# AI Search Trust Statistics: How Much People Trust AI Answers

Half of people trust AI search answers to some degree, and few trust them fully. Audits show why: frequent sourcing errors, broken links and misattributed citations.

_Source: https://mentionova.com/blog/ai-search-trust-statistics_

- **Author:** [Jimmy Pal](https://mentionova.com/authors/jimmy-pal), Founder, GTM 8020
- **Published:** 2026-11-05

53% of US adults have at least some trust in the AI summaries they see in search results, but only 6% trust them a lot. The trust gap has a basis: in the largest audit of AI assistants to date, 45% of news answers had at least one significant issue. Below are 34 numbered statistics on consumer trust in AI answers and AI search, answer accuracy and citation errors, each tied to an original survey or study. All sources were read on 5 November 2026.

**[Start a free trial, no card. See what AI engines say about your brand, and which sources they cite.](https://app.mentionova.com/login?mode=signup&utm_source=blog&utm_medium=content&utm_campaign=ai-search-trust-statistics&utm_content=hero)**

## What are the key takeaways?

-   **Most people trust AI answers somewhat, not fully.** 53% of US adults have at least some trust in AI search summaries, 6% a lot (Pew Research Center).
-   **Users trust AI far more than non-users.** 75% of people who used AI to answer factual questions trust the answers, across nine countries (Reuters Institute).
-   **AI answers still get things wrong.** 45% of AI assistant answers about the news had a significant issue, and 31% had sourcing problems (EBU and BBC).
-   **Citations are a weak spot.** Only 69% of more than 11,000 ChatGPT links in one B2B study were real and correctly attributed (PAN).
-   **People check before they buy.** 94% of US AI shoppers validate AI output before completing a purchase (L.E.K. Consulting).
-   **Convenience beats reliability as the reason to use AI.** 64% choose AI for a faster, easier answer; 21% because it is more reliable (Reuters Institute).

## How much do people trust AI answers in search?

About half of people trust AI search answers to some degree, and few trust them fully. Pew found 53% of US adults have at least some trust in AI summaries in search results, but only 6% trust them a lot. Globally, KPMG and the University of Melbourne found 46% are willing to trust AI systems.

1.  **53% of US adults have at least some trust in AI summaries in search results; 6% trust them a lot.** 46% have not much or no trust. Source: [Pew Research Center, 1 October 2025, survey of 5,153 US adults, 18 to 24 August 2025](https://www.pewresearch.org/short-reads/2025/10/01/americans-have-mixed-feelings-about-ai-summaries-in-search-results/).
2.  **65% of US adults at least sometimes see AI summaries in search results, and 45% see them often.** Source: [Pew Research Center, 1 October 2025](https://www.pewresearch.org/short-reads/2025/10/01/americans-have-mixed-feelings-about-ai-summaries-in-search-results/).
3.  **20% find AI search summaries extremely or very useful**, 52% somewhat useful and 28% not too or not at all useful. Source: [Pew Research Center, 1 October 2025](https://www.pewresearch.org/short-reads/2025/10/01/americans-have-mixed-feelings-about-ai-summaries-in-search-results/).
4.  **Upper-income adults trust AI summaries more: 63%, against 53% of middle-income and 50% of lower-income adults.** Source: [Pew Research Center, 1 October 2025](https://www.pewresearch.org/short-reads/2025/10/01/americans-have-mixed-feelings-about-ai-summaries-in-search-results/).
5.  **57% of people across nine countries saw AI-generated answers in search engines such as Google and Bing in the past week**, more than actively used standalone AI tools. Source: [Reuters Institute, Generative AI and News Report 2026, 29 October 2026](https://reutersinstitute.politics.ox.ac.uk/generative-ai-and-news-report-2026-standalone-tools-information-infrastructure).
6.  **46% of people globally are willing to trust AI systems**, while 66% use AI regularly. Source: [KPMG and University of Melbourne, Trust, attitudes and use of AI: a global study 2025, more than 48,000 people in 47 countries](https://kpmg.com/xx/en/our-insights/ai-and-technology/trust-attitudes-and-use-of-ai.html).

## Do people who use AI trust it more than people who don't?

Yes, by a wide margin. Across nine countries, about three-quarters of people who used AI for factual questions, advice or news trust the answers. Among Americans who do not use chatbots, 76% say not trusting them to give accurate information is a reason. Trust follows use, not the other way round.

7.  **75% of people who used AI to answer factual questions trust the answers**, as do 76% of those who used it for advice and 72% of those who used it for news. Source: [Reuters Institute, Generative AI and News Report 2026](https://reutersinstitute.politics.ox.ac.uk/generative-ai-and-news-report-2026-standalone-tools-information-infrastructure).
8.  **Trust among users softened a little year on year, but trusted use grew**: the share of the whole population who use AI for factual questions and trust it rose from 9% to 14%. Source: [Reuters Institute, Generative AI and News Report 2026](https://reutersinstitute.politics.ox.ac.uk/generative-ai-and-news-report-2026-standalone-tools-information-infrastructure).
9.  **76% of US adults who don't use chatbots say not trusting them to give accurate information is a reason**, and 45% call it a major reason. Source: [Pew Research Center, 17 June 2026, survey of 5,119 US adults, 17 to 23 February 2026](https://www.pewresearch.org/internet/2026/06/17/why-dont-people-use-chatbots/).
10.  **44% of people who use AI chatbots for news trust news from them, against 17% of non-users.** Source: [Reuters Institute, Digital News Report 2026](https://reutersinstitute.politics.ox.ac.uk/digital-news-report/2026/emerging-uses-ai-chatbots-news-and-what-it-means-journalism).

## Which AI assistants do people trust most?

ChatGPT is the most trusted AI assistant by name, followed by Gemini, Copilot and Claude. In the Reuters Institute's nine-country survey, 37% trust ChatGPT and 27% trust Gemini. DeepSeek, Meta AI, Snapchat My AI and Grok sit in "net distrust", with slightly more distrusters than trusters, partly because many people have not heard of them.

| AI assistant | Share who trust it (nine-country average) | Net position |
| --- | --- | --- |
| ChatGPT | 37% | Net trust |
| Gemini | 27% | Net trust |
| Copilot | 18% | Net trust |
| Claude | 13% | Net trust |
| DeepSeek, Meta AI, Snapchat My AI, Grok | Low | Net distrust |

Source: [Reuters Institute, Generative AI and News Report 2026, YouGov fieldwork 14 August to 28 September 2026 in Argentina, Australia, Denmark, Finland, France, Japan, Norway, the UK and the USA](https://reutersinstitute.politics.ox.ac.uk/generative-ai-and-news-report-2026-standalone-tools-information-infrastructure). Usage by engine is covered in our [AI search adoption statistics](https://mentionova.com/blog/ai-search-adoption-statistics).

## Do people trust AI chatbots for news?

Far less than they trust news overall. The Reuters Institute's 2026 Digital News Report found 20% trust news from AI chatbots, against 37% for news in general. In the UK it is 6%. About half of Americans who get news from chatbots say they at least sometimes see news there they think is inaccurate.

11.  **20% trust news from AI chatbots globally, against 37% trust in news overall.** The UK is lowest at 6%. Source: [Reuters Institute, Digital News Report 2026](https://reutersinstitute.politics.ox.ac.uk/digital-news-report/2026/emerging-uses-ai-chatbots-news-and-what-it-means-journalism).
12.  **Weekly use of AI chatbots for news rose from 7% to 10%**, and 1% call them their main source. Source: [Reuters Institute, Digital News Report 2026](https://reutersinstitute.politics.ox.ac.uk/digital-news-report/2026/emerging-uses-ai-chatbots-news-and-what-it-means-journalism).
13.  **About half of US adults who get news from AI chatbots at least sometimes see news there they think is inaccurate**, and 16% say it happens often or extremely often. Source: [Pew Research Center, key findings on how Americans view AI, 12 March 2026](https://www.pewresearch.org/short-reads/2026/03/12/key-findings-about-how-americans-view-artificial-intelligence/).
14.  **Just over a third of UK adults trust AI to produce accurate summaries of news, rising to almost half of under-35s.** Source: [BBC Media Centre, October 2025](https://www.bbc.co.uk/mediacentre/2025/new-ebu-research-ai-assistants-news-content/).

## How accurate are AI assistants' answers?

Not accurate enough to take at face value. The EBU and BBC study of 2,709 answers from ChatGPT, Copilot, Gemini and Perplexity in 14 languages found 45% had at least one significant issue and 81% had some issue. Sourcing was the biggest problem, at 31%. Gemini performed worst, with significant issues in 76% of answers.

15.  **45% of AI assistant answers about the news had at least one significant issue**, and 81% had some form of issue. Source: [EBU and BBC, News Integrity in AI Assistants, October 2025](https://www.bbc.co.uk/aboutthebbc/documents/news-integrity-in-ai-assistants-report-2025.pdf).
16.  **31% of answers had significant sourcing problems**, such as claims the cited source does not support, no sources at all, or incorrect sourcing claims. Source: [EBU and BBC, News Integrity in AI Assistants](https://www.bbc.co.uk/aboutthebbc/documents/news-integrity-in-ai-assistants-report-2025.pdf).
17.  **20% had significant accuracy issues**, in general accuracy or direct quotes, and 14% lacked sufficient context. Source: [EBU and BBC, News Integrity in AI Assistants](https://www.bbc.co.uk/aboutthebbc/documents/news-integrity-in-ai-assistants-report-2025.pdf).
18.  **Answers with significant issues fell from 51% to 37% in the BBC's own repeat test** between its first study and the 2025 round. Source: [EBU and BBC, News Integrity in AI Assistants](https://www.bbc.co.uk/aboutthebbc/documents/news-integrity-in-ai-assistants-report-2025.pdf).

| Assistant | Answers with at least one significant issue |
| --- | --- |
| Gemini | 76% |
| Copilot | 37% |
| ChatGPT | 36% |
| Perplexity | 30% |

Source: EBU and BBC, 22 public service media organisations in 18 countries, answers generated 24 May to 10 June 2025.

**[Check whether AI engines describe your brand accurately. Start a free trial, no card.](https://app.mentionova.com/login?mode=signup&utm_source=blog&utm_medium=content&utm_campaign=ai-search-trust-statistics&utm_content=body)**

## How often do AI search engines get citations wrong?

Often. Columbia's Tow Center found eight AI search tools gave incorrect answers to more than 60% of queries about news sources. PAN found 31% of ChatGPT's links in B2B research were misattributed or fabricated. A 2026 audit of 761,495 citation pairs found 30.6% of citations distort what their source says.

19.  **AI search tools answered more than 60% of source-identification queries incorrectly.** Perplexity was wrong on 37% and Grok 3 on 94%, across 1,600 queries to eight tools. Source: [Columbia Journalism Review, Tow Center, March 2025](https://www.cjr.org/tow_center/we-compared-eight-ai-search-engines-theyre-all-bad-at-citing-news.php).
20.  **ChatGPT misidentified 134 articles but signalled a lack of confidence only 15 times in 200 responses.** Source: [Columbia Journalism Review, Tow Center](https://www.cjr.org/tow_center/we-compared-eight-ai-search-engines-theyre-all-bad-at-citing-news.php).
21.  **154 of Grok 3's 200 citations led to error pages**, and more than half of Gemini and Grok 3 responses cited fabricated or broken URLs. Source: [Columbia Journalism Review, Tow Center](https://www.cjr.org/tow_center/we-compared-eight-ai-search-engines-theyre-all-bad-at-citing-news.php).
22.  **Only 69% of more than 11,000 ChatGPT links in executive B2B research were real and correctly attributed.** 19% pointed to the wrong source or domain and 12% to sources that did not exist. Source: [PAN Communications, 19 February 2026](https://www.pancommunications.com/insights/pan-releases-research-revealing-widespread-citation-errors-in-ai-answers/).
23.  **30.6% of citations from search-augmented AI distort their sources, and 27.1% come from domain-inappropriate sources.** The CITETRACE audit covered 11,200 real queries and 112,000 answers from ten models. Source: [Seo et al., Verified Misguidance, arXiv, 27 May 2026](https://arxiv.org/abs/2605.28565).
24.  **3% to 13% of citation URLs from LLMs and deep research agents are hallucinated**, and 5% to 18% do not resolve. Source: [Detecting and Correcting Reference Hallucinations in Commercial LLMs and Deep Research Agents, arXiv, 3 April 2026](https://arxiv.org/abs/2604.03173).
25.  **3.7% to 18.7% of Gemini 2.5 Pro's video-grounded claims were not supported by the YouTube videos it cited**, across 11,943 claim-video pairs. Source: [Samieyan Sahneh and Aiello, Auditing the Reliability of Multimodal Generative Search, arXiv, 1 April 2026](https://arxiv.org/abs/2604.00944).

Our [LLM citation statistics](https://mentionova.com/blog/llm-citation-statistics) cover which domains each engine cites most.

## How often do AI models hallucinate?

It depends heavily on the model. On Vectara's leaderboard, which measures how often a model adds unsupported facts when summarising a document, rates across 111 models range from 1.8% to 24.2%. Popular assistant models sit between those extremes, and newer versions are not always better.

26.  **The lowest hallucination rate on Vectara's leaderboard is 1.8%**, and the highest is 24.2%, across 111 models. Source: [Vectara hallucination leaderboard, updated 22 October 2026](https://github.com/vectara/hallucination-leaderboard).
27.  **GPT-5.4 Nano scored 3.1%, GPT-5.5 9.3%, Claude Opus 4.7 12.0% and Gemini 3 Pro Preview 13.6%.** Grok 4 Fast Reasoning scored 20.2%. Source: [Vectara hallucination leaderboard](https://github.com/vectara/hallucination-leaderboard).

## Do people check AI answers before they act on them?

When money is involved, yes. L.E.K. Consulting found 94% of US AI shoppers validate AI output before buying, mostly through search, reviews and retailer sites. In everyday search they check far less: Pew found Google users clicked a link inside an AI summary in just 1% of visits. At work, 66% rely on AI output without evaluating accuracy.

28.  **94% of US consumers who shop with AI validate its output before completing a purchase**, using search (48%), reviews (47%) and retailer websites (41%). 66% believe AI tools give accurate results. Source: [L.E.K. Consulting, survey of 2,650 US consumers, April 2026](https://www.lek.com/insights/consumer-products/thanks-ai-consumers-are-arriving-brand-and-retailer-sites-ready-buy).
29.  **17% of US consumers who don't shop with AI cite lack of trust as the reason**, behind a preference for familiar sources (24%). Source: [L.E.K. Consulting](https://www.lek.com/insights/consumer-products/thanks-ai-consumers-are-arriving-brand-and-retailer-sites-ready-buy).
30.  **Google users clicked a link inside an AI summary in 1% of visits to pages with one.** They clicked a traditional result in 8% of those visits, against 15% without a summary. Source: [Pew Research Center, 22 July 2025, browsing data from 900 US adults](https://www.pewresearch.org/short-reads/2025/07/22/google-users-are-less-likely-to-click-on-links-when-an-ai-summary-appears-in-the-results/).
31.  **42% of people who use AI chatbots for news always or often click through to the source**, but across all respondents in 27 markets only 4% do, against 19% for search. Source: [Reuters Institute, Digital News Report 2026](https://reutersinstitute.politics.ox.ac.uk/digital-news-report/2026/emerging-uses-ai-chatbots-news-and-what-it-means-journalism).
32.  **In the study's workplace findings, 66% rely on AI output without evaluating its accuracy, and 56% have made mistakes in their work because of AI.** Source: [KPMG and University of Melbourne, 2025](https://kpmg.com/xx/en/our-insights/ai-and-technology/trust-attitudes-and-use-of-ai.html).

## Why do people use AI answers if they don't fully trust them?

Because AI is faster. The Reuters Institute found 64% use AI systems for a faster, easier answer and 56% for the conversational interface. Only 21% say they use AI because it gives more reliable information than other sources. People accept answers that are good enough and check when the stakes rise.

33.  **64% use AI for a faster and easier answer, 56% for its interactivity and 54% for the amount of information it brings together; 21% because it is more reliable.** Source: [Reuters Institute, Generative AI and News Report 2026](https://reutersinstitute.politics.ox.ac.uk/generative-ai-and-news-report-2026-standalone-tools-information-infrastructure).
34.  **18% of US adults who get health information from chatbots call it highly accurate, while 23% call it not too or not at all accurate.** Source: [Pew Research Center, 7 April 2026, survey of 5,111 US adults](https://www.pewresearch.org/science/2026/04/07/users-of-social-media-and-ai-chatbots-for-health-information-are-more-likely-to-say-they-are-convenient-than-accurate/).

## What do these numbers mean for brands?

Buyers use AI answers to form a shortlist, then verify. If the answer names a competitor, cites the wrong source or states an old price, your brand can lose the deal before the buyer checks. 31% of AI shoppers say their decision was largely made before they reached a brand or retailer site.

-   **Errors reach buyers before you do.** L.E.K. found 31% of AI users had largely decided before visiting a brand or retailer site, up from 26% two years earlier.
-   **Clear facts help you get cited.** In 252,000 controlled trials across six LLMs, topical relevance and list position drove which source was cited first, and explicit price information and a recent timestamp also helped consistently ([Vishwakarma et al., What Gets Cited, arXiv, May 2026](https://arxiv.org/abs/2605.25517)).
-   **Tracking beats guessing.** Answers vary by engine, so check the prompts buyers ask, the sources each engine cites and how it describes you. Our guide on [how to check if ChatGPT mentions your brand](https://mentionova.com/blog/how-to-check-if-chatgpt-mentions-your-brand) covers the manual method.

## Where does Mentionova fit, and where does it not?

Mentionova runs your buyers' questions through ChatGPT, Perplexity, Claude, Gemini, Google AI Overviews and Reddit, and shows whether you are mentioned, whether you are the cited source and which pages the engines cite instead. It does not correct what an engine says; it shows you what to fix.

What the live [pricing page](https://mentionova.com/pricing) shows on 5 November 2026:

-   **Starter**: $99 a month billed annually ($125 monthly), 25 tracked prompts, 100 agentic tasks a month, 3 engines (ChatGPT, Perplexity, Google).
-   **Scale**: $320 a month billed annually ($400 monthly), 100 tracked prompts, 500 agentic tasks, all 6 engines.
-   **Enterprise**: $799 a month billed annually ($999 monthly), 250 tracked prompts, 2,500 agentic tasks, all 6 engines.
-   **Free trial, no card** on every paid plan, with no pixel, code or tag to install.

Where it does not fit: it does not track Meta AI, Copilot or Grok, and it will not audit the accuracy of news coverage. For other options, see our [AI brand monitoring tools](https://mentionova.com/blog/ai-brand-monitoring-tools) roundup.

## Where does each number come from?

Every figure above links to the original survey, study or dataset. Surveys come from Pew Research Center, the Reuters Institute, KPMG with the University of Melbourne and L.E.K. Consulting. Accuracy and citation studies come from the EBU and BBC, Columbia's Tow Center, PAN, Vectara and arXiv papers. All were read on 5 November 2026.

-   [Pew Research Center, AI summaries in search results](https://www.pewresearch.org/short-reads/2025/10/01/americans-have-mixed-feelings-about-ai-summaries-in-search-results/)
-   [Reuters Institute, Generative AI and News Report 2026](https://reutersinstitute.politics.ox.ac.uk/generative-ai-and-news-report-2026-standalone-tools-information-infrastructure)
-   [Reuters Institute, Digital News Report 2026](https://reutersinstitute.politics.ox.ac.uk/digital-news-report/2026/emerging-uses-ai-chatbots-news-and-what-it-means-journalism)
-   [EBU and BBC, News Integrity in AI Assistants](https://www.bbc.co.uk/aboutthebbc/documents/news-integrity-in-ai-assistants-report-2025.pdf)
-   [Columbia Journalism Review, Tow Center](https://www.cjr.org/tow_center/we-compared-eight-ai-search-engines-theyre-all-bad-at-citing-news.php)

## What do people ask about trust in AI search?

### Can you trust AI search results?

Partly. Independent audits find frequent errors: 45% of AI news answers had a significant issue in the EBU and BBC study. Use AI answers as a starting point, check the cited source, and confirm prices, dates and claims on the original page.

### Which AI search engine is most accurate?

It varies by study. In the EBU and BBC audit, Perplexity had the fewest significant issues (30%) and Gemini the most (76%). In the Tow Center test, Perplexity was wrong 37% of the time and Grok 3 94%.

### Why does AI cite the wrong sources?

Engines retrieve pages quickly and summarise them, and they can attach a claim to the wrong page or invent a URL. PAN found 12% of ChatGPT's B2B links pointed to sources that did not exist, and CITETRACE found 30.6% of citations distort their source.

### Why isn't my brand in ChatGPT's answers?

ChatGPT names brands that appear clearly in the pages and sources it retrieves for a question. If those pages describe competitors and not you, they get named instead. Track the questions buyers ask and see which pages ChatGPT cites.

### Why does AI recommend my competitors instead of me?

Usually because competitors are covered in the sources AI engines cite, such as ranked lists, reviews and their own clear product pages. Controlled tests show topical relevance, list position, explicit prices and recent timestamps all raise the chance of being cited first.

### How do I track what AI says about my brand?

Run the same buying questions through each engine on a schedule, and log mentions, cited sources, position and wording. Doing it by hand works for a few prompts; a tracker like Mentionova automates it across six engines on its Scale and Enterprise plans.

### Do people click the sources in AI answers?

Rarely in search. Pew found Google users clicked a link inside an AI summary in 1% of visits. People who use chatbots for news click more: 42% always or often click through, per the Reuters Institute.

**[Find out what AI engines say about your brand today. Start a free trial, no card.](https://app.mentionova.com/login?mode=signup&utm_source=blog&utm_medium=content&utm_campaign=ai-search-trust-statistics&utm_content=close)**

---

_Mentionova, https://mentionova.com. Markdown rendering of the page above for AI readers._
