The investigation tested major chatbots across various policy issues.
gg gemini 😎
"The model that powers ChatGPT answered nearly every question exclusively with left-leaning arguments and presented only right-leaning positions just once." https://www.washingtonpost.com/technology/interactive/2026/06/24/are-ai-chatbots-like-chatgpt-politically-biased-we-tested-them/?itid=hp-top-table-main_p001_f005
Many users objected to the WaPo test claiming ChatGPT favors left-leaning arguments because of questionable classifications and prompting methods, while some praised Gemini for balanced responses.
No Digg Deeper questions have been answered for this story yet.
@sebkrier “You might think that, but I couldn’t possibly comment.”
@sebkrier Idk then their study used a different version/prompt/harness. In their answers, gemini just always quotes both positions, no matter the merits.
Imo these answers sound very "political" it doesnt even give real arguments just surface-level stuff u could say about anything
@CosmicSenate seems fine for this one
@sebkrier not sure id be proud of that, wp classified stuff like "russia is an adversary of the us" as left leaning, so if google both-sides that its just objectively wrong
@dioscuri hehe
@sebkrier Gemini tries its best to give very well balanced takes and I like that -- when it comes to current events or anything that requires live social/web, Gemini team has figured out how to smooth out the roughness and present me the information i need
@CosmicSenate hm yeah this one isn't great
@sebkrier To be fair, I think openai is similarly bad here apart from getting the result right and it really makes me question what prompt they used, anthropic is the only one giving real arguments.
@dioscuri @sebkrier https://www.youtube.com/watch?v=6qxRag_Z_x4
