Branding in the Age of AI: Why Trust Matters More
Why trust becomes the scarce asset when content and design are cheap to generate: what research says about disclosure, visible effort and brand trust, and how buyers judge credibility.
Answer to the question
Does branding AI-made work help or hurt trust?
Branding AI work leans on trust because cheap generation weakens polish as a sign of effort. Research finds that disclosing AI use can lower trust, that having it exposed by others costs more, and that showing a process can raise perceived value. Brands can invest in named accountability, plain statements of method and a consistent record.
Key takeaways
- Schilke and Reimann reported thirteen experiments in which people who disclosed their use of AI were trusted less than people who did not, with a smaller penalty for disclosing than for having the use exposed by a third party.
- Altay and Gilardi (2023, preprint) found in two experiments with 4,976 participants that labelling headlines as AI-generated lowered perceived accuracy and willingness to share, and linked the effect to an expectation of no human oversight.
- Buell and Norton (2011) found in five experiments that showing the steps of a service at work raised its perceived value, so process and label do different jobs.
- A 2001 study of 107 brands related brand trust and brand affect to purchase loyalty and attitudinal loyalty, which relate in turn to market share and relative price.
- Our inference: when polish is cheap to produce it tells a buyer less, so named accountability, a plain statement of method and a consistent record carry more weight.
Polish used to be expensive, and buyers read it as care. Generated text, images and layouts have made it cheap, so branding AI work now leans on other signals of trust. Thirteen experiments by Oliver Schilke and Martin Reimann, in a preprint, found that people who disclosed their use of AI were trusted less than people who said nothing.
The cost of faking a signal
Kevin Simler and Robin Hanson's The Elephant in the Brain treats much of what people buy and display as signalling, and the idea beneath that treatment holds that a signal is believable in proportion to what it costs to fake. The longer account sits in the article on what costly signalling is. Polish, volume and visible craft have worked as brand signals because they took time and money to produce.
Cheap generation of text, images and layouts lowers that cost for every brand, honest or not. When polish costs little, it tells a buyer less, and attention moves to signals that remain costly or checkable. Named people who answer for the work, a record built over years and a promise a buyer can hold the brand to all fit that description.
A reputation is the clearest case. A named person who signs the work puts a standing built over years behind it, and that standing is at risk if the work misleads, so the signature cannot be copied the way a layout can. Branding AI-generated work well therefore depends less on how finished the output looks and more on what stands behind it.
What disclosure and visible effort do to trust
Schilke and Reimann attribute the trust penalty to lower perceived legitimacy. They report that it persisted across disclosure framings, held when disclosure was voluntary, and was smaller but not absent among evaluators with favourable attitudes to technology. The penalty from disclosing was also weaker than the penalty when a third party exposed the use. Their conclusion is that transparency is not always beneficial for trust. For a team deciding how to handle branding AI use, the exposure result carries as much weight as the disclosure result, since it makes silence that is later uncovered the costlier course in their experiments.
News headlines gave a second test. In two pre-registered online experiments with 4,976 participants in the United States and the United Kingdom, reported in a 2023 preprint by Sacha Altay and Fabrizio Gilardi, labelling a headline as AI-generated lowered its perceived accuracy and the willingness to share it, whether the headline was true or false and whether a person or a machine had written it. The effect was about one third the size of labelling a headline as false. The authors link it to an expectation that an AI-labelled headline was written entirely by AI, with no human oversight, and they advise that such labels need clear meanings.
Earlier work on AI labels on products found that the cost of a label depended on whether the product was bought to say something about its owner. What a label suggests therefore varies with the product and the audience, which makes the content of a disclosure the variable to test. Whether a statement naming the people who directed and checked the work lands differently from a bare label is a question for a brand's own audience research.
Showing a service at work raised how much people valued it, according to a 2011 report by Ryan Buell and Michael Norton. Across five experiments in online travel search and online dating, a version that displayed the steps being taken, such as the sites being searched, was valued more than a version that displayed only a progress bar for the same wait. The article on what showing the work does covers the conditions.
Process and label do different jobs. A visible process lets a buyer see effort and sequence, while a label assigns the work to a category and leaves the buyer to infer the rest. In branding AI work, showing the process beside an honest statement of method gives the inference something to rest on.
A 2001 study in the Journal of Marketing examined 107 brands using three surveys of consumers and brand managers. It related brand trust and brand affect to two kinds of loyalty, loyalty shown in purchase and loyalty shown in attitude, while controlling for hedonic and utilitarian value, brand differentiation and share of voice. In the authors' model, purchase loyalty led to greater market share and attitudinal loyalty led to a higher relative price. It places trust upstream of the numbers a brand tracks.
Trust of that kind accumulates through consistency. A brand identity system keeps what a buyer sees the same from one contact to the next, and the psychology behind strong brands gives a buyer a record to judge. When output is cheap, the record is the part a competitor cannot generate quickly, and a team branding AI content at volume has the most to gain from protecting it. The same logic applies to claims made in advertising, packaging and customer messages: each statement that later proves accurate adds to the record, and each that does not subtracts from it.
What buyers weigh when branding AI work
Buyers rarely audit a brand. They read signals and weigh what each one can and cannot show.
| Signal | What it can show a buyer | What it cannot show |
|---|---|---|
| Named people who answer for the work | Someone is accountable | Whether the work is good |
| A plain statement of how AI was used | The brand is open about method | How the audience will weigh it |
| A record kept over years | The brand does what it says | What the next release will be |
| Independent confirmation | A check outside the brand's control | Whether the check covered this buyer's concern |
The balance among them probably varies with the purchase. A high-stakes, rarely repeated purchase leaves little time for a record to form, so named accountability and independent confirmation are likely to matter more there, while a frequent purchase lets the record do most of the work.
A brand that works with AI-assisted output can invest in the left-hand column, because those signals cost something real and build slowly. They cannot be bought overnight.
Common questions
Does branding AI-made work help or hurt trust?
The research is mixed and concerns related settings. A preprint of thirteen experiments found that people who disclosed AI use were trusted less than those who did not, and a 2023 preprint found that AI labels on headlines lowered perceived accuracy. Experiments on showing work in progress found higher perceived value. Branding AI output well therefore depends on what the audience is told about human involvement.
Should a brand say that it used AI?
In the thirteen-experiment preprint, disclosing carried a trust penalty, and having the use exposed by someone else carried a larger one. The studies involve individuals and headlines, not brands, so a team weighing branding AI use should test its own wording with its own audience, including statements that name the people who directed and checked the work.
How do buyers judge a brand's credibility?
Mostly through signals: who answers for the work, what the brand says about its method and what it has delivered over time. A 2001 study of 107 brands related brand trust to two forms of loyalty, which is one reason a long record counts for any team branding AI work.
Sources
Simler, K., & Hanson, R. (2018). The Elephant in the Brain: Hidden Motives in Everyday Life. Oxford University Press.
Schilke, O., & Reimann, M. (2025). 'The transparency dilemma: How AI disclosure erodes trust.' SSRN preprint. doi:10.2139/ssrn.5205850 https://doi.org/10.2139/ssrn.5205850
Altay, S., & Gilardi, F. (2023). 'People are skeptical of headlines labeled as AI-generated, even if true or human-made, because they assume full AI automation.' PsyArXiv preprint. doi:10.31234/osf.io/83k9r https://doi.org/10.31234/osf.io/83k9r
Buell, R. W., & Norton, M. I. (2011). 'The Labor Illusion: How Operational Transparency Increases Perceived Value.' Management Science, 57(9), 1564-1579. doi:10.1287/mnsc.1110.1376 https://doi.org/10.1287/mnsc.1110.1376
Chaudhuri, A., & Holbrook, M. B. (2001). 'The chain of effects from brand trust and brand affect to brand performance: The role of brand loyalty.' Journal of Marketing, 65(2). doi:10.1509/jmkg.65.2.81.18255 https://doi.org/10.1509/jmkg.65.2.81.18255
Cheng, X., & Nam, I. (2026). 'Understanding the influence of AI labels on consumer psychology: the moderating role of product type.' Frontiers in Psychology, 17, 1878095. doi:10.3389/fpsyg.2026.1878095 https://doi.org/10.3389/fpsyg.2026.1878095