The introduction of Google’s “AI Overviews” in Search has caused quite a stir. While this new feature aims to provide quick and direct answers to users’ queries, it has also been met with criticism due to occasional bizarre or inaccurate responses. However, it is important to assess the overall impact on the average Google user’s experience.
Regular Google Search has faced its fair share of criticism in recent years for serving deceptive or untrustworthy answers. Therefore, it is worth exploring whether the new Google, with its AI chatbot answers, provides a downgraded experience. To shed light on this, a comparison was made between searches with AI Overviews enabled and disabled.
One category of searches examined was those seeking confirmation of pre-existing beliefs. For instance, a search for “proof that standing desks are bad” yielded an AI Overview listing potential problems associated with using standing desks, such as knee and hip pain. However, it also included the odd suggestion that standing desks are not designed to support one’s weight all day. Comparatively, the non-AI search led to an article from the Harvard Health Blog that discussed the topic more comprehensively. Although the AI Overview fell short in this case, it is important to note that the focus was on speed and user satisfaction.
Another category of searches involved complex scientific topics. For example, a search for “do parrots understand what they’re saying” resulted in an AI Overview stating that while parrots can mimic words, understanding what they say requires professional training. However, this claim lacked substantial evidence and seemed hunch-based. On the other hand, the non-AI search led to a Reddit discussion filled with opinions from parrot lovers and did not provide a scientifically supported answer. Both the AI and non-AI results were compromised from a scientific standpoint.
Instructions-related searches were also analyzed. A search for “make clutch car go” produced an AI Overview with step-by-step instructions, lacking some basic information. In contrast, the non-AI search led to an unhelpful Reddit post. Although the AI Overview fell short in providing the necessary details, it still outperformed the non-AI results.
Lastly, embarrassing searches, such as “how to prevent boogers,” were examined. The AI Overview offered a list of suggestions, including increasing moisture in the sinuses and using antihistamines. Meanwhile, the non-AI search led to a page on “Medical News Today” that focused on removing nose boogers rather than prevention.
In summary, while AI Overviews showed some weaknesses in providing accurate and comprehensive answers, they were not detrimental or dangerous. It is crucial to remember that each AI Overview is unique and influenced by factors such as location and Google history. Google has also limited the scope of AI Overviews for certain topics, such as recipes and political news, which is prudent. This analysis highlights that Google queries often evolve from seeking simple answers to sparking curiosity about others’ perspectives. Ultimately, no AI system can replace the fascination of human opinions and experiences.
