Miriam Shor Says She Has Never Used ChatGPT: 'Ask How Much Water It Uses Every Time You Use It'

Miriam Shor has revealed that she has never used ChatGPT — and when news anchors appeared surprised by the admission, the actress had a pointed question of her own.

During the exchange, Shor was asked whether she had ever used ChatGPT. She said she had not and, when asked whether she wanted to try it, indicated that she had no interest in doing so. Instead, she turned the conversation towards one of the increasingly debated issues surrounding generative artificial intelligence: its environmental footprint.

'Let's ask how much water it uses every time you use ChatGPT,' Shor said. The remark has attracted attention online, with the exchange being discussed in a popular Reddit community. The original post focused on the apparent surprise over Shor's admission and highlighted her response about water consumption.

Miriam Shor's Blunt Response to the ChatGPT Question

Shor's comments stand out because ChatGPT has become an increasingly familiar tool for many people, with users turning to AI for everything from brainstorming and writing assistance to research and everyday questions.

The actress, however, indicated that she has never incorporated the chatbot into her routine. Rather than treating her lack of experience with AI as something she needed to change, she questioned the environmental cost associated with the technology.

Her comment touches on a broader conversation that has grown alongside the rapid expansion of generative AI: how much energy and water are required to operate the data centres that power these systems.

How Much Water Does ChatGPT Actually Use?

This is where the issue becomes more complicated than a single viral figure might suggest.

A study by researchers from the University of California, Riverside and the University of Texas at Arlington estimated that GPT-3 could consume roughly 500 millilitres of water for every 10 to 50 medium-length responses, depending on when and where the model was deployed.

The estimate included water associated with data-centre cooling as well as water linked to electricity generation.

Importantly, this does not mean that every ChatGPT prompt consumes half a litre of water. The figure was an estimate for GPT-3, an earlier AI model, and covered a range of responses. Water consumption can vary significantly depending on the model, data-centre location, cooling technology, electricity source and workload.

OpenAI CEO Sam Altman later cited a much smaller figure for an average ChatGPT query. In June 2025, Altman said an average query used approximately 0.000085 US gallons of water, equivalent to about 0.32 millilitres.

However, the two figures should not be treated as directly comparable measurements. They were produced using different approaches and assumptions, and the exact methodology behind Altman's figure was not publicly detailed to the same extent as the academic research.

The difference nevertheless illustrates why claims about AI water consumption require context. There is no single figure that accurately represents the water footprint of every ChatGPT interaction.

Generative AI systems rely on powerful computer hardware housed in data centres. Those machines produce substantial amounts of heat while processing requests and therefore require cooling.

Water can be involved directly through some cooling systems and indirectly through electricity generation. The amount involved depends heavily on how a particular facility is designed, where it is located and how it is operated.

The technology industry is also developing systems intended to reduce freshwater consumption. OpenAI has also described some newer data-centre projects as using closed-loop cooling systems. At its Abilene site, for example, the company says water is recirculated through sealed pipes rather than continuously consumed for cooling.

That means the water footprint of newer AI infrastructure should not automatically be assumed to be identical to estimates based on older systems or different facilities.

The Environmental Debate Around ChatGPT

Shor's question nevertheless points to a genuine environmental debate.

AI's impact is not limited to water. Researchers and technology companies are also examining electricity consumption, carbon emissions, hardware manufacturing and the resources required to construct and operate large data centres.

The environmental impact can also vary depending on the source of electricity and the efficiency of the hardware and infrastructure being used. As AI systems become more widely adopted, researchers are continuing to examine how their environmental footprint changes as computing becomes more efficient while overall demand increases.

At the same time, the rapid growth of AI means that even relatively small resource requirements per individual interaction can become significant when multiplied across enormous numbers of users.