Before AI becomes more powerful, it must first sound powerful. Leading AI models like those from OpenAI, Meta, and Mistral are described using human like terms such as “thinker,” “reasoner,” and “helper.” These descriptions are more than just communication tools; they shape how much intelligence, trust, and authority people are willing to attribute to machines.
A recent study, “Constructing AI power: anthropomorphising imaginaries of generative AI by AI companies,” published in AI & Society, examines how these companies use such language. The study, authored by Anastasia Glawatzki from the University of Münster, reveals that presenting AI as human like helps build visions of its capabilities and roles in society.
Glawatzki analyzed 35 corporate documents and found a progression in how AI is framed. Initially, AI is seen as a productivity tool. Over time, it evolves into a collaborator and, ultimately, an increasingly autonomous actor. This shift in language transforms technological ambition into social expectations.
The study highlights how cognitive descriptions like intelligence, understanding, and reasoning are often attributed to AI systems. This language subtly changes the perception of the technology, making it easier to relate to familiar human categories.
According to Glawatzki, anthropomorphism operates beyond just communication. It establishes expectations about AI’s capabilities and trustworthiness, influencing how users interact with the technology. Corporate descriptions can preconfigure the user experience before they independently assess the technology’s limitations.
The study identifies three recurring visions of generative AI:
1. Instrumental AI: AI is seen as a tool with human capabilities, enhancing productivity and efficiency. This framing is described as instrumental, as it remains under human control but is increasingly valued for its supposedly human like performance.
2. Collaborative AI: Here, AI is portrayed as an empowering assistant, capable of supporting users in both professional and personal tasks. This relationship produces cooperative or collaborative AI power, where humans guide the system while AI expands what people can accomplish.
3. Autonomous AI: This vision envisions AI as moving beyond assistance, toward expertise and greater autonomy, potentially exceeding human capability. This framing shifts power from human control to AI self optimization and independent decision making.
Beyond marketing, these narratives can normalize AI’s expansion into more areas of life, narrowing the space for questioning the desirability of every application. For example, productivity narratives can redefine non-use, making AI adoption less of an option and more of a requirement. This can strengthen the market position of companies supplying AI infrastructure.
The study also warns of potential downsides. Claims about democratizing creativity or expertise may obscure unequal access, labor displacement, and the human work involved in training and maintaining AI systems. Additionally, anthropomorphic language can complicate accountability. Describing AI as independently reasoning or learning can obscure the human organizations designing and deploying it.
Companies can portray systems as increasingly autonomous and potentially dangerous while simultaneously presenting themselves as the best positioned to build safeguards. This creates a narrative where technological power and corporate authority reinforce each other.
The implications extend beyond marketing. Regulators, businesses, and public institutions need to examine the language through which capability claims are communicated, especially where terms like intelligence, understanding, expertise, or autonomy may influence trust and delegation.
For governments procuring AI systems, this raises practical questions about how vendor claims are evaluated and whether anthropomorphic terminology obscures limitations. Organizations using AI in education, healthcare, public administration, or professional services may also need clearer boundaries between conversational design and demonstrated competence.
The study also raises important questions for developing economies, where the reliance on externally developed AI infrastructure could involve more than just technology imports but also assumptions about productivity, expertise, and the social role of machines. Future research could explore whether ideas about AI intelligence, autonomy, and human machine relationships are globally converging or remain shaped by different cultural and institutional contexts.