Google has recently scaled back the implementation of AI-generated responses in search results, following the emergence of notable errors. This includes inaccuracies such as suggesting glue as an ingredient for pizza sauce.
Introduced merely two weeks ago, the AI Overviews feature, initially available to U.S. users, placed AI-generated summaries at the forefront of search results. However, a notable reduction in these overviews was observed by users, including SEO professionals, hinting at Google's decision to modify the feature amid critique. The AI functionality remains non-optional for search engine users.
Liz Reid, Google's Head of Search, acknowledged these issues in a recent blog post, confirming adjustments were being made.
The need for change was underscored by instances where AI Overviews malfunctioned spectacularly. Misinformation spread rapidly online, with erroneous claims about public figures and blatantly incorrect facts, such as a former U.S. president's religious affiliations, geographical inaccuracies about Africa, and bizarre dietary advice involving rocks.
To combat such issues, Google has introduced new measures. These include the identification and exclusion of "nonsensical queries" from AI-generated results, a reduction in the inclusion of satirical or humorous content, and constraints on prompts lacking sufficient data, where AI-generated results could mislead rather than inform.
Further, Google's safeguards extend to content sources, particularly limiting information from forums or social media, which, while sometimes offering valuable insights, can also propagate misleading advice. An example given was the inappropriate suggestion to use glue for culinary purposes, which Reid clarified in her post.
Google has established protocols to avoid displaying AI-generated results for sensitive topics like news and health, addressing potential risks such as harmful misinformation that could encourage hazardous behaviors.
These amendments follow a broader pattern observed with Big Tech's initial AI rollouts, which often require subsequent revisions to address unforeseen complications.
This trend continued earlier in the year when Google AI's capability to generate images faced backlash for bias in racial representation and historical inaccuracies, leading to a temporary suspension of the feature after public and critical feedback.
