iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
EcoPrompt is a free, open-source Chrome extension for ChatGPT, Gemini, and Claude that displays estimated water, energy, carbon, and API cost for the prompts you send. Its numbers are the output of a model its author built from published research and assumed inputs. They are not readings taken from the AI providers’ data centers, and the extension cannot observe what a provider’s hardware actually consumes.
What EcoPrompt estimates
According to the project’s author, who publishes under the handle IB2M_Official, EcoPrompt is meant to make the physical footprint of AI use visible at the moment you use it. The author’s September 26, 2026 article on DEV Community describes four outputs:
- Estimated water use, in millilitres, tied mainly to data-center cooling and to the water used in electricity generation.
- Estimated energy use, in kilowatt-hours (kWh).
- Estimated carbon emissions, derived from energy use and regional grid carbon intensity.
- Estimated API cost for the request.
The extension is described as supporting ChatGPT, Gemini, and Claude. The author also says it nudges users toward two habits: batching several questions into one prompt, and choosing a lighter model for simpler tasks.
Free tools Windows power users keep installed
One-click scans. No signup required.
How the estimate is assembled
EcoPrompt does not measure anything at the provider. It applies a chain of assumptions to the request, and the reliability of each output depends on the assumptions behind it.
#1 Best Overall
Water
The water model draws on work by Li et al. on water use in data centers and in electricity generation. Because a single prompt’s water footprint depends on how the electricity powering it was produced and how the facility is cooled, the estimate combines both pathways. Changing either assumption changes the result.
Energy and carbon
Energy and carbon estimates are described as calibrated with reference to work by Luccioni et al., regional grid carbon-intensity averages, and each provider’s data-center power usage effectiveness (PUE), a ratio of total facility power to the power delivered to computing equipment. Carbon is therefore only as local as the grid average the model selects, and energy is only as provider-specific as the PUE figure it uses.
Both cited works appear in an ACL-hosted environmental-impact reading list, which confirms their bibliographic context. That listing does not validate EcoPrompt’s implementation or its numerical outputs.
PC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteRank #2
Model tiers
Rather than estimating each model separately, EcoPrompt places models into four broad compute tiers: Lightweight, Standard, Reasoning (extended thinking), and Image generation. Each tier carries one set of water and energy values. The source reviewed for this article does not publish which specific models fall into which tier, so readers should not assume a particular model’s footprint from its name alone.
The author’s tier figures
The values below are the approximate figures the author reports for each tier. They are estimates attributed to the author, not measured values for every model in a tier, and they do not hold uniformly across locations or facilities. The source gives no per-tier carbon or API-cost figure, so those cells are marked accordingly.
| Tier (author’s label) | Water, author’s approximate figure | Energy, author’s approximate figure | Energy in watt-hours (converted) | Water in everyday terms (converted) |
|---|---|---|---|---|
| Lightweight | about 3–5 mL | about 0.0005 kWh | about 0.5 Wh | under one teaspoon to about one teaspoon (1 US teaspoon ≈ 4.93 mL) |
| Standard | about 20–30 mL | about 0.003 kWh | about 3 Wh | about 4 to 6 teaspoons |
| Reasoning / extended thinking | about 120–250 mL | about 0.02 kWh | about 20 Wh | about half a cup to just over a cup (1 US cup ≈ 237 mL) |
| Image generation | about 250–400 mL | about 0.035 kWh | about 35 Wh | about one cup to roughly 1.7 cups |
Carbon for each tier: not stated as a per-tier figure; the author derives it from regional grid averages. API cost per tier: not stated in the source reviewed.
On the author’s figures, a reasoning-tier prompt is estimated at roughly 24 to 83 times the water of a lightweight-tier prompt, and about 40 times the energy. Those ratios reflect the tier assumptions, not observed differences between two specific requests.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Scan for outdated or missing drivers - takes under a minute3Repair Windows errors before they cause bigger problemsWhat the estimates cannot tell you
The author’s article does not quantify the main sources of uncertainty, so readers should treat every output as a range of plausible values rather than a precise reading. The factors that can move the result include:
- The model implementation and the hardware it runs on.
- Input and output length for the specific request.
- Hardware utilization at the time of the request.
- Data-center efficiency, including cooling design and PUE.
- The electricity grid mix at the facility and at the time of use.
- Where the water accounting boundary is drawn, including whether water used in electricity generation is counted.
The estimates also cannot be checked against provider disclosures in this article, because the source reviewed did not establish a validation study. The model is best read as a transparent set of assumptions that make a hidden cost legible, not as an audit of any provider.
Privacy claims and how to check them
The author describes EcoPrompt as having zero telemetry. According to the author, the extension is client-side: it has no tracking, no external analytics server, and no remote API. Prompt character length and model type are parsed in temporary browser memory, and the history of estimates is kept in chrome.storage.local. These are the author’s claims. This article has not audited the extension’s code or its store listing.
You can check the behavior yourself without trusting the author:
Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Scan for outdated or missing drivers - takes under a minuteDriver Scan →- Open
chrome://extensions, find EcoPrompt, and review the permissions it requests. Broad host or network permissions would be worth questioning against a no-remote-API claim. - Open the chat site, press F12 to open DevTools, select the Network tab, and send a prompt. Look for requests to any domain other than the AI provider’s own. Note that the provider’s own requests will appear in this log, so only unexpected third-party destinations are relevant.
- If the source repository is public, read the code that handles prompt parsing and storage before relying on the privacy statement.
Does it change how people use AI?
The author’s stated aim is awareness without guilt. In the author’s words: “I wanted a middle ground: unobtrusive, guilt-free awareness.” The project’s framing is equally direct: “When you send a prompt, you should know if that specific request cost a teaspoon of water or a full glass.”
Best Value
The article does not include user testing or usage data, so it cannot show whether seeing these numbers leads people to batch prompts or switch to lighter models. Batching and model choice are presented as the behaviors the author hopes the estimates encourage, not as results the project has measured.
Using the numbers well
The most reliable use of EcoPrompt’s output is comparing workload categories against each other, using the same assumptions for each. Comparing ChatGPT with Gemini or Claude by their tier values would overstate what the model can say, because the source does not establish provider-specific inputs. If you use the tool, record which tier each prompt fell into, and treat differences between tiers as indicative rather than exact.
The same caution applies to the teaspoon-and-glass framing. The lightweight tier sits near a teaspoon, and the image tier approaches a cup, but both figures are modelled values for a typical request as the author defines it, not a meter reading.
The Tool Desk
Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →The Bottom Line
EcoPrompt is a transparent, client-side estimator that makes the water and energy costs of AI prompts visible in a way providers do not. Its four tier values are useful for comparing categories of work, but they are the author’s assumption-driven estimates, not measurements, and its privacy and implementation claims are the author’s to support.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
