Google Queries With Featured Snippets, Questions Likely Trigger AI Overviews
AI Overview, the search capability powered by Gemini, Google’s underlying foundational model, has begun to answer and predict follow-up questions before they are asked in a query.
BrightEdge Generative Parser released data on Tuesday shows Google’s ability to provide precise responses and anticipate follow-up questions through its generative artificial intelligence (GAI) technology.
“I have this vision, a metaphor in my mind, that through our AI parser we can peek inside Google Labs and see the changes as they work on them,” said Jim Yu, CEO at BrightEdge, adding that information from AI Overviews does not always serve up.
In fact, less than 15% of the keywords trigger AI Overviews, down from 84% when Google was testing it.
“This indicates Google is working on balancing information—the usefulness of when to trigger it vs. the risk of providing incorrect answers,” Yu said.
Google’s AI already predicts and answers follow-up queries before they are asked. Yu said since the I/O summit, the overlap between the citation in the AI on the exact match keyword is different than before.
“Now, if I search for a big screen TV, it determines the follow-up questions and serves them up before they are asked,” he said. “Questions such as where should I place my TV in the room and what things should I consider.”
Search queries with Featured Snippets and questions are two types most likely to trigger AI Overviews in general availability, compared with local-related queries and site links, which are less likely.
A snapshot through the parser of the data from May 30 shows AI Overviews will change and roll-out by industry.
Some 63% of the keywords in healthcare did show AI Overviews, but Google has said in the past it will tread lightly in this market segment. Yu expects to see less.
B2B tech has 32% of AI Overviews. Ecommerce has 23%.
Restaurants and travel serve up very little information from AI Overviews, Yu said. “Very different from when the technology was in labs, with location-based queries heavily triggering AI Overviews,” he said.
“They also are giving much less space to answers from AI Overviews in the search page,” he said, as the company fine-tunes search quality, flow of traffic, and moderation in terms of trying to determine how much space to give ads.
Yu believes Google is building different formats and experiences per sector, as well as deciding what keywords should trigger the AI technology.
Schema markup also will become important because the engines will need to pinpoint the original source of the information serving up.
“We are seeing the evolution in the data,” Yu said. “The days of focusing on one keyword is over. Now we must understand the entire conversation around the keyword including intent and series of actions. Then optimize around that.”
(18)
