Across various nations, a significant shift towards a sustainable and environmentally responsible economy is occurring. Nevertheless, this positive transition is marred by a rise in greenwashing, wherein organizations overstate their environmental commitments. To combat this issue and safeguard consumers, initiatives have been developed to authenticate green claims. Given the increasing prevalence of environmental and scientific assertions, there is a pressing need for automated methods capable of detecting and validating these claims on a large scale. Recent advancements in artificial intelligence, particularly large language models (LLMs), have revolutionized the processing of diverse textual data as these models are extensively trained on vast datasets spanning multiple domains. In this paper, we evaluate the ability of several open-source LLMs to detect claims within the environmental domain where we have used various prompt engineering techniques like zero-shot, few-shot and contextual learning. To enhance performance, we have also finetuned a lightweight Large Language Model, Mistral-7B, on domain-specific data for the claim detection task. This approach reduces hallucination while retaining the benefits of LLMs and can also be employed to discern the grounds of claims and the types of climate pollution they aim to address. Acknowledging the inherent challenge posed by LLMs’ tendency to generate false or misleading information, we introduce EnClaim—a transformer-based architecture augmented with stylistic features. EnClaim integrates various linguistic stylistic elements with language models to improve the accuracy of claim detection, providing a robust framework for evaluating environmental claims.