Recent national discussion about artificial‑intelligence water consumption has often cited a claim that each ChatGPT prompt uses a full bottle of water. Independent researcher Andy Masley, a former head of Effective Altruism DC, says that figure is inaccurate. A typical ChatGPT request—such as drafting an email—uses roughly 6 mL of water, about one percent of a standard bottle.
Why the myth spread
The exaggerated statistic gained traction because AI’s inner workings are opaque. Large tech firms rarely disclose the exact cooling methods or energy sources for their data centers, making precise calculations difficult. Media outlets amplified the claim, and organizations like the Environmental and Energy Study Institute repeated it as fact.
What the data actually show
In 2024, the Washington Post partnered with University of California, Riverside researchers to estimate water use for a 100‑word ChatGPT response. Their estimate was just over 500 mL—still far less than a bottle. Since then, EcoLogits, a public tracker of AI energy use, updated its calculator. For the current free model, ChatGPT 5.5, the tool estimates 6.11 mL per email‑length prompt, confirming Masley’s earlier figure.
Comparing AI to food production
While AI’s water use is modest, the water needed to produce meat is staggering. The Water Footprint Network reports that producing one kilogram of beef requires about 925 liters of water, and pork needs roughly 1,078 liters per kilogram. By contrast, a single chatbot query consumes only a few milliliters. The disparity grows when considering the average American diet: in 2023, each person ate about 82 pounds of beef and 64 pounds of pork annually.
Animal agriculture also impacts water quality. Manure and fertilizer runoff from livestock farms pollutes nearby streams and rivers, adding a hidden cost to the water footprint that pure volume measurements miss.
Where the water actually goes
Data centers cool their servers with water because it removes heat more efficiently than air. The amount of water required depends on factors such as query length, ambient temperature, and the proximity of cooling water sources. Moreover, the water used to generate electricity for a data center may be drawn from a power plant located in a different state, meaning the water impact can be felt far from the physical data center.
What consumers can do
Masley argues that the focus on AI’s water use distracts from higher‑impact choices. Reducing meat consumption, especially beef and pork, can dramatically lower an individual’s water footprint. Plant‑based foods like soy, potatoes, and bananas require less than 20 % of the water needed for the same weight of beef.
While AI developers continue to improve energy‑efficiency—Google reported a 33‑fold improvement in energy use for a typical Gemini text request between 2024 and 2025—consumers should keep an eye on larger, everyday decisions that affect water resources.
Bottom line
The myth that a single AI prompt wastes a whole bottle of water is unfounded. The real water challenge lies in the foods we place on our plates. By understanding the true scale of water use in meat production, families can make more informed choices that protect limited water supplies for future generations.
Original reporting: El Paso News (HLL/CB) — read the source article.
