Tale of Two Storylines：An NLP-assisted Corpus Study of Open AI’s Identity Construction

Every organization has a story it tells about itself. It is the answer to the question of who they are, what they do, and why they matter. For decades, researchers have studied how companies craft this identity, mostly by looking at their formal, official documents like annual reports or social responsibility statements. These are the polished, carefully edited versions of a company's life, designed to look good to investors and regulators. But in the modern world, much of a company's daily conversation happens in a different place: the blog. These are the informal, frequent updates where a company explains its latest experiments, hires new people, or reacts to news. They are less scripted and often reveal a more complex, sometimes contradictory, picture of who the organization really thinks it is. Understanding this gap between the formal image and the informal voice is crucial, because it is in these everyday words that a company's true character is often revealed to the public.

A team of researchers from Sichuan International Studies University decided to investigate this phenomenon using a specific, high-profile example: OpenAI, the organization behind some of the most famous artificial intelligence tools in the world. They wanted to see how OpenAI constructs its identity not just in its official press releases, but in the hundreds of blog posts it has published over nearly a decade. To do this, they used a method that combines computer science with the study of language. They gathered every blog post OpenAI had published from December 2015 to March 2024, a collection of 118 articles containing roughly 435,000 characters. Instead of reading every single word manually, they used a computer program to analyze the text. This program, known as a natural language processing tool, looked at how words appeared together in the text to find hidden patterns. It grouped the blog posts into three main clusters based on their content and then examined the specific words that OpenAI used most often when talking about itself.

The analysis revealed that OpenAI tells a very clear, three-part story on the surface. The first part of this story paints the organization as a guardian of safety and human benefit. In their writing, OpenAI frequently links its mission to the idea of protecting people from the risks of advanced technology. They describe themselves as ethical stewards who are building artificial intelligence that is safe, reliable, and designed to help all of humanity. The second part of the story presents them as a cutting-edge innovator. Here, the language focuses on research, engineering, and academic breakthroughs. The blog posts highlight their work with scholars, their technical goals, and their continuous efforts to improve their models. The third part of the story casts them as a shared advocate. In this narrative, OpenAI emphasizes collaboration, openness, and making their technology available to others through tools like application programming interfaces. They position themselves as a partner to developers and a bridge between complex research and the wider world.

However, the researchers found that beneath this polished, three-part narrative lies a second, quieter storyline. This underlying story is not explicitly stated but is revealed through the tensions and contradictions in the language. While OpenAI speaks loudly about safety and privacy, the researchers noticed a specific contradiction regarding data usage. In a blog post about "ChatGPT for Teams," the company highlighted that "No training on your business data or conversations" would occur. However, the researchers argue that this statement obscures the fact that user data is still used to train the broader GPT models. This creates a hidden identity the researchers call a "privacy peeper," where the drive for innovation seems to occasionally overshadow the protection of user information. Similarly, while the public story is one of open collaboration and shared benefits, the underlying text reveals a different dynamic regarding power. The researchers identified a hidden "technology monopolist" identity. This emerges from the way OpenAI talks about its exclusive partnerships, its shift from a non-profit to a for-profit structure, and its growing control over the artificial intelligence market. The language of open access sometimes sits uneasily alongside descriptions of exclusive agreements and the consolidation of resources.

The study suggests that these two storylines—the public face of the benevolent guardian and the private reality of the powerful monopolist—are not just accidental. They are woven together in the company's daily communication. The researchers argue that this dual narrative is a common feature of how modern organizations manage their identity. They present a unified, positive image to the world while simultaneously navigating the complex, often conflicting realities of running a high-stakes technology company. The findings indicate that a company's true identity is not a single, fixed thing, but a fluid construction that changes depending on the context. By looking at the informal blog posts rather than just the formal reports, the researchers were able to see the full picture of how OpenAI positions itself: as a hero of safety and innovation, but also as a powerful entity grappling with the very real pressures of market dominance and data control.

This research highlights an important lesson for how we understand the companies that shape our digital lives. It shows that identity is not just what a company says in its annual report; it is also what it says in its blog posts, its updates, and its casual explanations. The study suggests that for a company to maintain trust, it needs to ensure that these different voices tell a consistent story. If the informal, day-to-day communication contradicts the formal promises, it can create a gap in public perception that is hard to bridge. For OpenAI, and for any organization operating in the fast-moving world of artificial intelligence, the challenge is to align their public narrative with their operational reality. The researchers conclude that paying attention to these informal texts is essential for anyone who wants to understand the true character of an organization, because it is in the small, everyday words that the most honest version of a company's identity is often found.