Despite paying for the same subscription, a small group of ChatGPT users is extracting up to seven times more value from the technology than the average user, in what experts terms as a classic case of capacity overhang.
What this brings forth is the striking gap in how users engage with Artificial Intelligence (AI) tools, even when they are paying for the same subscription.
According to a recent OpenAI report, users on the same ChatGPT Plus plan are not using the Large Language Machine (LLM) equally. The top 5% of users, referred to as ‘power users’ consume nearly seven times more advanced AI ‘thinking capabilities’ than the median user. Even at the level of individual prompts, these users extract almost three times more value per interaction.
Beyond Basic Writing Assistance
This capability overhang presents the disconnect between what AI systems can do and how people actually use them. “The gap is not about access, but about usage,” the report suggests. While millions have access to the best class AI tools, most continue to use them for relatively simple tasks such as basic queries or casual assistance.
The divide becomes even more pronounced at the level of tasks. For the majority of users, ChatGPT functions mainly as a writing assistant, used to draft emails, summarise documents, or generate content. These interactions are mostly linear and require limited iteration, drawing relatively less on the model’s deeper reasoning capabilities. Writing and editing tasks account for a significant share of overall usage, reflecting high-frequency but relatively shallow engagement.
In contrast, power users are far more likely to deploy the same tools for coding and technical problem-solving. Programming-related tasks, such as debugging, building scripts, or developing applications form a smaller share of total usage but demand substantially more from the system. These interactions are inherently iterative, often requiring multiple rounds of refinement, and consume far greater thinking capability per task.
Productivity Divide
Industry executives say that this divide portrays that individuals and organizations that learn to integrate AI into their workflows stand to gain significant productivity advantages, while others risk falling behind despite having access to the same tools. To add, this difference becomes even more pronounced among OpenAI’s own employees, who use up to 15 times more of these advanced capabilities per user and over five times more per message than the average Plus subscriber, the report suggested.
The idea is that organisations no longer need to concentrate focus on building more powerful models of AI, rather to ensure that people know how to use what already exists.