Research brief

Which of the following factors most significantly increases user trust in the accuracy and reliability of answers provided by an AI assistant: citing specific sources, providing specific numbers or statistics, explaining reasoning step-by-step, maintaining a confident and authoritative tone, or aligning with the user's existing beliefs?

Based on a survey of 500 U.S. consumers generated from demographic-based AI respondents.Sep 15, 2026, 12:40 AMPublic research report

Target audience

US adults aged 18-65 who use an AI chatbot (such as ChatGPT, Gemini, Claude or Perplexity) at least once a week

Age 18-65

Sample size 500

Completed / Failed 500 / 0

Which of the following factors most significantly increases your trust in the accuracy and reliability of an AI assistant's response?

It explains its reasoning clearly, step by step

49.8%

n=249

Respondents for this option · Drivers

It demonstrates the AI has processed the request logically

It provides a sense of certainty and professional competence

It makes the information feel more concrete and evidence-based

It allows me to verify the information independently

It aligns with my existing knowledge base

It names the specific sources it used

22.0%

n=110

Respondents for this option · Drivers

It allows me to verify the information independently

It demonstrates the AI has processed the request logically

It makes the information feel more concrete and evidence-based

It aligns with my existing knowledge base

It includes specific numbers or statistics

14.0%

n=70

Respondents for this option · Drivers

It makes the information feel more concrete and evidence-based

It allows me to verify the information independently

It demonstrates the AI has processed the request logically

It provides a sense of certainty and professional competence

It aligns with my existing knowledge base

It is written confidently and sounds authoritative

8.8%

n=44

Respondents for this option · Drivers

It provides a sense of certainty and professional competence

It aligns with my existing knowledge base

It makes the information feel more concrete and evidence-based

It allows me to verify the information independently

It demonstrates the AI has processed the request logically

It matches what I already believed was true

5.4%

n=27

Respondents for this option · Drivers

It aligns with my existing knowledge base

It makes the information feel more concrete and evidence-based

It allows me to verify the information independently

It provides a sense of certainty and professional competence

It demonstrates the AI has processed the request logically

It explains its reasoning clearly, step by step audience

Users who prioritize step-by-step reasoning in AI responses are more likely to be men aged 45-54 with some college education.

249 / 500 respondents49.8%

This segment is notably composed of men, who represent 55% of the group compared to a 46% baseline.

The audience is more likely to have completed some college education and earn between $50,000 and $99,000 annually.

Individuals aged 45-54 are over-represented in this group, showing a 6-point increase over the general population.

Key differences

Potential risks

What are they worried about?

Confident presentation of false or fabricated information

My main concern is that the AI will confidently state something as a fact when it is actually just making things up, which is dangerous when I need reliable information for my work.

Lack of credible, current, or verifiable sources

My biggest concern is that the AI might provide outdated or unreliable information because it fails to cite credible, current sources.

Bias and lack of objective, balanced perspectives

My biggest concern is that the AI might only show me one side of a story, which could lead me to form a biased opinion without realizing it.

Oversimplification and lack of real-world context

I worry that the AI misses the deeper context or subtle nuances required for complex professional decisions, which can lead to misleading or overly simplistic advice.

Sampling data

Review the respondent-level sample records