Skip to content

Research

What Business Owners Ask AI About Their Google Presence

By Sergey Sulimko, Founder

How this was counted

Population
English-language US questions the labeller judged a business owner would ask about Google presence
Sample
1689
Collected from
2026-08-21
Collected to
2026-08-21
Instrument
scripts/aeo/question_inventory.py → scripts/aeo/evidence_gate.py --questions
Dataset
owner-questions-2026-08
Excluded
questions with our_audience != yes; exclude the unregistered 2026-09 inventory from every finding

The owner-questions-2026-08 dataset contains 2,984 harvested English-language US questions. Of those, 1,689 were classified as questions a business owner would ask, making the owner-audience share 56.6% of the registered corpus.1 The dataset defines its population as questions the labeller judged a business owner would ask.

Question dimension Count (n) Denominator Share
Owner-audience filter 1 1,689 2,984 56.6%
Intent — reviews 1 482 1,689 28.54%
Intent — optimise 1 369 1,689 21.85%
Intent — choose-vendor 1 215 1,689 12.73%
Form — question 1 740 1,689 43.81%
Form — noun-phrase 1 738 1,689 43.69%
Form — task 1 211 1,689 12.49%
Vendor nameable — no 1 907 1,689 53.7%
Vendor nameable — yes 1 782 1,689 46.3%
Maps Agent lane — yes 1 630 1,689 37.3%
Maps Agent lane — partial 1 758 1,689 44.88%
Maps Agent lane — no 1 301 1,689 17.82%
Vertical — none 1 1,593 1,689 94.32%
Vertical — real estate 1 7 1,689 0.41%
Vertical — plumbing 1 6 1,689 0.36%
Vertical — dental 1 5 1,689 0.30%

How large is the owner-question corpus about Google presence?

The registered owner-questions-2026-08 dataset contains 2,984 harvested questions, and its our_audience == yes filter retains 1,689 owner questions.1 The resulting owner-audience share is 56.6% (1,689 / 2,984), with a source-provided 95% Wilson interval of 0.5482–0.5837.1

The dataset defines the population as English-language US questions the labeller judged a business owner would ask, with the owner labels recorded on 2026-08-21.1 The analysis uses the registered corpus and excludes rows where our_audience != yes plus the unregistered September inventory.

Which intents appear most often in owner questions?

The owner-questions-2026-08 dataset labels reviews as the largest listed intent: 482 of 1,689 owner questions (28.54%).1 It records optimise at 369 (21.85%) and choose-vendor at 215 (12.73%) over the same owner-question denominator.1

These are classifier counts over this corpus. They are not search-volume shares, a purchase-intent measure, or a demand forecast.

Which forms do owner questions take?

The owner-questions-2026-08 dataset records question-form as 740 of 1,689 owner questions (43.81%) and noun-phrase as 738 (43.69%).1 Task form accounts for 211 items (12.49%) over the same denominator.1

The form label describes the observed question shape. It is not a measure of quality, urgency, or conversion.

How often can an owner question name a vendor?

The owner-questions-2026-08 dataset labels vendor naming as absent for 907 of 1,689 owner questions (53.7%) and present for 782 (46.3%).1

Here, vendor-nameable means that a question can name a vendor. It is not a named-company result, a recommendation, or a readiness-to-buy signal.

How much of the owner-question corpus falls inside Maps Agent's lane?

The owner-questions-2026-08 dataset's Maps Agent lane labels split the 1,689 owner questions into 630 yes (37.3%), 758 partial (44.88%), and 301 no (17.82%).1 The partial bucket is largest, followed by yes and no.1

This lane classification describes the question's relationship to the measured scope. It is not a ranking, forecast, customer count, or product-performance result.

How often do owner questions name a business vertical?

Vertical specificity is sparse in the registered dataset: 1,593 of 1,689 owner questions have no vertical label (94.32%), while real estate has 7 (0.41%), plumbing has 6 (0.36%), and dental has 5 (0.30%).1

These are buckets in this owner-question corpus. They are not industry prevalence or market-share estimates, and the listed verticals are not presented as the complete set of named verticals.

Limits

The owner-questions-2026-08 dataset covers English-language US questions observed through 2026-08-21 and includes 1,689 owner-labelled questions from 2,984 harvested questions.1 The our_audience == yes filter and explicit exclusion of the unregistered September inventory define the analysis population. The method reports six label-level agreement values from 0.85 to 1.00; those agreement figures describe classifier labels, not a business-owner measure.1 Questions were harvested from keyword corpora and assistant follow-ups. Choose-vendor is a classifier label for a question type, not a purchase-intent or paid-engagement measure. These count shares are not search-volume shares.

These results describe the declared corpus and scope. They do not establish prevalence in another language or market, industry market share, or future demand.

FAQ

How were the questions collected?

The owner-questions-2026-08 dataset was built from harvested English-language US questions observed in keyword corpora and assistant follow-ups through 2026-08-21.1 The named instrument is scripts/aeo/question_inventory.py → scripts/aeo/evidence_gate.py --questions; the filter retains rows with our_audience == yes.1

Were the questions asked directly by business owners?

The owner-questions-2026-08 dataset defines the population as questions the labeller judged a business owner would ask.1 Its owner designation is a classifier label on observed questions, with 1,689 rows in the retained owner population.1

What does the 56.6% owner share mean?

It means the owner-questions-2026-08 dataset retained 1,689 owner-audience questions from 2,984 harvested questions, a 56.6% share.1 The source-provided 95% Wilson interval is 0.5482–0.5837.1

It does not mean that 56.6% of business owners ask these questions.

Does a vendor-nameable question show purchase intent?

The owner-questions-2026-08 dataset labels 782 of 1,689 owner questions vendor-nameable and 907 not vendor-nameable.1 Vendor-nameable describes whether a question can name a vendor; it is not a recommendation or a readiness-to-buy signal.

Can the intent shares be read as search-volume shares?

The owner-questions-2026-08 dataset records reviews at 482 of 1,689 (28.54%), optimise at 369 (21.85%), and choose-vendor at 215 (12.73%).1 These are count shares over the classified owner-question corpus; they are not search-volume shares.

Can these results be generalized beyond English-language US questions?

The owner-questions-2026-08 dataset covers English-language US questions collected through 2026-08-21 and applies the owner-audience filter.1 Its results are bounded to that language, market, collection window, and classified population.

Sources

  1. Maps Agent Research — owner questions, August 2026. Public first-party dataset owner-questions-2026-08, checked 2026-09-17. Open the owner-question dataset

Cite this page with a link. More measurements in Maps Agent Research.

Last updated September 18, 2026. This page is maintained by the Maps Agent research team.