AI Insight
This experimental study with 618 short-form video users found that labeling content as "AI-generated" significantly reduced viewers' perception of creative effort and their willingness to strategically curate their content feeds, while "human-made" labels did not differ from unlabeled content. The effect operated through both rational and normative psychological pathways, with the AI label weakening users' intention to intervene in algorithmic recommendations regardless of content type. The research reveals an implicit default assumption that content is human-made unless otherwise indicated, and suggests that algorithmic knowledge does not increase user agency in curation decisions.
Why it matters
These findings have direct implications for platform labeling policies surrounding AI-generated content, suggesting that disclosure labels may inadvertently devalue creative work and reduce user engagement with curation tools. The results indicate that effective AI content policies should emphasize human creativity certification rather than simply flagging AI involvement, potentially reshaping how platforms approach content attribution and user empowerment.
IntroductionThis study examined how content provenance labeling on short-form video platforms shapes users’ perceived creator effort, algorithmic curation efficacy, and strategic curation intention. Whereas prior research has largely framed such labels as instruments for risk disclosure or warning, this study reconceptualizes provenance labels as value signals (i.e., attribution cues that signal and assign creative effort) and investigates the psychological mechanisms through which these signals create beliefs about algorithmic intervention and affect subsequent behavioral intentions.MethodsA between-subjects experiment was conducted with 618 short-form video users using a 3 (label type: human-made vs. AI-generated vs. unlabeled) × 2 (content type: eudaimonic vs. hedonic) factorial design to test labeling effects and the dual-pathway mechanism linking perceived effort with strategic curation.ResultsThe AI-generated label significantly reduced perceived effort, whereas the human-made label did not differ from the unlabeled condition, providing empirical evidence for an implicit human-made default assumption. Perceived effort increased strategic curation intentions via both rational and normative pathways. However, the AI-generated label weakened both pathways, producing an asymmetric dual-path effect that systematically undermined users’ willingness to intervene in algorithmic curation.DiscussionThe effort devaluation induced by AI labeling operated as a context-independent heuristic, unaffected by content type or perceived platform degradation. Moreover, an exploratory finding showed that greater algorithmic knowledge was associated with lower intervention intention, suggesting that user agency may be grounded less in technical knowledge than in subjective efficacy beliefs. This implies that future labeling policies should move beyond passive risk disclosure toward a human-centered value-certification framework that foregrounds human creativity and effort.
Source: Human-made vs. AI-generated: how provenance labels drive strategic curation via perceived effort