To disclose or not disclose, is no longer the question – effect of AI-disclosed brand voice on brand authenticity and attitude

Alexandra Kirkby, Carsten Baumgarth, Jörg Henseler•View original
HighlightsBalancedharry voice
A brand manager writes product copy with artificial intelligence, hits publish, and then wonders: should I have told anyone? For a long time, the working assumption was that transparency about AI authorship would damage the brand — that consumers would feel something was missing, something less genuine. Kirkby, Baumgarth, and Henseler decided to actually test that assumption. What they found should change how brands think about this. The study centers on three concepts worth pinning down. Brand voice authenticity asks whether the text sounds like the brand — whether it reflects the brand's real identity, its attitude, tone, and language. Brand authenticity is broader: it's whether the brand itself feels genuine, original, and reliable. Brand attitude is simpler still: how favorably consumers feel. The fear was that labeling copy as "AI" would erode all three. The researchers also manipulated emotionality — texts ranged from a dry product specification at the low end, through a product description in the middle, up to a chatbot conversation at the high end — because theory suggests people are more skeptical of machines handling emotional content than factual content. To test this, they ran a three-by-three online experiment using real text from Adidas.co.uk. Six hundred twenty-four English-speaking students each saw one stimulus, labeled either "generated by artificial intelligence," "written by a human," or not labeled at all, crossed with one of the three emotionality levels. Manipulation checks confirmed participants actually registered what they read: when the label said AI, 92 percent of participants identified the source as AI. The method used to analyze the results was structural equation modeling — a statistical approach that tests how well a proposed chain of cause and effect fits the actual data. The headline finding is clear. AI-disclosed text was not rated as less authentic than human-disclosed text on any of the three measures. Mean brand voice authenticity scores were nearly identical: 3.15 for human, 3.22 for AI, and 3.24 for undisclosed, on a one-to-five scale. Brand attitude scores were similarly flat: 3.98, 3.86, and 3.91. The model fit was acceptable — a Comparative Fit Index of 0.94, a Root Mean Square Error of Approximation of 0.04 — and none of the disclosure coefficients reached significance. Emotionality didn't moderate the effect either. Even in the chatbot condition, where the text was most emotionally charged, the AI label didn't move the needle. The limitations are real: a student sample, one brand, English-language texts only, and a moment in time when attitudes toward AI are still shifting. Higher-involvement categories — think financial advice or healthcare — remain untested. But within the scope of what was studied, the disclosure penalty brands have been dreading simply did not appear. The question of whether to disclose is no longer the one worth losing sleep over. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.

A brand manager writes product copy with artificial intelligence, hits publish, and then wonders: should I have told anyone? For a long time, the working assumption was that transparency about AI authorship would damage the brand — that consumers would feel something was missing, something less genuine. Kirkby, Baumgarth, and Henseler decided to actually test that assumption. What they found should change how brands think about this. The study centers on three concepts worth pinning down. Brand voice authenticity asks whether the text sounds like the brand — whether it reflects the brand's real identity, its attitude, tone, and language. Brand authenticity is broader: it's whether the brand itself feels genuine, original, and reliable. Brand attitude is simpler still: how favorably consumers feel. The fear was that labeling copy as "AI" would erode all three. The researchers also manipulated emotionality — texts ranged from a dry product specification at the low end, through a product description in the middle, up to a chatbot conversation at the high end — because theory suggests people are more skeptical of machines handling emotional content than factual content.

To test this, they ran a three-by-three online experiment using real text from Adidas.co.uk. Six hundred twenty-four English-speaking students each saw one stimulus, labeled either "generated by artificial intelligence," "written by a human," or not labeled at all, crossed with one of the three emotionality levels. Manipulation checks confirmed participants actually registered what they read: when the label said AI, 92 percent of participants identified the source as AI. The method used to analyze the results was structural equation modeling — a statistical approach that tests how well a proposed chain of cause and effect fits the actual data. The headline finding is clear. AI-disclosed text was not rated as less authentic than human-disclosed text on any of the three measures. Mean brand voice authenticity scores were nearly identical: 3.15 for human, 3.22 for AI, and 3.24 for undisclosed, on a one-to-five scale. Brand attitude scores were similarly flat: 3.98, 3.86, and 3.91. The model fit was acceptable — a Comparative Fit Index of 0.94, a Root Mean Square Error of Approximation of 0.04 — and none of the disclosure coefficients reached significance. Emotionality didn't moderate the effect either. Even in the chatbot condition, where the text was most emotionally charged, the AI label didn't move the needle.

The limitations are real: a student sample, one brand, English-language texts only, and a moment in time when attitudes toward AI are still shifting. Higher-involvement categories — think financial advice or healthcare — remain untested. But within the scope of what was studied, the disclosure penalty brands have been dreading simply did not appear. The question of whether to disclose is no longer the one worth losing sleep over. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.

More in Computer Science →

Related lectures