Excitement, fear, distrust: ChatGPT's debut fueled global emotional reckoning

A new study sheds light on the emotionally charged landscape surrounding the release of ChatGPT, offering the most detailed empirical portrait to date of how the public reacted to one of the most transformative AI technologies ever launched. The findings, based on over 5 million tweets, show that the reaction to OpenAI’s generative AI chatbot was far from uniformly enthusiastic, with excitement frequently intertwined with fear, anxiety, and ethical unease.

The study, titled “The Emotional Landscape of Technological Innovation: A Data-Driven Case Study of ChatGPT’s Launch”, is published in Informatics (2025, Vol. 12, Article 58). Going far beyond traditional sentiment analysis, the researchers used emotion classification hierarchies and topic modelling to map not just the public's emotional polarity but the specific psychological responses triggered by ChatGPT’s release.

How did users really feel about ChatGPT? 

The study identifies a complex and evolving emotional response among early adopters. While the initial launch in November 2022 sparked a wave of enthusiasm, dominated by expressions of joy, love, and fascination, this quickly gave way to a more conflicted emotional climate. Users began expressing anxiety, suspicion, frustration, and even despair over ChatGPT’s potential societal consequences.

Of the 5.26 million tweets analyzed between November 2022 and May 2023, over 1 million contained at least one identifiable emotion, with many containing multiple overlapping emotional cues. Through a detailed lexical mapping using the WordNet-Affect hierarchy, the researchers identified 33 distinct emotion categories. These ranged from optimistic sentiments like “Enjoyment/Liking” and “Glee/Joy/Jubilance” to darker tones of “Afraid/Fear,” “Ashamed/Disgraced/Shame,” and “General Dislike.”

Chronological tracking revealed that emotional reactions fluctuated throughout the study period. While enjoyment and surprise spiked shortly after release, sustained expressions of anxiety and distrust persisted throughout the dataset. These findings support the idea that user emotions toward AI adoption are not fixed but dynamic and highly context-dependent.

What triggered positive and negative reactions?

To understand the drivers behind these emotions, the study incorporated topic modelling using Latent Dirichlet Allocation (LDA). This allowed the researchers to connect specific emotional responses to topics that users were discussing.

Topics that elicited excitement and affection typically involved ChatGPT’s perceived utility, especially its capabilities in creative writing, customer service, and productivity. For example, tweets associated with themes like “AI-generated blog content” and “ChatGPT’s role in transforming retail” generated strong positive reactions. These sentiments align with innovation diffusion theory, where early adopters embrace technologies that solve existing problems or enhance daily tasks.

By contrast, negative emotions were most strongly associated with conversations about job displacement, academic integrity, surveillance, and misinformation. Themes like “AI replacing humankind,” “disruption in education,” and “ethical risks of AI-generated text” generated high volumes of fear, anxiety, and moral discomfort. The combination of hope and dread in these reactions suggests that ChatGPT sits at a critical junction in public trust in artificial intelligence.

The topic–emotion alignment matrix revealed duality even within single themes. For example, the idea of AI replacing human jobs simultaneously triggered “Excitement” among innovation enthusiasts and “Fear” among those concerned about employment insecurity. This bifurcated sentiment underlines the necessity of nuanced emotional analysis when assessing technology adoption patterns.

What are the broader implications for AI governance?

Emotional responses are not just public reactions, they are indicators of acceptance thresholds, trust dynamics, and friction points that could influence long-term technology adoption.

The persistent presence of anxiety and ethical discomfort highlights the need for better communication around the limitations, capabilities, and risks of AI tools like ChatGPT. While traditional tech adoption models often assume a linear progression from novelty to acceptance, this study underscores that emotional volatility can slow or even reverse adoption if not addressed.

The authors recommend that future AI rollouts adopt an emotionally intelligent strategy - one that proactively monitors public sentiment in real time, and incorporates feedback loops to address fears, correct misconceptions, and increase transparency. Emphasizing digital responsibility and public accountability, they argue, can convert fearful resistors into cautious but willing users.

Another implication is the need for emotional risk assessments in digital product deployment. Just as companies model cybersecurity vulnerabilities or environmental impacts, they must also model how innovations might provoke emotional and psychological reactions that affect public trust and reputational stability.

Future studies, as the researchers envision, could expand this emotional monitoring approach to other emerging technologies, or even track how public emotions evolve in response to critical events like feature updates, media coverage, or corporate controversies. Additionally, tracking how emotions correlate with long-term user retention could help tech developers build more sustainable engagement strategies.