What is SGE / AI Overviews, how does it work and what makes it different to standard organic search

11th Jun 2024

What is SGE / AI Overviews?

Google’s Search Generative Experience (SGE) / AI Overviews is a new beta program that uses the power of generative AI to revolutionise the SERP (Search Engine Results Page) and how we as searchers will interact with it. Instead of a standard search, that gives you a page of links to sift through to find the information you need, SGE / AI Overviews aims to provide you with AI generated summaries directly on your results page. 

In terms of healthcare, imagine a search around ‘best treatments for acne’. Google will now provide concise, high-level summaries from credible sources, such as brief overviews of the top acne treatments, maybe excerpts from authoritative dermatological sources and embedded videos from expert dermatologists explaining different treatment options.

In essence, SGE / AI Overviews aims to provide more relevant, comprehensive, and contextually appropriate search results by leveraging the capabilities of AI and machine learning.

How does SGE / AI Overviews work?

Google’s SGE / AI Overviews harnesses the power of generative AI. Generative AI is a type of artificial intelligence adept at creating new content, to provide users with a more comprehensive search experience. Large language models (LLMs) run generative AI – for example, an LLM called GPT-4 powers Chat GPT – and are adept at processing and understanding massive amounts of text data. 

Through a process called training, these models learn to identify patterns, relationships, and the nuances of human language.  This training empowers them to not just regurgitate information but to analyse and synthesise it in order to create new content.

SGE/ AI Overviews is powered by multiple LLMs, including MUM, PaLM2, LaMDA and more. 

• Google’s Multitask Unified Model (MUM) is trained across 75 languages and has been integrated into Google Search to enhance the overall search experience. It’s able to figure out related topics within video content, even in cases where these topics aren’t explicitly mentioned. 
• Pathways Language Model 2 (PaLM 2) tackles over 100 languages. It excels in multilingualism, its ability to handle mathematical concepts and reasoning and how it can understand and even write code. 

SGE / AI Overviews is also employing Gemini to make it faster for users. Gemini was announced at the end of 2023 and is natively multimodal. Multimodal AI means it can understand and generate data across various modalities, including text, imagery and audio. Google noted they achieved a 40% reduction in latency for users in SGE / AI Overviews.

What makes SGE / AI Overviews different to standard organic search?

For decades, organic search has relied on a well-established paradigm. When you enter a query, Google employs a complex algorithm to scour the web, identify the most relevant web pages, and present them as a list of links. This approach serves us well, offering a gateway to a vast ocean of information. However, it requires users to navigate through multiple websites, sifting through content to extract the answers they seek. This process can be time-consuming, especially for those seeking quick insights or basic information.

SGE / AI Overviews emerges as a transformative alternative. It aims to empower users with a more direct and comprehensive search experience. Utilising the prowess of generative AI, SGE / AI Overviews doesn’t just provide links; it strives to offer a concise summary of the topic directly on the search results page. 

Can SGE / AI Overviews be Wrong or Biased?

SGE / AI Overviews is susceptible to inaccuracies and biases, primarily due to the inherent complexities of natural language understanding and the vast amount of data it processes. Google has even stated itself that SGE/ AI Overviews may have gaps in knowledge for certain areas, and has only been designed to only sometimes produce a result if it needs more knowledge to properly answer the question. 

Additionally, AI hallucinations are worth noting. AI hallucinations refer to when generative AI systems (particularly LLMs) produce information, images, or text that may not be accurate or real. The AI algorithm makes mistakes or misinterprets data, which leads to content that’s not actually based on factual information. 

To counter biased results, Google employs various strategies, including adversarial testing within SGE / AI Overviews. This method involves deliberately attempting to challenge the system by feeding it data that is likely to provoke problematic outputs. By doing so, Google aims to pinpoint instances of bias and safety issues within the model, using these findings to enhance its performance and reliability.

Google’s SGE comes at an exciting time in the evolution of web search. Varn Health is committed to staying ahead of the SEO curve, constantly monitoring the updates of SGE to make sure that clients will show up in AI-generated results. We recently took a closer look and researched the overlap between SGE and organic search results, and will continue to investigate as new updates roll out. 

If you’d like to learn more about SGE and how it works as well as how Varn Health can help you on your SEO journey. don’t hesitate to get in touch.