AI-Driven Context Interpretation in Influencer Marketing with ChatGPT Pro: Exploring Semantic Boundaries

This study investigates the ability of ChatGPT o1 Pro, Large Language Model (LLM), to identify and interpret advertising embedded in influencer marketing posts on Instagram. Using a dataset of 100 posts from the beauty, technology, and food sectors, each labeled with the hashtag #reklama, this research evaluates whether the model relies on explicit cues or can discern implicit promotional intent across varying levels of content complexity. Results show that the model does not depend solely on surface-level markers, but its accuracy declines sharply as advertising is more seamlessly integrated into influencer narratives. A strong positive correlation (r = 0.8601) and regression analysis (R² = 0.7583) confirm that complexity is a major predictor of AI misclassification. Logistic regression further demonstrates that the risk of error rises most steeply in technology and food categories. Semantic and contextual errors are most prevalent, highlighting the limitations of current large language models in parsing nuanced and context-rich advertising. These findings underscore the need for industry-specific fine-tuning and multimodal approaches that combine textual and visual cues for more reliable AI-driven marketing analysis. The study informs practitioners and regulators of both the promise and current limitations of generative AI for advertising compliance and campaign monitoring.