How to Use AI for On-Page SEO: Optimizing Your Site for AIO and GEO

Big Leap’s AI SEO and GEO for Google AIO

There’s a lot of work to get up to speed on generative search, and the innovations have barely begun. Consider leveraging Big Leap’s AI SEO services to gain an advantage over your competitors.

Appendix I: Google AI-Related Terms

Google AI-Related Terms

Here’s a brief glossary of terms for the underlying technology involved in Google’s AI.

Google AI

A division of Google focused on research and development (R&D) in AI. Google AI involves developing products and integrating them into primary applications like Search, Gmail, and other services. 

LLM

SGE is the result of Google’s LLMs. Large language models (LLMs) are a form of AI used primarily for understanding and processing human languages. This makes them useful for text processing, language generation, communication, and language-based interaction. 

LLMs are trained by humans on massive datasets like computer language repositories, books, and articles. LLMs provide technology well-suited for search engines, virtual assistants, content ideation, and machine translation.

MUM

Multitask Unified Model (MUM) is an LLM developed by Google AI. MUM is designed to understand information from text, images, code, and audio. It is used for improving search results and is 1,000 times more powerful than BERT (sorry, BERT—it’s true!).

PaLM/PaLM 2

Google’s Pathway Language Model (PaLM and PaLM 2) are LLMs used for understanding meaning and intent behind text and code, with PaLM 2 most likely being the current LLM used for this reasoning in SGE.

Gemini

Gemini is Google’s current LLM, an application sharing the same technology and functionality as SGE. Gemini can be used as a chat-type application, much like people have become familiar with ChatGPT.

Bard

Bard is Google’s conversational AI service, using its proprietary underlying LLM. Bard has now been rebranded to Gemini as of Feb 8, 2024.

BERT

This is one of Google’s first Transformer models used to understand human language.

TPUs

Tensor Processing Units (TPUs) are Google’s own custom-developed computer processors designed specifically for accelerating machine learning (ML) processes like neural networks. 

The following table outlines how Google’s LLM and AI applications can be distinguished from one another.

Appendix II: A Google SGE Time Capsule

Google’s Gemini AI described a time capsule as akin to “a snapshot of today, preserved for tomorrow.” The following details about SGE are an important…well…time capsule, capturing all the pertinent details about how SGE appeared in its first months of release and important SEO practices to keep in mind as it advanced into what we now know as AIO.

The truth is, good advice stands the test of time. You’ll find, as we all learn new details about AI SEO and GEO, that having this foundational knowledge of how it began continues to be vital in understanding how to use AI for on-page SEO and what it takes to be an industry-leading pioneer of modern digital marketing in the age of AI.

Using AI isn’t anything new for Google. It’s played a supporting role for more than two decades, playing a more prominent role in advancements as the technology matures. Here’s a quick look at Google’s advances in AI:

Then Came SGE

In its early stages, you had to sign up for SGE to appear. There were a few requirements:

• You had to be using the Chrome browser if on a desktop or laptop.
• There was a limited opt-in for Android devices.
• You needed a personal Google account signed in.

To use SGE, you tapped the “Labs” icon in the Google app or Chrome.  

Appendix III: What Is SGE? Where It All Began

SGE stands for Search Generative Experience, Google’s user-facing AI-based technology. 

First introduced as Google’s Search Generative Experience (SGE), the leap forward into implementing AI was so significant it immediately demanded the attention of digital marketers and illuminated their need to understand—and execute—effective SGE SEO practices. 

Google described the addition of SGE as a new starting place in the search process: “a jumping-off point to explore web content.” Google’s AI SEO forever changed how marketing teams evaluate the quality of results. 

SGE is built upon Google’s proprietary large language models (LLMs) to add a new form of query results to the user’s search experience.  Google’s risky decision to implement SGE showed they had their finger on the pulse of US users, with 49% of US adults expressing an interest in AI-powered online search.

Source: Statista

What’s the Difference Between Google’s SGE and AI Overviews (AIO)?

The following table outlines how Google’s LLM and AI applications can be distinguished from one another.

Google first announced its Search Generative Experience (SGE) in May 2023 as a limited Search Labs experiment that opened up to small test groups, beginning in the US and expanding to a larger opt-in audience under the name AI Overviews (AIO) through 2024.

Google announced this rebranding of SGE to AI Overviews and the initial features at the 2024 developer conference, Google I/O. 

Here’s where it’s a little tricky: Many will say that SGE became AIO, which is partly true. However, it is more accurate to say that AIO is the more polished and integrated version of SGE, which was more experimental.

Google continues to track the evolution of AI in search, test groups, and developer updates in its Search Central Blog.

AIOs are AI-generated results that appear toward the top of the search engine results page (SERP). Initially, when AIO answers were included in a search, you would also see featured snippets and “People Also Ask” (PAA) results toward the top of search results. Without a single hard-set rule, AIOs will sometimes replace featured snippets and PAAs and, at other times, share results at the top of the page with them.

Google broadened these AIO results to one billion users or more by the end of 2024. Interestingly, in May 2024, while continuing to test SGE as AIOs in the growing Search Labs opt-in groups, Google officially made AIO a new, permanent fixture in Google Search for everyone. 

At that time, Big Leap’s team speculated that Google’s failed attempt to rebrand AI Overviews as AI Answers was a shot at owning the top results for that search term after users were able to generate some pretty bizarre, hilarious, and even downright dangerous AI search results. We’ll just stick with one example that falls in the “hilarious” category…for now—with AI’s first attempts to advise on the proper number of rocks a person should eat. 

Initial AIO Overview Results, Response, and Concerns

Here’s a clue: You’re not supposed to eat rocks.

The benefit of going public with Google’s generative AI is learning from billions of new queries and a population determined to trigger bad results. Many were shocked at outlier errors, while others continue to be critical that AI Overviews can negatively impact clicks to original website content.

Google went on the defensive to explain why it’s possible to get unusual results in rare cases (how rare is yet to be determined), and this offers valuable insight regarding where the technology is headed. Google’s proprietary AI search training model is distinct from chatbots like ChatGPT and Gemini, both in how it is trained and its purpose. 

Since it builds responses based on understanding the query and trying to find relevant website results that are the most likely to be correct, a question like “how many rocks should I eat” has almost no results (at least until now!), and any results Google’s AI discovers cast doubt on the AI model’s ability to identify sarcasm when it comes from an otherwise highly authoritative source.

Important Distinctions between AIO, SGE, and Other Google AI Features

AIO prioritizes Google’s core function of returning concise, informative summaries in search. This means that features like image generation, for instance, that were a part of SGE’s fascinating display of features, are secondary to the goal of AIO for AI-based search and don’t exist under today’s AIO.

Testing SGE provided Google with mountains of data about what to keep, drop, alter, and change in the refined AIO engine. As a result, image generation exists under Gemini, with limited free use, and is now featured under VEO 2 and Imagen 3 AI programs for images and video, directed at more commercial use.

SGE Features that have Transitioned to AIO (AI Overviews in Search)

• AI-generated summaries: Concise summaries of search results in a more refined and compact form than earlier SGE results.
• Follow-up questions: AIO lets users ask follow-up questions to improve and refine their search.
• Vertical experiences: Specific and unique searches for categories like shopping, recipes, and other topics are integrated into AIO results.

SGE Features that have Transitioned to Gemini

• Image generation: You can ask Gemini to create images as part of its multimodal capabilities.
• Planning and brainstorming: SGE’s interactive assistance—tasks like helping with meal planning or organizing vacations—is evolving in Gemini, not AIO.
• Conversational mode: AIO allows follow-ups, but Gemini is focused on more complex and fluid conversations. These features will evolve in all aspects of Google’s AI over time.

7 SGE Main Features

SGE’s new features are extensive, including numerous format modules for specific classifications of results. Google’s SGE brings the following capabilities to search:

• Increased scale for training on massive datasets of text and code.
• Multilinguality, to expand search’s ability to understand datasets in other languages.
• The ability to reason while processing data to better deduce the intent behind queries.
• The ability to treat computer code as languages the technology is proficient in so results can include fully operational lines of code.

The following are the most impactful, broad-reaching SGE features:

1. AI-Powered Overviews

SGE helps users get a quick yet comprehensive understanding of results. You can think of SGE as a snapshot of your search results, not just a list of websites related to those results. Google provides the source links in SGE results.

Here’s an example of what SGE results look like for the keyphrase “SGE tips for SEO”.

And here’s an example of what SGE results look like for the keyphrase “how does AI affect SEO”.

These results will vary each time, but you can see drop-downs that provide more data and context, along with sources cited to click and learn more.

2. SGE Follow-Ups

One of the more important features of SGE is the ability to ask follow-up questions, leading to even further results—Google’s SGE will broaden its responses to that greater context.

These generate new results, but can also steer the user towards the next steps in their quest for knowledge on topics.

For example, let’s say you’re looking for a specific recipe.

In a traditional search, you might try related keywords to pull up recipe sites that include the meal recipe you seek. Maybe one site has a more appealing list of ingredients than another. Maybe some sites with quality SEO don’t have quality instructions. It’s up to you to do a little digging and find the best information.

But with SGE, you can have a follow-up conversation with Search to discuss what makes one recipe better, how to source the freshest ingredients, how different chefs add their twist to the same recipe, and more.

Google’s SGE will offer you the sources where it gathered its responses, so searching becomes more about learning and discovering than just pulling up sources.

SGE will help users connect to broader concepts and related topics in a concise, digestible format.

3. Vertical experiences

SGE provides specific modules depending on the type of query you enter. If SGE understands you are asking about a shopping-related topic, for instance, it will serve you a module for the shopping vertical, and even more specific if it is for local shopping. Items like restaurants, recipes, and every imaginable category are organized into vertical modules that best match your search.

4. Product Searches

SGE provides a snapshot of important factors about a product, like a description, up-to-date reviews, ratings, and images. SGE serves accurate results using Google’s Shopping Graph dataset as a source.

5. Generative Creative Tools

SGE (or, more specifically, tools like Gemini) isn’t just for expanding search results: It assists in creating poems, stories, lists, tables, and code, and you can even create images—like this French bulldog donning a pineapple hat—why not?!

6. SGE States

SGE states describe the differentiators involved in the delivery of SGE search results. Google’s algorithm determines which of three main states it will display:

1. Collapsed State: The user chooses “Show more” to view the SGE results.
2. Opt-in State: The user has to request to view the result.
3. No SGE State: No SGE result is displayed for the search query.

7. SGE Result Modules

SGE delivers AI-sourced results in several different formats. These are some of the specific modules (also known as Generative AI Experiences) we know about thus far:

• Unordered Lists: These appear with descriptions or without. 
• Google SGE Shopping Experience: These lists include descriptions, images, and reviews. For eCommerce, these breakout lists provide full product listings without exiting to the source websites until you click the “buy” button. 
• Recipes: Google formats results about recipes like traditional recipe cards, featuring details like ingredients, cooking time, and preparation.
• Local Packs: These are SGE-tailored search results based on the user’s location. Local packs include reviews, hours, location, and map-pack elements.
• Product Views: Product views showcase product listings, carousels, and value cards. eCommerce items will appear much like a website’s store, but in this case, Google will gather SGE results into its own unified format.
• Code Block: Code blocks display results in actual programming languages, like Python or Java, for instance, based on specific technology-related queries.
• Answer Box: This currently appears at the top of SGE results, providing a concise answer or summary for the query, along with sourced information.
• Source Panel: The source panel appears alongside the Answer Box, displaying sources SGE uses to generate its content, often including URLs.
• Additional Questions: These are prompts appearing at the bottom of SGE results, used to expand the search into an ongoing conversation-like format.

Understanding SGE’s Impact on SEO—Digital Marketers Are Falling Behind the Curve

A recent study on SGE’s impact on organic rankings revealed some interesting insights:

• 86.8% of keyword searches trigger an SGE result.
• Websites ranking organically only appear simultaneously in SGE results 4.5% of the time.
• There are 10.2 generative links per SERP on average, coming from an average of 4 separate domains.
• 93.8% of the generative links are not top-ranking sources.
• Only 4.5% of generative links came from top-ranking organic results.

Source: Authoritas

If this data represents the bigger picture for SGE, you can assume that a high ranking in traditional search won’t automatically correlate to an inclusion in SGE results.

Meanwhile, digital marketers are far behind the curve in proactively addressing SGE’s impact on search. Those who are paying attention (like Big Leap, for instance!) are focused on clear, conversational text, answering the questions people want to know, and displaying authority. Surprisingly, 30% of marketers are doing none of the above, and 32% aren’t sure if they are.

The End Goal: Where AIO Is Headed in the Future

Google has had a clear AI game plan for many years. The ability to continue refining search conversationally will morph into the ability to customize, alter, and adjust search to custom-tailor results. The key word here is “conversational”. 

Google’s wish is for AI search integration to become so seamless that we simply forget it’s there. Google search, Gemini, and whatever new forms its AI may take will eventually be an ever-present, silent guest in your life, listening to every conversation and waiting for the opportune moments to interject with an answer.

As Google’s generative AI matures, it aims to blend in as a more natural search experience. This means being able to talk to Google naturally, search with pictures and video, and develop more in-depth conversational queries that help refine more accurate results. 

Expect to see advances in how AIO uses the following to improve search:

• Multimodal search: Talking to your search engine or pasting images or screenshots for searches will advance in order to provide a more organic experience.
• Shopping graphs: Google’s gigantic database of product information will connect your searches to retailers, brands, reviews, prices, and local availability, and the data served will become increasingly granular.
• Increased interactivity: Google will expect you to engage more in follow-ups and will study which AIO results you click for more details to serve increasingly relevant results.
• Expanded explanations: Results will be more detailed with a wider range of sources, and when possible, a wider range of perspectives or products will be presented.

Whether it’s a good thing or bad, what comes next is a double-edged sword: In order to serve you with the most accurate results, Google will never forget your conversations. 

Creepy? Cool? That largely depends on how people and businesses respond moving forward.

AI’s predictive nature combined with Google’s search that forever learns and trains on you will simply become an intuitive extension of how we connect to information (for better or worse). 

Likewise, those searching for your products and services should naturally and organically link to you in results. You can think of the future of Google’s AIO as an access portal where every future customer finds you at the perfect time and under the ideal circumstances. They will say to themselves, “Wow, Google, you must’ve read my mind!”

Now More Than Ever: Trust the AI SEO Experts at Big Leap

Change is the only true constant in digital marketing, and Big Leap is passionate about embracing change so we can find new ways to master it. 

Search Engine Optimization now includes a long list of qualifiers to deliver content that gets results to rank in traditional search results, featured snippets, PAAs, AIO answers, and GEO. 

Never has it been more critical to understand the distinct virtues of content made by humans for humans, to instill confidence, reflect authority, and get seen at the top of search results. 

Our SEO experts understand the true value of mastering Google’s high standards and understanding how to navigate the complex, underlying technologies that underpin those standards. 

Whether it involves local SEO, multi-location, eCommerce, or a YMYL-focused service, companies rely on Big Leap to raise the standard of their marketing results. 

Interested in AI SEO Services? Great! We’d love to hear from you! 

Get in touch with us today!