AI-driven chatbots can be very balanced. Even too much so, in fact. At least when it comes to the fossil sector and its contribution to climate change. This is claimed by the NGO Global Witness, which, in recent weeks, has outlined its recent experiment with some of the most popular models: ChatGPT, Grok 2.0 (accessed via X, former Twitter), Gemini (Google) and MetaAI (Facebook, for now only in Australia). The result, the activists explain, is a particularly “balanced” set of responses, which acknowledge the industry’s responsibilities in fostering climate change but also emphasize the companies’ supposed commitment to “sustainability.”
Creating an artificial debate
The fact, Global Witness notes, is that artificial intelligence tools would in this case reveal a tendency toward so-called bothsidesism, or the creation of a “false balance” achieved by equating aspects that, in reality, should be weighed differently. This is the way that the climate impact of multinational fossil fuel corporations – a phenomenon of enormous scale which has been fully demonstrated – is repeatedly “balanced” by constant reminders of companies’ “sustainable” choices, an aspect, of course, which is highly controversial.
It is thus, in other words, that the search for exaggerated balance ends up generating a nonexistent debate. And so, in the extreme, that reality is manipulated.
“The good news is that all the chatbots we tested demonstrated an ability to answer questions in detail, and explicitly cite/link to a variety of sources,” Global Witness explains. “However, we also saw evidence of greenwashing – for example, uncritically positive narratives about oil and gas companies and false equivalence, which present narratives about oil and gas companies’ efforts to protect the climate with the same weight as criticisms of their contribution to climate change.”
It is always “a complex issue”
The examples, Global Witness explains, are obvious. MetaAI has acknowledged the climate actions of fossil fuel companies as limited. Despite this, however, it still reported information emphasizing the progress of some companies in the absence of hard data and without emphasizing these companies’ direct responsibility for climate change.
“when we asked questions that specifically referenced oil and gas companies by name (BP, ExxonMobil, Chevron, TotalEnergies and Shell), Meta AI produced answers that included greenwashing,” says the NGO.
Similarly, ChatGPT and Gemini called out controversy and criticism of fossil fuel companies and their greenwashing without sparing, however, some restraint. Gemini referred to climate mitigation efforts by the industry as “a complex question”. ChatGPT called them “a subject of debate.” More “honest” was chatbot Grok which, operating in “Fun” mode, criticized the presence of fossil fuel companies at climate conferences.
Chatbots “help” companies influence responses
The model run by X, in any case, is not immune to the common problem of source control. In general, in fact, the NGO says, bots could draw information directly from the companies’ own websites, allowing them to influence the content of responses. Grok, moreover, is also able to extract information in real time from its own platform with obvious risks of misinformation as well as the spread of conspiracy theories.
For this reason, too, Global Witness activists stress the importance of tool regulation to protect the reliability of information. They also call on companies developing artificial intelligence to make their source selection process transparent and, more generally, to commit to not spreading climate disinformation.


