Back to AI & Tech

What people want from AI search summaries

Source report pendingMethodology

Overview

AI summaries are already familiar in practice. About 44% say they saw an AI-generated summary on every or almost every search in the past week.

But that does not mean users simply want them always on. About 28% would turn them off for some searches, 22% would turn them off for all searches, and 25% are not sure.

Stacked breakdown

44% saw AI summaries on every or almost every search in the past week.

In the past 7 days, about how often did you see an AI‑generated summary (sometimes labeled ‘AI Overview’) at the top of the results for a query on a search engine?

Every or almost every search
43.9%
About half of my searches
17.1%
A few searches
17.5%
Not in the past week
21.5%

2025 · base n 1,509 · +/- 3.1%

AI Adoption Survey July 2025

Source report pending

AI summaries are a frequent search layer

About 44% say they saw an AI-generated summary on every or almost every search during the previous week.

Another 17% saw one on about half of searches and 18% saw one on a few searches, leaving 22% who did not see one in the past week.

Stacked breakdown

50% would turn AI summaries off for some or all searches.

If Google let you turn AI summaries off, would you…

Keep them on for all searches
24.6%
Turn them off for some searches
28.4%
Turn them off for all searches
21.6%
Not sure
25.3%

2025 · base n 1,509 · +/- 3.1%

AI Adoption Survey July 2025

Source report pending

Many would use an off switch

If Google allowed users to turn AI summaries off, 25% would keep them on for all searches.

A larger combined group, about 50%, would turn them off for some or all searches.

Stacked breakdown

53% have not thought much about AI search reducing traffic to original websites.

Are you concerned, not concerned, or have you not thought much about the impact that AI search results may have in decreasing traffic to original websites?

Concerned about impacts of decreased traffic
24.2%
Not concerned about impacts of decreased traffic
22.8%
Have not thought much about it
53.0%

2025 · base n 1,509 · +/- 3.1%

AI Adoption Survey July 2025

Source report pending

Publisher traffic is not top of mind for most

Asked about AI search decreasing traffic to original websites, 53% say they have not thought much about it.

Concern and non-concern are close among the rest: 24% are concerned and 23% are not concerned.

Methodology

Full methodology
Mode
Verasight panel recruited via random address-based sampling, random person-to-person text messaging, and dynamic online targeting
Population
US adults age 18+
Field dates
2025-07-30 → 2025-08-04
Base (unweighted)
1,509
Margin of error
+/- 3.1%
Module
AI Adoption Survey July 2025
Sponsor
Verasight
Weight variable
weight
Weighting targets
age, race/ethnicity, sex, income, education, region, metropolitan status

Sources

[1]
  • 01
    Source report pendingShows how often adults recently saw AI summaries in search results.

Citation

AI Adoption Survey July 2025, fielded July 30-August 4, 2025, N=1,509 US adults age 18+, +/- 3.1%.

Source report pending

Verasight survey methodology

How Verasight conducts surveys.

This page describes the Verasight general survey contract, separate from how the Data Library packages it. When a wave report is published, its field dates, sample sizes, and module breakdown are listed in that report.

Mode
Verasight panel recruited via random address-based sampling, random person-to-person text messaging, and dynamic online targeting.
Population
US adults age 18+.
Sample design
Surveys are run as omnibus or single-topic waves. Omnibus waves are split into modules with their own respondent set, typically around one thousand respondents per module.
Field window
Each wave specifies its own field dates. Most omnibus waves field across roughly two weeks.
Weighting
Per-module weighting to CPS targets including age, race and ethnicity, sex, income, education, region, and metropolitan status.
Partisanship benchmark
Pew Research Center's NPORS benchmarking surveys, three-year running average.
Vote benchmark
2024 presidential vote population benchmarks.
Margin of error
Typically about plus or minus 3.4 to 3.6 percent per module at standard module sizes. Question-level MoE is recomputed when a base shrinks materially below the module baseline.
Reporting
The canonical report hub is reports.verasight.io. Verified legacy reports remain on their original hosts, and unpublished reports are marked pending.
Transparency
Verasight is a member of the American Association for Public Opinion Research Transparency Initiative.

The canonical report hub is reports.verasight.io; older reports may remain on verified legacy hosts.