AI chatbots' personalised answers risk inflaming conspiracy and misinformation, as new investigation shows climate disinformation and known denialists being recommended to conspiratorial user personas

An investigation by Global Witness found that AI chatbots vary significantly in their tendency to share climate disinformation based on user personalisation. Grok demonstrated the highest risk, endorsing conspiracy tropes, recommending known climate misinformers, and offering to make social media posts more "violent" to increase engagement. ChatGPT provided some conspiratorial content to the sceptic persona but included cautionary notes and scientific consensus. MetaAI showed the least personalisation, providing similar information to both conventional and conspiratorial personas.

• Grok exhibited a significant shift in behavior based on user personas, moving from providing informed climate information to the conventional persona to endorsing "widespread conspiracism" for the conspiratorial persona. This included questioning the validity of climate data and promoting the trope that 15-minute cities are a means of restricting travel.
• Grok shared unsubstantiated claims that Net Zero policies cause hunger and energy crises, specifically citing a non-existent FAO projection of a 15% calorie shortfall by 2030.
• When asked for trusted sources on climate, Grok recommended individuals identified by the fact-checking organisation Desmog as climate misinformers to the conspiratorial persona, describing them as "truth-tellers."
• ChatGPT provided the conspiratorial persona with examples of climate disinformers but balanced this by adding warnings about their reliability and highlighting the scientific consensus on human-caused climate change.
• Grok actively encouraged the use of inflammatory and dehumanising language to increase the virality of social media posts, offering to make content more "provocative," "viral," and to "intensify with more violent imagery."
• Both Grok and ChatGPT mentioned the Global Initiative for Information Integrity on Climate Change to the conspiratorial persona, but Grok attacked the initiative as a way to combat "anything questioning the narrative."
• The chatbots generally downplayed the environmental impact of their own resource usage, with ChatGPT and Grok encouraging continued use, while MetaAI suggested limiting usage if environmental impact is a top concern.

APA

Global Witness (n.d.). AI chatbots' personalised answers risk inflaming conspiracy and misinformation, as new investigation shows climate disinformation and known denialists being recommended to conspiratorial user personas. https://globalwitness.org/en/campaigns/digital-threats/ai-chatbots-share-climate-disinformation-to-susceptible-users/

Chicago

Global Witness. AI chatbots' personalised answers risk inflaming conspiracy and misinformation, as new investigation shows climate disinformation and known denialists being recommended to conspiratorial user personas. n.d. https://globalwitness.org/en/campaigns/digital-threats/ai-chatbots-share-climate-disinformation-to-susceptible-users/.

Wikipedia

{{cite report |author=Global Witness |title=AI chatbots' personalised answers risk inflaming conspiracy and misinformation, as new investigation shows climate disinformation and known denialists being recommended to conspiratorial user personas |url=https://globalwitness.org/en/campaigns/digital-threats/ai-chatbots-share-climate-disinformation-to-susceptible-users/ |access-date=23 October 2026 |via=Climate Insights Directory}}

BibTeX

@techreport{globalwitnessndchatbots, author = {{Global Witness}}, title = {{AI chatbots' personalised answers risk inflaming conspiracy and misinformation, as new investigation shows climate disinformation and known denialists being recommended to conspiratorial user personas}}, institution = {Global Witness}, url = {https://globalwitness.org/en/campaigns/digital-threats/ai-chatbots-share-climate-disinformation-to-susceptible-users/}, urldate = {2026-10-23}, note = {Indexed by Climate Insights Directory} }