The Rise of “GEO” – or How AI is Transforming Search Engine Optimization

I’m talking about this thing:

This is what’s known as generative search and it’s changing how we optimize content for search engines. Or at least it is by those in the know – which will soon include you. Instead of traditional SEO, savvy content marketers are now employing GEO – generative engine optimization – to boost their visibility.

If you don’t want to be left behind, then keep reading. I’ll explain how traditional SEO differs from GEO and where the concept of GEO originated from. Then I’ll go over some best practices so you can improve your odds of getting featured in generative search results.

How GEO differs from traditional SEO 🔄

Let’s begin by reviewing the differences between traditional SEO and GEO. This is important to understand because the two of them have fundamentally different objectives. And different objectives require different approaches.

With traditional SEO, the focus was primarily on three things – optimizing for keywords, ranking in search results for those keywords, and increasing the clickthrough rate (CTR) to your website by writing a compelling enough title with an intriguing enough meta description.

In contrast, generative engine optimization doesn’t focus on things like keyword density or similar measurables. Instead, GEO is concerned with citation inclusion, content synthesis, and providing authoritative content to user queries.

If you’ve done any kind of content publishing online then the traditional SEO approach is not news to you. And if you’re a WordPress user in particular, then you’ve most likely relied on plugins to do the brunt of your SEO work. Unfortunately, as convenient as it would be, there are no GEO plugin equivalents to Yoast or Rank Math. At least not right now anyway. Maybe some smart developers will create the first one in 2025.

But in the meantime, it means you need to do this work manually. And to do this work manually, you need to understand what the aforementioned citation inclusion, content synthesis, and providing authoritative content look like on a deeper level.

Understanding the new search landscape 🔍

The concept of generative engine optimization was first formally introduced in a groundbreaking research paper at the end of 2023. That paper was further refined in the following months, with the most recent version being revised in the summer of 2024. It represents a fundamental shift in how we think about search engine visibility.

As the researchers noted, “The advent of large language models (LLMs) has ushered in a new paradigm of search engines that use generative models to gather and summarize information to answer user queries.”

You might be wondering why this is such a big deal.

Here’s the thing: while the traditional search engine approach simply provided a list of relevant websites to a user query, these new generative engine models (like Google’s SGE, Perplexity, and others) are doing something completely different. They’re creating comprehensive, AI-generated responses by synthesizing information from multiple sources. This is the content synthesis element of the GEO equation:

According to the papers’ authors, this shift “significantly improves user utility” but poses “a huge challenge for the third stakeholder – website and content creators.”

Why?

Because when users get their answers directly in the search results, they’re less likely to click through to individual websites.

Although I wouldn’t disagree with this on the whole, speaking for myself, I actually do click on the websites that Google, Perplexity, and Grok show me in these generative search results. The reason I do it is because the sources themselves might not always be the most accurate or reliable – which means the answer given to me could be incorrect (or outdated). Checking the links gives me peace of mind.

So even though not everyone is going to do what I do, there are some that will – and these are the people you want to reach by showing up in generative search results. Here’s how to do it…

Successful GEO strategies based on experimental research 🎯

Remember those three key elements I talked about in the introduction – citation inclusion, content synthesis, and providing authoritative content? The research paper I mentioned earlier reveals specific strategies to target them. Let’s review the strategies and afterwards I’ll share some data with you to show you what the results of using them looks like.

Statistical integration

When generative engines combine information from multiple sources (i.e., content synthesis), they love concrete data points they can work with. Here’s how to incorporate them:

• Include quantitative data that tells a story (real numbers, not fluff).
• Present verifiable statistics from credible research.
• Build data-driven arguments that support your main points.

The main takeaway is that generative engines are particularly good at recognizing when statistics add meaningful context versus when they’re just decorative. Quality and relevance matter more than quantity.

Citation and quotation strategy

Using this strategy strengthens your content’s authority. In turn, this increases the odds that your article will be cited in a generative search result (i.e., citation inclusion). Here’s the winning formula:

• Reference authoritative sources that truly add value.
• Include direct quotes that enhance your arguments.
• Provide clear, proper attributions for all citations.

Content fluency optimization

Content fluency is particularly fascinating because it shows that generative engines can actually recognize good writing. The research found a visibility boost of 15-30% just from improving content fluency. But what does “fluency” really mean in practice? It means to:

• Write with clarity and purpose (e.g., avoid unnecessary jargon).
• Use natural language patterns.
• Create logical flow between ideas.

The key insight here is that you’re writing for an AI that’s been trained on millions of examples of human communication. It knows good writing when it sees it – just like we do. Simply put, if you write well then you improve your odds of being featured in a generative summary snippet.

Measuring the real impact of these GEO strategies

Okay, I promised to share some data with you, but first I need to acknowledge something important: certain parts of the researchers’ methodology explanation are about as clear as a foggy morning in London. This doesn’t mean I’m discounting the results – I wouldn’t be sharing them if I was. But let’s begin by breaking down what we know and what we can reasonably infer about their process.

Understanding the GEO-bench dataset

The researchers worked with something they called “GEO-bench.” You can think of it as their testing toolkit – a carefully curated set of queries and search results they could use to evaluate their optimization strategies.

In their own words, GEO-bench consists of:

“10K queries from multiple sources, repurposed for generative engines, along with synthetically generated queries. [It] includes queries from nine different sources, each further categorized based on their target domain, difficulty, query intent, and other dimensions.”

They split these 10,000 queries into three groups:

• 8,000 training queries
• 1,000 validation queries
• 1,000 test queries

For each query in the GEO-bench, they pulled the top five Google search results to use as source material.

How they might have tested it all

While the researchers weren’t explicit about their testing system, we can make some reasonable assumptions based on standard research practices and the nature of their work.

My best guess is that they likely created their own synthetic GE (generative engine) setup – essentially a controlled testing environment. This makes sense because testing with real-world generative engines like Google’s or Perplexity.ai would be impossible – you can’t know exactly what sources they’re pulling from, when or how they’re using them.

They probably used this controlled environment to feed their GE those top five Google search results as a knowledge base. Then, by modifying pieces of content in specific ways and observing how these changes affected the GE’s responses, they could measure the impact of different optimization strategies.

The 8,000 “training” queries likely helped them develop and refine these strategies – experimenting with different content modifications and observing patterns in how their GE responded. The 1,000 validation queries then served as a first checkpoint to verify these patterns weren’t just random flukes. Finally, they used the 1,000 test queries – completely new ones their system had never seen before – to prove their strategies worked consistently in fresh contexts.

This is, of course, just my interpretation of their process. I could be completely wrong (if you’ve got a different opinion, drop it in the comments – I’d love to hear it).

Regardless of the exact methods though, the findings of their research were fascinating…

Average improvement when combining GEO methods

Below is a matrix that shows how much better content performed in generative search results compared to unoptimized content when combined in pairs of two. For example, when content incorporated a combination of fluency and statistics, it improved by an average of 35.8%. This was the most effective pairing:

In this context, “improved,” is based on two key metrics that the researchers measured against baseline (non-optimized) content. These two metrics were position-adjusted word count and subjective impression.

Position-adjusted word count examines two crucial factors together: how much of your content appears in the generative response AND where it shows up. Content appearing earlier in the response gets weighted more heavily since that’s what users typically see first.

The subjective impression metric digs deeper into quality factors that affect visibility. These include:

• Relevance to the search query.
• Impact on the final response.
• Information uniqueness.
• Likelihood of user engagement.
• Integration with other sources.

Overall, despite the lack of clarity on certain aspects of their methodology, we can see that combining different GEO strategies can lead to some pretty impressive results.

What you can do right now to implement GEO 💡

At this point you’re probably ready to optimize your content for generative search.

I’ll explain how to break it down into manageable steps, but first I want to quickly mention that Google specifically tends to favor generative responses for certain niches, while avoiding it for others.

For example, I tested a number of queries dealing with food recipes and didn’t get a single generative response.

You might want to run a similar test for your own niche to see what Google serves you for results. Try different types of queries (e.g., what is X, how to do Y, what’s the best solution for Z, etc) to really give it a thorough test.

Regardless of the results, it’s also good to keep in mind that Google is not the only generative search engine on the web and a lot of this GEO strategy is about making your content better for human readers as well. So just because you might not see a large volume of generative responses doesn’t immediately mean you need to abandon the idea of doing this.

However, depending on your time and resources, it could be a decision-making factor for some of you. Hence why I brought it up. Now on to the fun stuff:

Step 1: Map your optimization opportunities

First things first: you need to look at your existing content with fresh eyes. Think of it like decluttering your closet – you’re not throwing everything out, just figuring out what needs a refresh.

For example, if you run a travel blog and you’re writing a post about taking some specific journey, then include precise details. Those details will increase your odds of getting scooped up and featured in a generative response:

Imagine in the example above, if any of the website sources Google pulled from just said to rent a motorbike and drive but without any of the other details. Google would have skipped over them and found some other blogs or vlogs that did have those details and would have featured them instead.

Also note how the top citation has a first person point of view title. This signals to Google that the person providing the information has some authority on the subject, since they actually drove the Mae Hong Son loop and are speaking from direct experience. In other words, their article wasn’t generated by ChatGPT and possibly filled with AI hallucinationsAI hallucinations occur when an AI model generates incorrect, misleading, or nonsensical results in response to a prompt..

The key here is identifying opportunities to add substance to your content.

Where to start

A good place to begin would be with recently decaying contentDecaying content is what happens when the usefulness and impact of a piece of content gradually fade over time. It could be because the subject matter is no longer current, new competitors offer fresher information, or audience interests have shifted. That content then slips down the search rankings and attracts fewer visitors. on your site. You can use Google Analytics to run reports that will show you which pages are experiencing declining views and engagement. If you’re not sure how to do this, ask ChatGPT or Claude and either tool should point you in the right direction.

Once you make your list of five or ten posts that you think would make good candidates for some GEO experimentation, then it’s time for the next step.

Step 2: Implement your power combos

Here’s where the research really comes in handy. We know that combining certain GEO strategies packs an extra punch. For example, we saw in that matrix earlier that pairing fluency optimization with statistical integration showed an impressive 35.8% improvement in visibility.

So go through your list of posts, one by one, and as you are reading through them, think about what combinations would make sense for each post.

I already gave the travel blog example. So if you run a travel blog, you can analyze your content and look for similar opportunities. Identify where it’s possible to improve fluency optimization by switching to a first-person perspective and where you can integrate statistical information by adding any specific details where you currently have vague explanations. In other words, find your own 150cc motorbike opportunities.

Regardless of your specific niche, you can apply a similar thought process:

• For tech review content, combine citation strategy with statistical integration to create a potent mix.
• For fitness content creators, provide your personal experience and back it up with quotes from exercise scientists or studies from reputable journals.

…and so on.

These are just a few examples to get your brain moving. The idea is to think about how different combinations can work together to enhance your specific type of content.

💡 Pro tip: After you’ve made your optimizations, read your content out loud. Does it maintain a natural flow with your newly added elements? If the answer is “yes,” then that’s your sweet spot.

Step 3: Test, monitor, and refine

Once you’ve implemented your chosen strategies, you need to track what’s working. Are certain combinations performing better than others? Are some pieces of content suddenly getting more traction in generative search results?

What about traffic in Google Analytics and Google Search Console? Is it going up or down?

Keep a close eye on your content’s performance and be ready to adjust your approach. The generative search landscape is still evolving, and what works today might need tweaking tomorrow.

Looking to the future 🔮

Generative engine optimization represents the next evolution of SEO, requiring a fresh approach to how we create and structure online content. Since we’re still in the early days of this thing, starting now can give you a competitive edge.

Treat it as an opportunity: by curating authoritative sources, offering meaningful statistics, and writing with clarity, you can significantly improve your odds of being found and featured in generative search results. Then people like me – who actually click on the source links – will find your website and hopefully stick around for a bit. That’s the goal. I hope what you read here helps you reach that goal.

Is this your first time reading about GEO or have you already started experimenting with it? Let me know in the comments if it’s working for you or if you have any questions about it.