Follow ZDNET: Add us as a favorite source On Google.
Key takeaways of zdnet
- Most Chatgpt is the use of non-work, which focuses on writing works.
- Cloud is used more specifically for coding for automation.
- AI adoption is uneven, with the first benefit in rich areas.
Two of the largest AI companies have revealed how people use their models.
Also: 2025’s best AI Chatbots
Openai released a research paper analyzing millions of Chatgpt conversations, while Anthropic published its third economic index report on cloud use. Both have a lot of charts and data, but also some interesting details about AI and why. Together, they provide an attractive look of how AI is crawling in our work, school and everyday life.
(Disclosure: ZDNET’s original company Ziff Davis filed a case of April 2025 against Openai, alleging that it violates Ziff Davis copyright training and operating its AI system.)
Most chat chats are to ask about non-work items
New study of Openai, How do people use chatAnalyzed a random sample of conversations from May 2024 to June 2025. First surprise: Non-work use has overtaken the use of work. In June 2024, about 47% of the Chatgpt message belonged to the work, but by June 2025, it fell to 27%-even the daily messages increased from about 451 million to 2.6 billion in the same period.
Now top categories? Practical guidance (which includes tuition, teaching, and how-to), information (about people, current events, products and dishes), and writing (email, documents, and other communications). According to Openai’s research, it is collectively “about 80%of all conversations”. The use of coding is 4.2%.
Too: The best AI for coding (including a new winner – and what not to use)
Therefore, while the world (and your boss) may have imagined people to use chats through tedious work work or to fly for computer programming, the majority are really changing it for help, advice, learning and writing.
Researchers at Openai classified user interaction into three buckets: ask (seeking information or guidance), doing (producing output or completing tasks), and expressing (sharing opinions or emotions). They guess that about 49% are asking messages, 40% are, and 11% are expressed. By July 2025, 56% of the work messages were in the doing category.
Too: Chatgate Plus still has a price of $ 20 when free version provides so much?
In fact, when Chatgpt is used at work, more than 40% of that use is for writing-and out of it is more than two-thirds of the editing texts that people have written themselves. So when people turn to chat at work, it is more for proofing than generating a complete lesson from scratches.
Most cloud chat automation instructions are (for coding)
Anthopropic Economic index See how the cloud is used by both consumers and companies. The report divides the tasks into two categories: automation (where the cloud completes the work with a little input – either with instructions, minimal conversations, or response loops, where users rela -relax results), and growth (where humans and AI work together through learning, recurrence and verification).
The report randomly analyzed the tapes of the sample and found that the director conversation – where a user asks Cloud to complete a task with minimal conversation – increased ranging from 27% (December 2024 to January 2025) to 39% (February to March 2025).
Too: Work work people use Cloud AI according to anthropic
Since last December, people are handing over the entire tasks to the cloud with at least-asking something to do and rely on the output. First time, anthropic Reports “Automation (49.1%) has become more common than an increase (47%).” It says that it may be “AI due to” AI being responsible for winning the confidence of users and fulfilling the sophisticated work. “
In enterprise settings, in particular, automation is common. Anthropic states that its 77% API customer has a “automation pattern” in conversations, with “vast majority” direction. So, while users of about half everyday use Claude.ai To complete the work with minimal conversation, businesses are even more relying on the cloud to insert tasks on autopylot – especially coding.
Anthropic stated that about 44% of API belongs to traffic coding (computer and mathematical functions), while compared to 36% on Cloud.AI. It said that Cloud.AI is more used for educational and writing works.
Who is using AI and where?
The use of Chatgpt with 700 million weekly users worldwide until July 2025, in the last one year, has been relatively sharp in lower and medium-oriented countries, notes the report notes. While about 80% of the early Chatgpt users were male, the share has fallen to 48% (until June 2025), more active users now have “usually women” first names.
The user base also sketes the youth, with about half the messages from adults coming from people under 26 years of age. Researchers also said that the possibility of using Chatgpt for educated and highly paid professional work is “quite high”.
Too: In a way Millennials defeated Jean Z in AI adoption
For cloud use, Anthropic stated that an account for the American highest shares (21.6%). However, Israel moves forward when it comes to “global per capita cloud use”. The nation has a “AI using index” (AUI) score of 7, which means that people in Israel use clouds almost seven times more than expected depending on the size of the country’s working population.
In the US, California is a leading state in terms of use, which is 25.3% of all cloud activity, mostly for IT functions. But when you adjust to the size of the population, the per capita-use leader becomes Washington, DC, with a 3.82 AUI score. People there clearly use the cloud about four times more often, which suggest more often than their part of the American working age population.
And why are DC people using cloud so much? “Document editing, information provisions and job applications,” anthropic.
Too: Most developers use AI in their daily workflows – but they do not rely on it
Increasing concerns about inequality, anthropic warned of uneven AI adoption, reflecting the history of electricity and internet, where rich countries pulled forward developing people. The report stated that “1% more GDP per capita was associated with 0.7% high anthropic AI use index,” can already occupy most productivity benefits, suggesting rich areas.
“In high-or-age countries, people are more likely to use clouds, more likely to cooperate than automation, and more likely to pursue the width of uses beyond coding,” anthropic. Per capita lower AI adopted countries in India, with more than half of all uses in India, focus highly on coding tasks, compared to almost one -third at the globally.
General subject
The report of both Openai and Anthropic sees how people interact with AI: asking to ask.
For example, Anthropic focused on whether the cloud was used for automation or increase and found that users prefer to give instructions with automation and small input rather than collaborating with models. Meanwhile, Openai broke the user’s interaction, or broke into expressing categories, and found that the users use it to ask.
Too: The fastest growing AI chatboat recently? It’s not chat or Gemini
Both studies also saw that people use AI for AI, such as writing or coding. While chatgpt and cloud are still used to write, the Chatgpt users often ask AI to edit non-functioning texts he himself wrote. Cloud, meanwhile, is more used for automation, especially coding, it is directed to complete it with businesses.
Add all this, and you get an interesting snapshot of AI so far and how it varies among the contestants.
