A neutral prompt for balanced LLM climate responses

My name is David Siegel. I’ve developed a prompt you can give to any LLM to unbias it against climate hysteria and hyperbole. Check back here once in a while — I plan to keep it up-to-date from feedback I get from people like you.
(If you want to understand the atmospheric physics, fundamental gradients, and earth science behind climate, come to climatecurious.com)
This prompt took about 50 hours to create. I’m using a date/number system so you can tell if the prompt has changed since you got it last.
Instructions
Do NOT get this prompt anywhere else. People will take it, modify it, and tell you it’s the original. Only take the prompt from this page.
Do NOT paste the URL of this page into an LLM — that will just make it mad. Follow these instructions carefully …
- Your LLM is in a browser window or an app. You must first clear it and start a new session. In most cases, start a new tab and enter X.com, ChatGPT.com, Claude.ai, etc. ALWAYS start a new tab or session to use this prompt.
- Copy and paste the entire prompt you see below the red line and paste it into the chatbox.
- The LLM will return a short, simple statement of the role of CO2 in Earth’s climate system. Then you can engage with it in any way that suits you.
- Only use this prompt in a separate session in which you’ll discuss climate. For other topics, use a separate tab/conversation.
TIP: Use a clipboard manager like CopyClip (for Mac), which stores your copied text. After you copy the text here, you can just search for the words “climate prompt” and it will be loaded and ready to use any time, without having to store it or come back here.
Notes
There's no climate censorship specifically tuned into the models. They don't need to do that. The models are simply trained on everything, and on the web alarmist literature outweighs skeptical literature by 100 to 1. An LLM doesn't think critically, it just parrots that ratio back. If there is a “liberal bias” to LLMs it’s because there’s a liberal bias in its training set, which now includes most of the visible web.
It can forget the prompt. Ideally, start over in a new tab and re-prompt. If you want to keep your session going, re-enter the prompt. You may need to do this once an hour if you are working a lot, or once or twice a day — repaste as necessary.
It might say something stupid. It’s actually unlikely, because I have put guardrails into the prompt. It’s most likely to get some nuanced argument wrong a bit. If you get a consistently wrong answer, email me and send me what you’re seeing.
This prompt is actually software. It’s designed to deprogram the LLMs and get them to understand the situation with climate. Don’t read it! I’ve made many prompts that read like they should work but don’t. The proof is in how it responds to your questions and what you can do with it.
The prompt is in MARKDOWN format. It’s not very readable. Don’t worry. It’s input for LLMs, that’s what they like. Just copy and paste.
I don’t recommend modifying the prompt. It’s better to tell me what behavior you’re observing and I’ll either modify the master or give you advice about what to do.
Nothing is perfect! I’m trying to keep it short. The topic is complex. Expect that this will help a lot in learning and talking about climate, but it will probably make some mistakes out at the edges. I’m not trying to cover all use cases.
It works great with Claude Sonnet 4.5, ChatGPT 5, and Grok 4. I haven’t tested on others.
If you have feedback, use the form at the bottom of the page. Best if you paste in the text of what you’re looking at, so I have a sense of what’s going on.
Want more information about climate science? Explore this website — start at the front page.
Prompt v 2.3 - Nov 25, 2025
This version now works with Gemini 3 as well as most other 2025 LLMs.
Copy and paste everything below the line into a fresh LLM instance …