ChatNetZero Boosts Climate Transparency with New Updates

Climate chatbot now shows the energy cost of every question

ChatNetZero has become the first AI chatbot to display the energy consumption of each individual query. Users can now see exactly how much power their questions about climate targets and net-zero pledges consume in real time. This matters because AI systems use substantial amounts of electricity, yet most tools hide these costs from view.

The chatbot specializes in answering questions about decarbonization commitments from businesses and governments. Developed by DDL and Arboretica with Net Zero Tracker, it draws on verified data covering over 4,000 organizations worldwide. These include nearly 200 countries, major cities, and the 2,000 largest publicly listed companies.

Unlike general-purpose AI tools that sometimes provide inaccurate information, ChatNetZero uses a technique called retrieval-augmented generation. This approach pulls answers from a curated database of expert-vetted sources. Consequently, every response includes citations to original documents. An academic evaluation published in August 2024 confirmed the system delivers reliable information on both specific organizations and general climate policy questions.

New transparency features address AI energy concerns

The recent update introduces two significant changes. First, the system now retrieves only the specific documents needed to answer each question, reducing unnecessary computation. Second, it displays an estimated energy figure alongside every response.

This addresses a common criticism within climate research communities. Many experts have questioned whether energy-intensive AI tools should be used to discuss climate action at all. By showing energy consumption directly, ChatNetZero allows users to make informed decisions about when queries justify their environmental cost.

The developers describe the update as strengthening both transparency and efficiency. Users can see the trade-off between getting information and consuming electricity. This visibility represents a departure from standard practice across the AI industry, where energy costs remain largely invisible to end users.

What the energy figures actually mean

A typical text query to modern AI systems consumes between 0.24 and 0.4 watt-hours of electricity. For context, that sits somewhere between a brief smartphone notification and a few seconds of laptop use. However, these small amounts add up quickly when millions of people use AI tools daily.

Epoch AI reported in February 2025 that a standard GPT-4o query uses approximately 0.3 watt-hours. This represents a substantial improvement from 2023 estimates of 3 watt-hours per query. The reduction comes from more efficient model designs and newer NVIDIA H100 graphics processors in data centers.

OpenAI’s Sam Altman stated that a typical ChatGPT text query consumes 0.34 watt-hours. Google’s Gemini uses slightly less, with a median of 0.24 watt-hours. Recent benchmarking published in ArXiv found short GPT-4o queries use 0.42 to 0.43 watt-hours, roughly 40% more than a Google search at 0.30 watt-hours.

These figures vary based on several factors. Longer prompts require more processing. Complex questions that need multiple reasoning steps consume more energy. Different AI models have different efficiency profiles. Data center infrastructure and cooling systems also affect the total power draw.

Carbon emissions from individual queries

Energy consumption translates to carbon emissions depending on how electricity is generated. Using a baseline of 0.34 watt-hours per query, standard calculations estimate approximately 0.16 grams of CO₂ per prompt. This assumes a Power Usage Effectiveness multiplier of 1.3 for data center infrastructure and US grid carbon intensity of 367 grams of CO₂ per kilowatt-hour.

Academic research published by Jegham and colleagues in 2025 arrived at a similar figure of 0.15 grams of CO₂ per query. Individual queries produce minimal emissions. However, scale changes the picture dramatically.

ChatGPT serves approximately 87.5 million users as of April 2025. If each user makes an average of eight queries per day, that totals 700 million daily queries. Over a year, this accumulated usage creates substantial energy demand, water consumption for cooling, and carbon emissions. Therefore, small per-query figures become significant when multiplied across global usage patterns.

How ChatNetZero differs from general AI tools

Most large language models draw on vast training datasets that include unreliable or outdated information. This leads to what researchers call hallucinations, where the AI generates plausible-sounding but factually incorrect responses. For climate policy questions, this creates serious problems because inaccurate information about net-zero commitments can mask greenwashing or misrepresent actual progress.

ChatNetZero addresses this by limiting its scope deliberately. It only answers questions using its curated database of verified climate commitments and policy documents. When it lacks relevant information, it says so rather than guessing. Each response links directly to source documents, allowing users to verify claims independently.

This focused approach serves researchers, policymakers, and businesses trying to assess climate commitments accurately. For example, a procurement team evaluating supplier climate claims can ask specific questions about company net-zero targets and receive answers grounded in publicly available corporate documents. Similarly, policy analysts can compare commitments across different countries or regions using consistent criteria.

Coming improvements expand the chatbot’s capabilities

The development team has outlined several planned enhancements. The first adds live internet connections to access the most current data. Climate commitments change frequently as governments update policies and companies revise targets. Real-time access ensures users get up-to-date information rather than relying solely on periodic database updates.

Another planned feature allows users to upload their own policy documents for benchmarking. A company could upload its sustainability report and ask how its commitments compare to industry peers or regional standards. This turns the chatbot into a comparative analysis tool, not just an information retrieval system.

The team also plans to expand document coverage. Currently, the database focuses on large organizations and national governments. Future versions will include smaller companies, local authorities, and sector-specific initiatives. This broader coverage will make the tool useful for a wider range of queries about climate action across different scales and contexts.

Essential facts about AI energy disclosure

• ChatNetZero displays estimated energy consumption for every query, making it the first climate-focused AI tool with transparent environmental metrics.
• A typical text query to modern AI systems uses between 0.24 and 0.4 watt-hours, with GPT-4o averaging 0.3 watt-hours per standard question.
• Each query generates approximately 0.15 to 0.16 grams of CO₂ emissions based on current US grid carbon intensity and data center efficiency factors.
• The chatbot draws on verified data covering over 4,000 entities including countries, major cities, and the world’s largest companies with public climate commitments.
• Retrieval-augmented generation ensures responses come from curated sources rather than general training data, reducing inaccuracies in climate policy information.
• Planned features include real-time data updates, document upload for benchmarking, and expanded coverage of smaller organizations and local authorities.

Why transparency matters for business users

For UK businesses working on net-zero compliance, this development carries practical implications. Many organizations now face requirements to report on supply chain emissions under regulations like PPN 06/21 for public sector suppliers. Checking supplier climate commitments quickly and accurately helps meet due diligence requirements without extensive manual research.

Visible energy costs also inform decisions about when AI tools add genuine value. A sustainability manager might use ChatNetZero to verify specific claims in a tender response, where accuracy justifies the query. Conversely, casual browsing or repeated similar questions might not warrant the environmental cost. This mirrors broader decisions businesses make about resource use and efficiency.

The transparency principle extends beyond individual queries. As AI adoption grows across business operations, understanding aggregate energy impacts becomes important for Scope 2 emissions reporting. Companies using AI extensively may need to account for this consumption in their carbon footprints. Tools that disclose energy use help businesses quantify these impacts accurately.

ChatNetZero also serves as a fact-checking resource against greenwashing. When suppliers or competitors make climate claims, businesses can verify these statements against documented commitments. This proves particularly valuable in procurement decisions where sustainability criteria influence contract awards. Having reliable information about actual commitments rather than marketing language supports better commercial decisions.

Setting expectations for the AI industry

This feature may influence how other AI developers approach environmental disclosure. Currently, most AI services provide no visibility into their energy consumption or carbon footprint. Users have no basis to compare different tools or understand their environmental impact. ChatNetZero demonstrates that technical barriers to disclosure are surmountable.

The climate technology sector faces particular scrutiny on this point. Tools designed to address environmental challenges lose credibility if their own resource consumption remains hidden. By modeling transparent practices, ChatNetZero establishes a baseline expectation for similar services. This could accelerate adoption of energy disclosure across climate-focused AI applications.

Broader AI industry adoption seems less certain. Large commercial platforms may resist disclosure due to competitive concerns or complexity in attribution. Energy consumption per query depends on infrastructure, model version, and operational factors that vary across providers. Nevertheless, growing regulatory interest in AI environmental impacts may eventually mandate some form of transparency regardless of voluntary industry action.

For businesses evaluating AI tools, this sets a useful precedent. When comparing similar services, energy efficiency and transparency can become selection criteria alongside accuracy and cost. This particularly matters for organizations with strong sustainability commitments or reporting requirements that include digital infrastructure emissions. Vendor questions about energy consumption and disclosure practices now have a reference point.

Further information and technical resources

ChatNetZero is accessible at its dedicated website for anyone researching climate commitments and net-zero pledges. Net Zero Tracker, one of the collaborating organizations, maintains comprehensive data on climate commitments at the national and subnational level. Their research underpins much of the chatbot’s knowledge base.

For technical details on AI energy consumption, Epoch AI publishes regular analysis of model efficiency and data center performance. Their February 2025 report provides current benchmarks for major AI systems. The ArXiv preprint server hosts academic papers on AI environmental impacts, including detailed benchmarking studies comparing different models and query types.

Businesses needing support with carbon reporting and net-zero compliance can access guidance through our compliance services covering ESG reporting and regulatory requirements. For organizations developing procurement criteria around supplier sustainability claims, our sustainable procurement support helps evaluate environmental commitments in tender processes. Understanding how to verify climate claims accurately matters increasingly for public sector suppliers and their supply chains.