Explainer
AI search engines explained: AI Overviews, AI Mode and accuracy
AI search engines answer a query with a generated summary and links instead of only a list of results; Google's AI Overviews and AI Mode may run many related searches across subtopics to build each answer.[1][2] In a 2025 Pew Research Center study, Google users who saw an AI summary clicked a traditional result link in 8% of visits, against 15% without one.[3] A Tow Center test of eight AI search tools found they gave incorrect answers to more than 60% of queries asking them to identify news sources.[4]
AI search engines combine a web search with a language model that writes an answer and links to supporting pages.[1] Google rolled out AI Mode to US users in May 2025, and studies since then have looked at how these tools change what people click and how reliably they cite sources.[5][3][4]
How AI search builds an answer
Google says its AI Overviews and AI Mode generate responses with links to supporting websites, and may use a “query fan-out” technique that issues multiple related searches across subtopics and data sources.[1] For AI Mode, Google describes this as breaking a question into subtopics and issuing many queries at once on the user’s behalf.[2] Google says there are no extra requirements for websites to appear in these features beyond normal search best practice.[1] How ordinary search engines crawl, index and rank pages is explained in how search engines work, and how the models generate text in how large language models work.
Google’s AI Overviews and AI Mode
Google began rolling out AI Mode in Search to US users on 20 May 2025, calling it its most powerful AI search, with follow-up questions and links to the web.[5] The company said at the time that in its biggest markets, such as the US and India, AI Overviews were driving an increase of more than 10% in Google usage for the types of queries that show them.[6]
What happens to clicks
Pew Research Center analysed the browsing data of 900 US adults and found that 58% ran at least one Google search in March 2025 that produced an AI-generated summary.[7] Users who saw an AI summary clicked a traditional search result in 8% of visits, against 15% for those who did not, and clicked a link inside the summary itself in 1% of visits.[3] They ended their browsing session after 26% of pages with an AI summary, compared with 16% of pages with only traditional results.[8]
Pew also found that Wikipedia, YouTube and Reddit were the most frequently cited sources in both Google’s AI summaries and its standard results.[9]
How accurate the citations are
In March 2025 Columbia University’s Tow Center for Digital Journalism tested eight generative search tools with 1,600 queries, asking each to identify the headline, original publisher, publication date and URL behind a news excerpt.[4] Collectively, the tools gave incorrect answers to more than 60% of queries, ranging from 37% for Perplexity to 94% for Grok 3.[4]
The researchers found that the tools often offered incorrect or speculative answers instead of declining, fabricated links and cited syndicated copies of articles.[10] Several appeared to bypass publishers’ robots.txt preferences, the standard that lets sites block crawlers.[10] Why language models produce confident errors is explained in AI hallucinations.
Read next
Retrieval-augmented generation (RAG) explained
How retrieval-augmented generation lets AI chatbots look up documents before answering: chunking, embeddings, hybrid search, reranking and limits.
Questions readers ask
What is an AI search engine?
A search tool that answers with a generated response and links to supporting websites. Google's AI Overviews and AI Mode, for example, may use a "query fan-out" technique that issues multiple related searches across subtopics and data sources to develop a response.[1]
What is Google AI Mode?
An AI search mode Google began rolling out to US users on 20 May 2025. Google describes it as its most powerful AI search, with follow-up questions and links to the web, and says it breaks a question into subtopics and runs many searches at once.[5][2]
Do AI summaries reduce clicks to websites?
In Pew Research Center's analysis of March 2025 browsing data, users who saw a Google AI summary clicked a traditional result in 8% of visits, against 15% for those who did not, and clicked a link inside the summary in 1% of visits.[3]
Are AI search engines accurate?
In a March 2025 Tow Center test, eight generative search tools gave incorrect answers to more than 60% of 1,600 queries asking them to identify the source of a news excerpt, with error rates from 37% for Perplexity to 94% for Grok 3.[4]
Which sources do Google's AI summaries cite most?
Pew found that Wikipedia, YouTube and Reddit were the most frequently cited sources in both Google's AI summaries and its standard results.[9]
Sources
· 10 claims from 4 publishers
Each numbered claim is a statement we checked against the sources listed with it. Tap a number in the text to see its sources in place.
- [1]
Google's AI Overviews and AI Mode generate AI responses with links to supporting websites, may use a "query fan-out" technique that issues multiple related searches across subtopics, and have no extra requirements for sites beyond normal SEO best practice. confirmedas of 2026-11-07
- AI features and your website · Google Search Central · Primary source · How AI features work in Search (retrieved 2026-11-07)
- AI features and your website · Google Search Central · Primary source · How AI features work in Search (retrieved 2026-11-07)
- AI features and your website · Google Search Central · Primary source · Introduction (retrieved 2026-11-07)
- AI features and your website · Google Search Central · Primary source · How AI features work in Search (retrieved 2026-11-07)
- [2]
Google says AI Mode uses a query fan-out technique that breaks a question into subtopics and issues many searches at once on the user's behalf. confirmedas of 2025-05-20
- AI in Search: Going beyond information to intelligence · Google · Primary source · 2025-05-20 (retrieved 2026-11-14)
- [3]
Pew found that users who saw an AI summary clicked a traditional search result in 8% of visits, against 15% for those who did not, and clicked a link inside the summary in 1% of visits. confirmedas of 2025-07-22
- Google users are less likely to click on links when an AI summary appears in the results · Pew Research Center · Primary source · 2025-07-22 (retrieved 2026-11-14)
- Google users are less likely to click on links when an AI summary appears in the results · Pew Research Center · Primary source · 2025-07-22 (retrieved 2026-11-14)
- Google users are less likely to click on links when an AI summary appears in the results · Pew Research Center · Primary source · 2025-07-22 (retrieved 2026-11-14)
- [4]
In a March 2025 test by Columbia's Tow Center of eight generative search tools on 1,600 queries asking them to identify the headline, publisher, publication date and URL of news excerpts, the tools collectively gave incorrect answers to more than 60% of queries, ranging from 37% for Perplexity to 94% for Grok 3. reportedas of 2025-03-06
- AI Search Has a Citation Problem: We compared eight AI search engines. They're all bad at citing news. · Columbia Journalism Review (Tow Center for Digital Journalism) · Secondary source · 2025-03-06 (retrieved 2026-11-14)
- AI Search Has a Citation Problem: We compared eight AI search engines. They're all bad at citing news. · Columbia Journalism Review (Tow Center for Digital Journalism) · Secondary source · 2025-03-06 · Methodology (retrieved 2026-11-14)
- AI Search Has a Citation Problem: We compared eight AI search engines. They're all bad at citing news. · Columbia Journalism Review (Tow Center for Digital Journalism) · Secondary source · 2025-03-06 (retrieved 2026-11-14)
- AI Search Has a Citation Problem: We compared eight AI search engines. They're all bad at citing news. · Columbia Journalism Review (Tow Center for Digital Journalism) · Secondary source · 2025-03-06 (retrieved 2026-11-14)
- [5]
On 20 May 2025 Google began rolling out AI Mode in Search to US users, describing it as its most powerful AI search with follow-up questions and links to the web. confirmedas of 2025-05-20
- AI in Search: Going beyond information to intelligence · Google · Primary source · 2025-05-20 (retrieved 2026-11-14)
- AI in Search: Going beyond information to intelligence · Google · Primary source · 2025-05-20 (retrieved 2026-11-14)
- [6]
Google said in May 2025 that in its biggest markets, such as the US and India, AI Overviews were driving an increase of more than 10% in Google usage for the types of queries that show them. confirmedas of 2025-05-20
- AI in Search: Going beyond information to intelligence · Google · Primary source · 2025-05-20 (retrieved 2026-11-14)
- [7]
In a Pew Research Center analysis of the browsing data of 900 US adults, 58% conducted at least one Google search in March 2025 that produced an AI-generated summary. confirmedas of 2025-07-22
- Google users are less likely to click on links when an AI summary appears in the results · Pew Research Center · Primary source · 2025-07-22 (retrieved 2026-11-14)
- Google users are less likely to click on links when an AI summary appears in the results · Pew Research Center · Primary source · 2025-07-22 (retrieved 2026-11-14)
- [8]
Pew found that users ended their browsing session after 26% of search pages with an AI summary, compared with 16% of pages with only traditional results. confirmedas of 2025-07-22
- Google users are less likely to click on links when an AI summary appears in the results · Pew Research Center · Primary source · 2025-07-22 (retrieved 2026-11-14)
- [9]
Pew found that Wikipedia, YouTube and Reddit were the most frequently cited sources in both Google's AI summaries and its standard results. confirmedas of 2025-07-22
- Google users are less likely to click on links when an AI summary appears in the results · Pew Research Center · Primary source · 2025-07-22 (retrieved 2026-11-14)
- [10]
The Tow Center found that the generative search tools it tested often gave speculative answers instead of declining, fabricated links, cited syndicated copies of articles, and in several cases appeared to bypass publishers' preferences set in robots.txt, the web standard that lets sites disallow crawlers. reportedas of 2025-03-06
- AI Search Has a Citation Problem: We compared eight AI search engines. They're all bad at citing news. · Columbia Journalism Review (Tow Center for Digital Journalism) · Secondary source · 2025-03-06 (retrieved 2026-11-14)
- AI Search Has a Citation Problem: We compared eight AI search engines. They're all bad at citing news. · Columbia Journalism Review (Tow Center for Digital Journalism) · Secondary source · 2025-03-06 (retrieved 2026-11-14)
- AI Search Has a Citation Problem: We compared eight AI search engines. They're all bad at citing news. · Columbia Journalism Review (Tow Center for Digital Journalism) · Secondary source · 2025-03-06 (retrieved 2026-11-14)
- AI Search Has a Citation Problem: We compared eight AI search engines. They're all bad at citing news. · Columbia Journalism Review (Tow Center for Digital Journalism) · Secondary source · 2025-03-06 · Methodology (definition of robots.txt) (retrieved 2026-11-14)
Revision history (1)
- Page created.
Created Nov 14, 2026. Last reviewed by an editor on Nov 14, 2026. Next scheduled review: Mar 16, 2027.
Cite this page
"AI search engines explained: AI Overviews, AI Mode and accuracy." ContentLora, updated Nov 14, 2026. https://contentlora.com/explain/ai-search-engines-explained
Spotted an error? Suggest a correction or emailcorrections@contentlora.com.