Audiences increasingly expect to know when AI has been used to create or significantly alter what they see. What happens after the label appears is less predictable. Recent research suggests that AI disclosure can measurably change how audiences engage with content, and for reasons that have little to do with the technology itself.
Brands are having to think about this more carefully as AI disclosure becomes a normal part of social platforms. Meta now uses an “AI info” label for content it identifies as AI-generated, while lighter AI editing may be disclosed less prominently. Meanwhile, TikTok and YouTube have introduced their own systems for identifying and labeling generated or significantly altered content.
There is also a clear gap between wanting transparency and feeling able to judge AI content independently. According to a 2025 Pew Research Center survey, 76% of US adults said it was extremely or very important to tell AI-generated content apart from human-made content, but only 12% felt extremely or very confident that they could do that themselves.
Do Audiences Want AI Content Labels?
Yes. Support for disclosing meaningful AI use is consistent across markets. Whether that disclosure earns trust is a separate question.
The Ipsos AI Monitor 2025 found that 79% of respondents across 30 countries agreed that companies using AI should have to disclose it, with 40% strongly agreeing.
AI labels also have limits. Saying that something was “AI-generated” describes how it was made. It says nothing about whether the information is accurate.
How Do AI Labels Affect Engagement On Social Media?
The evidence here points in more than one direction. Several studies record fewer likes and weaker overall engagement. Others find that people register the disclosure without changing how much they interact.
7% to 8% fewer likes
The Journal of Consumer Research study on AI disclosure and social media engagement analysed TikTok posts and found that those carrying an AI disclosure received around 7% to 8% fewer likes and about 7% less combined engagement than comparable posts without one. Because the researchers also ran eight preregistered experiments, they were able to test the difference against normal variation in social media performance.
Lower affective and behavioural engagement
Researchers in Electronic Markets also found lower engagement when people were told that content had been AI-enhanced or AI-generated. The reaction was more negative when AI appeared to have done most of the creative work, pointing to an important distinction between using AI as a tool and using it as the main creator.
No significant overall engagement change
Another CHI 2025 study on AI content labels and engagement, this time with 911 participants, found that labels helped people identify AI-generated or altered content. Their overall likelihood of liking, commenting, or sharing did not shift significantly.
For communicators, the practical takeaway is to stop benchmarking labelled and unlabelled content against each other. The comparison that matters is between two labelled posts, one that explains the work and one that does not.
Why Can AI Disclosure Reduce Engagement?
The drop in engagement does not seem to come simply from people disliking AI. The Journal of Consumer Research study found that one of the bigger issues is perceived effort, because when participants saw an AI disclosure, they tended to assume the creator had put less work into the post, which weakened their sense of connection with that person and made them less likely to engage.
The researchers also found that this reaction could change when the AI tool itself seemed more demanding to use. A disclosure mentioning Photoshop neural filters led participants to perceive more effort than a generic AI label, and the difference in parasocial connection disappeared.
There is also a psychological effect that helps explain why people can strongly support AI disclosure without feeling personally vulnerable to AI content. Saifuddin Ahmed’s research on deepfake influence found a third-person effect, where people believed deepfakes were more likely to influence others than themselves, and they also tended to think they were better than others at spotting them, even though that confidence did not predict their actual detection ability.
That gives communicators something concrete to work with: identify the parts of the process that took genuine skill or time, and make sure audiences can see them.
Does AI Label Wording Affect How People React?
Yes, because the wording around AI use can shape how much of the work people think came from a person and how much came from a machine. “AI generated” can suggest that AI did most of the creative work, while “AI enhanced” or “AI assisted” leaves more room for visible human involvement.
The Electronic Markets study on AI content labels found the biggest drop in engagement for fully AI generated content. AI enhanced content performed better, though still below content presented as human created.
Brands do not always control the label itself, because platforms such as Meta, TikTok, and YouTube may apply their own terminology and disclosure systems. What brands can control is the language around that label, so if a platform uses a broad term, the brand can add context about whether AI generated the content, supported editing, or played a smaller role in a human-led process.
For brands working across markets, this is also a translation problem. Disclosure wording rarely carries the same weight in Portuguese and Spanish as it does in English, and a phrase that reads as neutral in one market can read as an admission in another. It is worth treating disclosure copy as copy, not as a compliance string to be translated once.
What Types Of Content Get The Strongest Reaction To AI Labels?
The penalty is not evenly distributed. Four contexts concentrate it.
Emotional and Craft-Based Content
The Electronic Markets study found a bigger drop in engagement for emotional content than for rational content once AI use was disclosed. That pattern makes sense in areas such as fashion, art or storytelling, where human taste and creativity are part of the appeal.
That was the problem Valentino faced in late 2025. Its AI generated DeVain handbag campaign was clearly disclosed, yet followers still reacted negatively because they felt the execution clashed with the human craft they associated with luxury fashion.
Influencer And Creator Content
The same issue can appear with influencers, as people often follow creators for personal experience, taste and authenticity. If AI disclosure makes that contribution feel less direct, it can affect both creator credibility and the brand behind the partnership.
Existing Brand Relationships
Research on AI disclosure and brand loyalty found that existing followers were more resistant to the negative effects of AI disclosure than people without a prior relationship with the brand.
Audiences Already Skeptical Of AI
Reaction also depends on what people already think about the technology, so the same disclosure can feel neutral to one audience and off putting to another.
What Could Latin American Attitudes Toward AI Mean For AI Labels?
Latin America appears to start from a more positive view of AI than many Western markets. The Ipsos AI Monitor 2026 found that people across LATAM and Asia are, on average, more likely to see AI products and services as bringing more benefits than drawbacks, while audiences in Europe and North America are more likely to feel nervous about them.
Within Latin America, those attitudes vary quite a bit. Ipsos data across the region found positive views ranging from 71% in Peru and 66% in Mexico to 56% in both Brazil and Argentina.
A 2026 study from Ecuador points to another interesting tension. 65% of respondents said they had recognised political content generated with AI, while 74% said it had not influenced their own vote, and only 12% acknowledged some influence. At the same time, people still broadly supported AI labelling, regulation and ethical limits, which suggests that support for transparency can coexist with the belief that AI is more likely to affect someone else.
For international brands, a warmer attitude toward AI in one market is not a licence to reuse the same disclosure across the region. In our experience, the human proof point also has to be local: regional audiences respond to seeing the team that actually made the work, not a global credit line.
How Should Brands Label AI Generated Content?
AI disclosure is becoming part of normal content governance, with platforms tightening their policies and regulations beginning to set legal requirements in some markets. Under the EU AI Act’s Article 50 transparency requirements, relevant obligations have applied since 2 August 2026, with some transparency infringements carrying fines of up to €15 million or 3% of worldwide annual turnover.
1. Make the Work Visible
If AI makes the process look effortless, audiences may assume that less skill, thought, or investment went into the content. Showing that a person was involved is useful. Showing the work itself is stronger: research, art direction, editing, iteration, expert judgment, and technical skill.
2. Describe AI’s Role Accurately
Use the language you control to clarify what AI actually did, especially when a platform label is broad. If AI generated the asset, say so while editing, or assistance should not imply full automated authorship. More detail does not always build trust, so the goal is accuracy without overexplaining.
3. Match The Disclosure To The Content
The level of disclosure should reflect what audiences expect from the format. A product explainer or localization asset may need less context than an influencer testimonial, artistic campaign, or executive thought leadership post, where human expertise, emotion, or craftsmanship is part of the value. This is also why localization tends to be the safest place for AI to sit, which matters in Latin America, where localization is how most international brands enter the region in the first place.
4. Test the Approach by Market
International brands should be careful about turning a global transparency standard into one global communications template. Reactions vary by market, and the evidence from Latin America is still too limited to assume that language tested in the US or Europe will work the same way in São Paulo, Mexico City, or Bogotá.
The competitive question is no longer whether brands disclose AI use. It is what audiences understand about the brand once they do, and that answer changes from one market to the next. Sherlock’s Social Media and Research and Insights teams work with international brands on exactly that: shaping disclosure language market by market and testing it before it ships.