of consumers suspected an AI-generated error in brand content in the past three months
The Consumer Trust Gap in AI Content
65% of consumers have suspected an AI-generated error in brand content in the past three months. A survey of 501 US adults on trust, disclosure and what makes brand content credible.
More brands are using AI to produce their content. But do the people reading it trust what they find? To find out, Cherry Data Signals surveyed 501 US adults who had read brand websites, product information or marketing content online in the past three months.
The short version: readers are wary of AI-written content, they rarely check claims themselves, and evidence is what wins them back.
Key findings
- 65% suspected an AI-generated error in brand content in the past three months (see the statistic)
- 40% would trust an article less if told it was written mainly by AI and reviewed by a person (see the statistic)
- 60% want a person writing or checking brand content when AI is used (see the statistic)
- 30% say unsupported claims are what most makes brand content feel generic (see the statistic)
- 70% would be more likely to consider buying from a brand that published useful research they could not find elsewhere (see the statistic)
Most readers cannot tell, and disclosure costs trust
Fewer than half (45%) are confident they can recognize marketing content produced using AI. A quarter (25%) say they are unconfident.
| Confidence in recognizing AI-produced marketing content | % |
|---|---|
| Very confident | 10% |
| Somewhat confident | 35% |
| Neither confident nor unconfident | 25% |
| Somewhat unconfident | 15% |
| Not at all confident | 10% |
| Not sure | 5% |
Telling readers does not settle it. If a brand disclosed that an article was written mainly by AI and reviewed by a person, 40% say their trust would decrease, against 20% whose trust would increase.
| Effect of AI disclosure on trust | % |
|---|---|
| Greatly increase my trust | 5% |
| Slightly increase my trust | 15% |
| Make no difference | 35% |
| Slightly decrease my trust | 25% |
| Greatly decrease my trust | 15% |
| Not sure | 5% |
The higher the stakes, the lower the tolerance
45% would be least comfortable relying on AI-written advice about an important personal decision. Only 5% picked product descriptions.
| Brand content least comfortable relying on if written mainly by AI | % |
|---|---|
| Advice about an important personal decision | 45% |
| Instructions for using a product | 20% |
| Comparisons between products or services | 15% |
| A company's views on an industry issue | 10% |
| Product descriptions | 5% |
| I would be uncomfortable with all of these | 3% |
| I would be equally comfortable with all of these | 1% |
| Not sure | 1% |
Readers are already spotting errors
65% had come across brand content in the past three months that they suspected contained an AI-generated error. Most (40%) did not check whether it was wrong.
| Suspected an AI-generated error in brand content (past three months) | % |
|---|---|
| Yes, and I checked whether it was wrong | 25% |
| Yes, but I did not check | 40% |
| No | 25% |
| Not sure | 10% |
That fits a wider pattern. Only 20% always or often check the source behind a statistic a brand uses, while 45% rarely or never do. Readers are relying on the brand to have done the checking.
| How often consumers check the source behind a brand's statistic | % |
|---|---|
| Always | 5% |
| Often | 15% |
| Sometimes | 30% |
| Rarely | 30% |
| Never | 15% |
| I do not recall seeing statistics used by brands | 5% |
Evidence is what reassures
Asked what would most reassure them about the accuracy of a brand's article, 30% chose links to relevant supporting sources and 25% a review by a named subject expert. 15% chose original research with an explanation of how it was conducted.
| What would most reassure you about the accuracy of a brand's article | % |
|---|---|
| Links to relevant supporting sources | 30% |
| Review by a named subject expert | 25% |
| Original research with an explanation of how it was conducted | 15% |
| The brand's existing reputation | 15% |
| A clear corrections and fact-checking policy | 10% |
| A recent publication or update date | 4% |
| None of these | 1% |
The same holds when two articles offer similar advice: 35% would trust the one supported by several independent sources.
| Which of two similar articles you would trust more | % |
|---|---|
| An article supported by links to several independent sources | 35% |
| An article supported by a named expert's experience | 25% |
| An article supported by detailed real-world examples | 20% |
| An article supported by a clearly explained customer survey | 10% |
| I would trust them equally | 8% |
| I would need more information | 2% |
Generic content is a substance problem
What most makes brand content feel generic is making claims without supporting evidence (30%), followed by repeating information readers have already seen (25%). Unnatural language comes well behind (10%).
| What most makes brand content feel generic or unhelpful | % |
|---|---|
| Making claims without supporting evidence | 30% |
| Repeating information I have already seen | 25% |
| Offering advice that is too broad | 20% |
| Using repetitive or unnatural language | 10% |
| Focusing too heavily on selling | 10% |
| Failing to answer the question I came with | 4% |
| None of these | 1% |
Readers want people in charge
Only 5% accept AI creating content without human review, even for straightforward topics. 60% want a person writing the content or a qualified person checking it, and 20% would rather brands avoided AI altogether.
| View of AI use in brand content | % |
|---|---|
| Brands should avoid using AI to create content | 20% |
| AI is acceptable for editing, but people should write the content | 35% |
| AI can draft content if a qualified person checks it | 25% |
| I do not mind how content is produced if it is accurate and useful | 12% |
| AI can create content without human review for straightforward topics | 5% |
| Not sure | 3% |
Original research moves buyers
70% say a brand publishing useful research they could not find elsewhere would make them more likely to consider it for a future purchase. Only 5% say it would make them less likely.
| Effect of unique research on likelihood of considering a brand | % |
|---|---|
| Much more likely | 25% |
| Somewhat more likely | 45% |
| No difference | 20% |
| Somewhat less likely | 3% |
| Much less likely | 2% |
| Not sure | 5% |
What this means for brands
- Back every claim. Unsupported claims are the top reason content feels generic, and sources are the top reassurance.
- Keep a person accountable. Most readers accept AI only with a qualified human writing or checking.
- Take extra care with high-stakes advice. That is where tolerance for AI is lowest.
- Publish something no one else has. Original research with a clear methodology earns trust and moves buyers.
Related: What makes original research worth linking to, our survey of 863 content professionals, and the full statistics library.
Methodology
Cherry Data Signals surveyed 501 US adults who had read brand websites, product information or marketing content online within the past three months. The survey was conducted in September 2026. Figures are rounded and may not total exactly 100%.
Want insights like these for your brand?
Cherry Data Signals runs custom surveys and delivers data-ready results. See how our survey research works, what a study costs in our survey cost guide, or how we turn results into original research content for PR, SEO and AI search.
Request a Survey