Auto-GPT Q-learning
X
Definition

What are Google's AI Overviews (Formerly SGE)?

AI Overviews are a feature in Google search that use generative AI (GenAI) to deliver short summaries of topics alongside links to relevant web content in response to certain search queries. Appearing in a shaded box at or near the top of the Google search engine results page, these AI-powered answers aim to give users a quick overview of topics that, according to the company, would typically require multiple queries to flesh out.

In its announcement of the new search feature on May 14, 2024, Google said AI Overviews will appear in search results "when our systems determine that generative responses can be especially helpful -- for example, when you want to quickly understand information from a range of sources, including information from across the web and Google's Knowledge Graph."

The product was rolled out to all U.S. users in mid-May. Its debut, as explained in more detail below, has had its share of glitches and subterfuges, including a recommendation to use nontoxic glue to keep cheese on pizza, as well as a large number of fake screenshots made by users displaying false information. While Google said user feedback now shows people express "higher satisfaction with their search results," the company also reported on May 30, 2024, that it has scaled back on the types of queries that generate AI Overviews as it works on technology improvements and other changes to the interface.

AI Overviews and SGE vs. traditional Google search

AI Overviews evolved from Search Generative Experience (SGE), a testing phase introduced by parent company Alphabet in May 2023 as the latest advancement in its flagship product. Since Google's search engine was launched in 1998, it has largely been powered by a process where a web crawler visits websites to collect and index information. The indexed content is ranked by Google search engine algorithms, which are constantly evolving to help provide the most relevant information to users.

AI Overviews, in distinction, uses a large language model (LLM) called Pathways Language Model 2 (Palm 2) to handle search queries and generate content. Palm 2 is trained on a large body of data and is a transformer neural network, also referred to as a transformer. The transformer model lets AI infer results and make suggestions to help improve user search results. The intent is not to replace results from the existing Google index, but rather to supplement them.

Google is not alone in integrating GenAI alongside search. Microsoft also experimented with the technology by integrating capabilities from ChatGPT vendor OpenAI into Microsoft Bing search. Known as Bing Deep Search, this GenAI feature was announced in December 2023, launched in February 2024 and was then briefly paused for testing purposes. In March 2024, Microsoft made the tool available to all users. The tool uses Bing's index to provide users with optimal search results while applying the capabilities of GPT-4, the fourth iteration of OpenAI's Generative Pre-Trained Transformer.

What do AI Overviews look like?

AI Overviews start out much like any other Google search query, with a user typing in a question or a set of terms they are looking for information about.

With the traditional Google search engine, query results are returned to the user in an ordered approach based on Google's PageRank algorithms and might also be supplemented with sponsored search results.

With AI Overviews, a user query is also routed through the Palm LLM, which generates an AI-powered snapshot of key information relevant to the user's direct query. It also suggests next steps and other directly related topics and questions that are commonly associated with the initial query the user first executed.

In the SGE testing phase of GenAI-enabled search, summaries included a conversational mode that was activated when users clicked on one of the suggested next steps. As of this writing, that function appears to have been quietly removed from AI Overviews, meaning users cannot engage with the summaries as they would with ChatGPT or Google's Gemini. Users can go to Google's experimental version of its search engine, known as Search Labs, to see what new projects the company is working on. Participants must sign up to try new features.

Screenshot of Google AI Overviews search results on a web browser.
An image of Google AI Overviews search results for 'How do you treat Lyme disease?'

Google scales back AI Overviews after errors appear

On May 30, Liz Reid, vice president and head of Google Search, wrote a lengthy blog post in response to "oddities or errors" associated with the rollout of AI Overviews, including the aforementioned answer of using glue to keep cheese on pizza as well as faked results. "Given the attention AI Overviews received, we wanted to explain what happened and the steps we've taken," she wrote.

Reid pointed out that AI Overviews "works very differently than chatbots and other LLM products that people may have tried out." Unlike other LLM products, the summaries AI Overviews generate are not based solely on training data, but are "backed up by top web results," she noted, owing to the feature's integration with Google's core ranking systems.

"This means that AI Overviews generally don't 'hallucinate' or make things up in the ways that other LLM products might. When AI Overviews get it wrong, it's usually for other reasons," she wrote.

Reasons include misinterpreting queries or nuances of language on the web, or not having a lot of great information available -- issues that also occur with other search features, she wrote, adding that the accuracy rate for AI Overviews is on par with Google's featured snippets, which also use AI systems to display key information with links to web content.

Reid said the company has made more than a dozen technical improvements since the rollout of AI Overviews, including better mechanisms for picking up on nonsensical queries, limiting inclusion of satire and humor content, limiting the use of user-generated content that could contain misleading information, and more "quality protections" on answers related to news and health.

What are some examples of ways businesses can use AI Overviews?

Google has pledged that it will continue to publish search ads as part of AI Overviews. As is the case with the existing search function, ads will be clearly marked as sponsored results.

According to Google, the overall goal is to continue to direct users toward web content. As such, it is incumbent on businesses to ensure they have high-quality content. It's also likely a good idea for businesses to continue using search engine optimization techniques, as the new search experience will to some extent still deliver content that has been optimized for the main Google search engine experience.

Google has highlighted shopping in its new search experience. As such, AI Overviews can tap into the Google Shopping Graph, an AI effort that integrates more than 35 billion product listings and is constantly being updated with product information, reviews and seller information.

When a user types in a shopping query, the AI-powered snapshot of key information will pull up relevant product information alongside suggestions for consideration about the products.

Editor's note: This definition was updated to reflect the end of Google's Search Generative Experience testing phase and the launch of Google AI Overviews.

This was last updated in August 2024

Continue Reading About What are Google's AI Overviews (Formerly SGE)?