Better Together Agency's 2025 Biases in Generative AI Survey
Generative AI, bias and consumer trust
Findings from an online 2025 survey of 1,010 U.S. adults on generative AI bias, trust and stated product preferences. Read Better Together Agency’s second annual report.
What respondents reported about bias and trust
Better Together Agency’s 2025 survey examines how U.S. adults view generative AI bias, fairness and company responsibility. Respondents reported their experiences, trust and stated product preferences.
The findings support a business case for reviewing bias, accuracy and accountability. Stated preferences should be read with the study’s method and limitations, rather than treated as observed revenue or customer loss.
Reported trust
The 2025 report describes respondents’ views on how fairness and bias relate to trust in companies.
Stated product choice
Respondents were asked whether bias would affect their willingness to keep using a product.
Company responsibility
The report examines expectations for companies to review and address generative AI bias.
Consumer views in the 2025 study
The report presents respondents’ expectations for companies using generative AI. These are reported views from the 2025 sample.
Study population
1,010 U.S. adults age 18 or older completed the 2025 online survey.
1/3
Reported intention
About one in three respondents said they might stop using a product if its generative AI showed bias. This is a stated preference from the 2025 study.
Perceived sector risks
Respondents named healthcare, education, finance and hiring among areas of concern.
A business case for reviewing generative AI bias
The 2025 survey reports U.S. adults’ views on generative AI bias, trust and product use. Stated willingness to stop using a product may indicate a concern for organizations, but the survey does not measure actual revenue loss or market share.
Organizations can use these findings when planning bias review, accuracy checks and accountability. The full report includes the study method, respondent population and limitations.
How respondents think about AI bias
The survey asks about trust, fairness and stated product preferences. These responses help organizations understand consumer expectations. Evaluating customer retention, sales or other business outcomes requires separate evidence.
How bias manifests in generative AI
Generated text
Review drafts for stereotypes, missing perspectives and assumptions about people.
Image representation
The 2025 report includes an individual example of an image request producing an inaccurate portrayal of Maya Angelou.
Review AI-assisted communications
Check generated text and images before publication. Review accuracy, language, representation and context. Set a process for reporting and correcting problems.
2025 survey findings
The 2025 findings describe respondents’ views on fairness, bias, trust and product choice. Use them alongside the method and limitations in the full report.
Keep examples separate from survey findings
The report includes examples from outside research and company materials. Those examples provide context and should be read separately from Better Together Agency’s 2025 survey results.
Steps for reviewing generative AI bias
The report recommends clear responsibility, careful source review, testing and user feedback. These steps support review and correction. They do not establish a guarantee of fairness, customer loyalty or growth.
Commit and Set the Vision
Make fairness in generative AI a core brand value with leadership backing
Diversify Your Data and Team
Audit data sources and bring diverse perspectives to development
Test, Audit, Repeat
Regularly and transparently audit generative AI systems for bias
Review outputs and controls
Evaluate the checks available for your workflow and keep people responsible for reviewing and correcting outputs.
Engage Users and Continuously Improve
Listen to feedback and adapt generative AI tools accordingly
Ready to Act?
How to Execute Your Generative AI Bias Action Plan
What does the 2025 survey examine?
The 2025 report examines U.S. adults’ views on generative AI bias, fairness and trust. It reports experiences and stated product preferences. It does not measure a company’s revenue or prove commercial effects.
How can bias appear in generative AI?
Generated text or images can contain stereotypes, inaccurate portrayals or missing perspectives. Review the output in its intended context. The 2025 report includes individual examples and outside research, each separate from the survey responses.
Why does review need to continue?
Review AI-assisted materials before publication, record problems and correct them. Keep the process current as the organization’s tools and workflows change.
Why should organizations review AI-assisted communications?
Organizations are responsible for the communications they publish. Reviewing accuracy, language and representation helps them identify problems before materials reach audiences. The 2025 survey provides context on respondents’ trust and expectations.
What did respondents say about product use?
The 2025 report describes respondents’ views on fairness, trust and company responsibility. About one in three said they might stop using a product if its generative AI showed bias. This is reported intent from the 2025 study, rather than observed customer loss.
What should organizations consider when reviewing risk?
A biased or inaccurate communication can require correction and raise questions about an organization’s review process. The survey reports consumer views. Claims about fines, revenue loss, customer lifetime value or market capitalization require separate evidence.
What does the report suggest for communications practice?
The report supports reviewing bias, accuracy and accountability as part of communications practice. Specific claims about increased sales, loyalty or market share need evidence from the organization making the claim.
What practical review steps does the report recommend?
Set clear responsibility. Review source materials and perspectives. Test the workflow and monitor outputs. Give people a way to report problems. Record corrections and review the process regularly.
Why include different perspectives in review?
People with different experience can notice assumptions or omissions others miss. Include relevant perspectives in source selection, testing and review. A diverse team still needs clear responsibility and a process for checking outputs.
What should an organization explain about its AI use?
Explain how the organization uses AI, what people review and how an audience can report a problem. The 2025 report discusses respondents’ expectations for company responsibility and trust.
How should readers interpret examples in the report?
The 2025 report includes an individual example of an image request producing an inaccurate portrayal of Maya Angelou. Treat individual accounts and outside studies as context. They do not estimate how often a model produces biased output across every use.
