OpenAI says nearly 1.3M weekly users are discussing “advanced topics in hard science” in January, with an average of 8.4M ChatGPT messages on those topics
AI is increasingly being used as a research collaborator for mathematicians and scientists, per a new report from OpenAI shared exclusively with Axios.
Context & Ripple Effects
This is a more granular view of ChatGPT’s expanding usage base: OpenAI had previously reported more than 200M weekly ChatGPT users and later said users were sending over 2.5B prompts per day globally. The hard-science cohort matters because it identifies research-oriented work as a measurable use case within that consumer-scale distribution.
OpenAI’s reported audience growth to 700M weekly users had established reach; this disclosure adds evidence that some of that reach is being applied to technically demanding inquiry rather than only general-purpose assistance.
First-order effects
- Researchers and mathematically or scientifically focused users gain evidence that ChatGPT is already being used by a substantial peer cohort as a research collaborator.
- OpenAI gains a defined high-value usage segment, with message volume that can inform how it evaluates and prioritizes science-oriented interactions.
Second-order effects
- Competing AI providers targeting researchers will face greater pressure to show credible performance and usability on advanced technical work, not merely broad consumer adoption.
- Research organizations and individual scientists may put more weight on AI tools’ ability to support iterative technical dialogue, raising the importance of reliability and domain-specific validation in procurement and use.
Third-order effects
- If sustained, this pattern would move general-purpose AI further into knowledge-work infrastructure for science, while making verification, attribution, and human review central constraints on adoption.
- The market may increasingly differentiate models by their effectiveness on useful expert tasks rather than headline user counts alone, though usage volume does not by itself establish research quality or scientific outcomes.
The trend: Consumer-scale AI platforms are becoming workflow tools for specialized professional and research tasks, with value increasingly judged by performance in high-stakes domains.