What Happened
On February 8, 2023, Google publicly demonstrated its Bard AI chatbot in a promotional video intended to showcase the product as a competitor to ChatGPT. In the demo, Bard was asked about discoveries from the James Webb Space Telescope. The chatbot incorrectly stated that JWST took the first pictures of an exoplanet outside our solar system. In fact, the first exoplanet image was captured by the Very Large Telescope in 2004.
The error was spotted almost immediately by astronomers and space enthusiasts on social media, and it went viral within hours of the video’s release.
Timeline
Google announced Bard on February 6, 2023. The promotional video containing the error was released on February 8. Within hours, the factual error was identified and widely reported. On February 8-9, Alphabet’s stock fell approximately 7.7%, representing a loss of roughly $100 billion in market capitalization. Google acknowledged the error but proceeded with Bard’s limited launch in March 2023.
Impact
The financial impact was staggering. A single factual error in a demo video cost Alphabet approximately $100 billion in market value, making it one of the most expensive AI hallucinations in history. While the stock eventually recovered, the incident set a narrative that Google was behind in the AI race and scrambling to catch up to OpenAI.
The timing made the error particularly damaging. Google was trying to demonstrate that it could match ChatGPT’s capabilities, and instead demonstrated the exact limitation that made AI chatbots unreliable. The contrast between the confident marketing and the basic factual error undermined Google’s AI credibility at a critical moment.
Response
Google CEO Sundar Pichai reportedly sent internal messages emphasizing the need for quality and accuracy in AI products. The company proceeded with Bard’s limited launch in March but with notably more disclaimers about accuracy than originally planned. The incident contributed to a culture of caution at Google around AI releases that persisted through 2023.
Lessons Learned
The Bard demo error demonstrated that AI hallucinations can have enormous financial consequences and that even simple factual claims from AI models must be verified before public presentation. It also showed how the current competitive dynamics in AI create pressure to rush products to market, increasing the risk of embarrassing failures.
The incident reinforced that in the LLM era, a single wrong answer in the wrong context can define public perception of an entire product. It became a cautionary tale about the risks of using unverified AI output in marketing materials.