Google is testing hiding the AI origins of a new Search feature (Updated)

Update, September 13, 2024 (07:24 PM ET): In the original article below, we surmised that Google’s new test of unlabeled AI search results was a part of AI Overview due to its similarity in both function and placement on the search results page. However, we have updated the article to reflect that this is not a part of the AI Overview system. Instead, this is a new feature in testing that is unrelated to AI Overview. However, the core of the issue — that this new feature uses generative AI but does not clearly label the created content as AI-generated — remains unchanged.

Original article, September 13, 2024 (01:06 PM ET): Google seems to be making it more difficult for users to know that the summaries it provides at the top of search results are generated by AI. It’s testing a new system that appears to be very similar to the current AI Overview feature, but it lacks that system’s prominent labeling that the content depicted is AI-generated.

In its current form, AI-generated summaries in Search feature a prominent AI Overview header with the Gemini logo. The feature also has a “Generative AI is experimental” tag at the bottom that appears when you tap the large “Show more” button. However, as per Barry Schwartz on X, Google is testing a new system that could possibly replace or augment AI Overview. It has no header, removes the large “Show more” button, and brings in a smaller “More” button in its place.

You can see screenshots below of how AI Overview currently looks (left) and how this new system compares to it (right):

Although the “Generative AI is experimental” tag is still present in the new feature, it remains hidden until you tap the small “More” button. Since the Gemini logo is missing and this feature is not named something as forthright as “AI Overview,” there’s no obvious indicator for users to know that the summaries are AI-generated without directly interacting with the content.

On multiple occasions in the past, AI Overview created inaccurate information that could prove harmful to the user. However, at least with AI Overview, it was pretty obvious that the content was AI-generated. With this new system’s lack of prominent labeling, it’s possible — even likely — that it could prove detrimental to the user experience.

Fortunately, this is just a test and isn’t widely available at the moment. We hope Google realizes it’s a bad idea and doesn’t give it the green light, or at least makes it more obvious that the presented content was created by AI.