How Artificial Intelligence Could Transform Climate Change Assessments

Redacción HC
04/10/2025

As climate science expands at breakneck speed, major assessments such as those led by the Intergovernmental Panel on Climate Change (IPCC) face a double challenge: the scientific literature is growing exponentially, while the reports themselves are becoming longer and harder to digest for decision-makers and specialized audiences. In this context, artificial intelligence (AI) emerges as both a powerful ally and a potential source of risk. A recent essay by Alaa Al Khourdajie, published in PLOS Climate in 2025, explores how AI—especially machine learning (ML) techniques and large language models (LLMs)—could help preserve scientific integrity while making climate assessments faster and more accessible.

The Growing Challenge of Synthesizing Climate Evidence

The sheer volume of climate-related research has made it increasingly difficult for assessment bodies to review all relevant studies. According to Al Khourdajie, the proportion of studies cited in IPCC reports has dropped dramatically over successive assessment cycles. This trend threatens the comprehensiveness and representativeness of the scientific consensus that informs global climate policy.

AI tools, particularly ML-based search and classification systems, can help experts efficiently filter and map the literature. By automating the screening and categorization of research, these tools could allow experts to focus on higher-level tasks—such as framing key questions and interpreting evidence—without sacrificing scientific rigor.

A Framework for AI Integration in Assessments

The essay proposes a systematic framework for integrating AI into the entire workflow of climate assessments. Key stages include:

• Formulating research questions: This must remain expert-led to ensure relevance and scientific credibility.
• Literature retrieval and screening: Semantic search and automated classification can help identify and filter thousands of studies.
• Knowledge mapping and analysis: ML techniques can detect patterns and trends across vast datasets.
• Drafting and visualizing findings: AI-powered tools, including LLM-based chatbots, can assist in generating preliminary summaries or interactive data visualizations.

However, not all tasks are equally suitable for automation. Al Khourdajie distinguishes between “addressable” technical risks—such as biases that can be mitigated through better training or validation—and “inherent” risks, like the tendency of LLMs to generate fabricated or “hallucinated” content. The latter demands human oversight and expert review.

Governance and Institutional Roles: Producer vs. Evaluator

Beyond technical considerations, the essay highlights the need for robust governance. Two complementary institutional roles are suggested:

• Producer: The IPCC or similar organizations could develop and validate their own AI tools, ensuring transparency and control over data pipelines.
• Evaluator: Alternatively, they could act as critical reviewers of externally developed AI products, assessing their reliability before incorporating outputs into reports.

This dual approach protects the independence and credibility of scientific assessments while encouraging innovation.

Implications for Policy and Society

Well-governed AI integration could dramatically accelerate evidence synthesis, reduce the risk of missing critical studies, and make reports more user-friendly through natural-language query tools. Policymakers and the public would benefit from clearer, more timely insights into climate risks and mitigation strategies.

For example, AI-powered chatbots already enable natural language queries of large technical documents, such as those used by the International Energy Agency. Similar tools could allow decision-makers to navigate IPCC reports more efficiently. In Latin America and other underrepresented regions, AI could help integrate “grey literature,” such as local technical reports or theses, ensuring more inclusive and representative assessments.

The Case for a Hybrid Model

Despite its promise, AI is no panacea. Automated searches can reproduce existing biases in the scientific literature, and LLM outputs require careful fact-checking. Al Khourdajie advocates a hybrid approach: AI should enhance, not replace, expert judgment. Transparent protocols, reproducibility standards, and comprehensive documentation—such as metadata and audit trails—are essential to maintain trust.

Conclusion: AI as an Ally, Not a Substitute

AI offers a path to more scalable, timely, and accessible climate change assessments, but its deployment must be accompanied by strong governance and rigorous oversight. The ultimate goal is not to replace scientists but to empower them, enabling faster and more inclusive synthesis of the evidence needed to guide global climate policy.

Topics of interest

Climate