updated 10 December 2025. Available at: https://developers.google.com/search/docs/appearance/ai-features (Accessed: 19 November 2026).
  3. Schwartz, B. (2025) ‘Google Search Console Now Counting AI Mode Clicks, Impressions & Positions’. Search Engine Roundtable, 17 June 2025. Available at: https://www.seroundtable.com/google-search-console-counting-ai-mode-39599.html (Accessed: 19 November 2026).
  4. Google Search Console Help (n.d.) ‘Generative AI performance report (Search)’. Available at: https://support.google.com/webmasters/answer/16984139?hl=en (Accessed: 19 November 2026).
  5. Google Search Console Help (n.d.) ‘Performance report (Search results): Advanced filtering and comparison’. Available at: https://support.google.com/webmasters/answer/17011165?hl=en (Accessed: 19 November 2026).
  6. Google Search Central (2026) ‘Optimizing your website for generative AI features on Google Search’. Google Search Central Documentation, last updated 10 July 2026. Available at: https://developers.google.com/search/docs/fundamentals/ai-optimization-guide (Accessed: 19 November 2026).
  7. Google Search Console Help (n.d.) ‘What are impressions, position, and clicks?’. Available at: https://support.google.com/webmasters/answer/7042828?hl=en (Accessed: 19 November 2026).
  8. Schwartz, B. (2026) ‘Google Shows AI Mode Queries In Search Console, Just Not In Generative AI Reports’. Search Engine Roundtable, 6 August 2026. Available at: https://www.seroundtable.com/google-search-console-ai-mode-queries-41821.html (Accessed: 19 November 2026).
  9. Maoz, H. and Samet, M. (2026) ‘Introducing Search Generative AI performance reports in Search Console’. Google Search Central Blog, 3 June 2026. Available at: https://developers.google.com/search/blog/2026/06/gen-ai-performance-reports (Accessed: 19 November 2026).
  10. Schwartz, B. (2026) ‘Google Search Console Generative AI Performance Report & AI Control Live’. Search Engine Roundtable, 1 September 2026. Available at: https://www.seroundtable.com/google-search-console-generative-ai-tools-live-41984.html (Accessed: 19 November 2026).
  11. Google Search Console Help (n.d.) ‘Performance report (Search results): Overview and basic setup’. Available at: https://support.google.com/webmasters/answer/7576553?hl=en (Accessed: 19 November 2026).
  12. Google Analytics Help (n.d.) ‘[GA4] Queries report’. Available at: https://support.google.com/analytics/answer/13682862?hl=en (Accessed: 19 November 2026).
  13. Google RE2 (n.d.) ‘Syntax’. GitHub, google/re2 Wiki. Available at: https://github.com/google/re2/wiki/Syntax (Accessed: 19 November 2026).
  14. Mohanadasan, S. (2026) ‘The AI Conversations Leaking Into Your Search Console’. Search Engine Journal, 17 August 2026. Available at: https://www.searchenginejournal.com/the-ai-conversations-leaking-into-your-search-console/585663/ (Accessed: 19 November 2026).
  15. Waisberg, D. (2021) ‘Search Console regex filters update and quick tips’. Google Search Central Blog, 2 June 2021. Available at: https://developers.google.com/search/blog/2021/06/regex-negative-match (Accessed: 19 November 2026).
  16. Huzman, M. (2025) ‘Introducing the branded queries filter in Search Console’. Google Search Central Blog, 20 November 2025. Available at: https://developers.google.com/search/blog/2025/11/search-console-branded-filter (Accessed: 19 November 2026).
  17. Google Search Console Help (n.d.) ‘Performance report (Search results): Dimensions and data groupings’. Available at: https://support.google.com/webmasters/answer/17011259?hl=en (Accessed: 19 November 2026).
  18. Google for Developers (2026) ‘Search Analytics: query’. Search Console API, last updated 16 November 2026. Available at: https://developers.google.com/webmaster-tools/v1/searchanalytics/query (Accessed: 20 November 2026).
  19. Jarboe, G. (2026) ‘AI Mode Queries Are 3X Longer – Why Your Page Should Lead With The Answer’. Search Engine Journal, 21 August 2026. Available at: https://www.searchenginejournal.com/ai-mode-queries-are-3x-longer-the-case-for-leading-with-the-answer/585990/ (Accessed: 19 November 2026).
  20. Google Search Console Help (n.d.) ‘Performance report (Search results): Troubleshooting data discrepancies’. Available at: https://support.google.com/webmasters/answer/17010575?hl=en (Accessed: 19 November 2026).
  21. Google Search Console Help (n.d.) ‘Search generative AI control’. Available at: https://support.google.com/webmasters/answer/16908024?hl=en (Accessed: 19 November 2026).
Sebastian Hesse
Sebastian HesseMarketing Communications & Campaign Architect
Connect on LinkedIn