DMS25 Part 2: New Era of Search Marketing: Understanding Generative Engine Optimization (GEO)

A collage of views of how it was at the lobby and entrance for the Digital Marketing Summit 2025 in Seoul

This is Part 2 of my DMS 2025 Seoul report. You can read Part 1 here.

One of the talks that I attended was by Sebastien Edgar who deep dived into how AI (or more appropriately LLMs) looks at data and how that affects our exposure to them.

When it comes to search marketing, we're witnessing a significant shift in how users discover information online. While Google still dominates with around 90% of the search market, AI chatbots like ChatGPT are rapidly gaining ground, particularly among younger users. This evolution requires us to adapt our optimization strategies beyond traditional SEO.

The Changing Search Landscape

The data clearly shows we're at an inflection point. Google remains the dominant player, but AI-powered search is growing exponentially. According to the presentation, Gen Z users (18-24) are increasingly turning to AI chat platforms instead of traditional search engines. In fact, 46.7% of this age group prefers using chatbots for searches - a trend that will likely increase as these platforms evolve.

One particularly concerning statistic for SEO professionals: when Google displays AI Overviews (AIO) in search results, click-through rates drop dramatically from 3.97% to just 0.64%. This represents a massive reduction in organic traffic potential when AI summaries appear.

Another interesting insight is that 52.2% of ChatGPT searches are informational, whereas Google searches tend to be more branded or navigational. This distinction is crucial for understanding how to optimize for different platforms.

I've also written in the past about how AI is impacting online search.

How LLMs Work: The Technical Foundation

To effectively optimize for AI systems, we need to understand how Large Language Models (LLMs) function. These models:

1. Are trained on common crawl data, fine-tuned datasets, and reinforcement learning
2. Process text through tokenization (converting words into numerical representations)
3. Don't actually "know" information but predict the next most likely word sequences
4. Associate brands with queries based on proximity and frequency in training data

This fundamentally changes our approach to optimization. Traditional keyword stuffing is ineffective for LLMs. Instead, they respond better to:

• Clear, high-quality content with statistics and quotes
• Content with proper readability scores (around Flesch-Kincaid 8)
• Structured formatting using HTML tables and proper citations
• Content from authoritative domains (particularly Wikipedia)

Optimization Strategies for LLMs

Based on research presented at the conference, several optimization techniques show promising results:

Content Optimization

The data is compelling. Adding certain elements to your content can dramatically increase visibility in AI responses:

These numbers should immediately shift your content strategy. By implementing these techniques, you can significantly improve your chances of being referenced in AI-generated responses.

One thing of interest though: When I saw these in Sebastien's slides, I immediately thought of SEO's best practices of giving objective value to readers. They are basically the same!

Technical Optimization

Several technical factors also influence how LLMs interpret and recommend your content:

1. Check your robots.txt - Ensure you're allowing LLM bots to crawl your content
2. Reduce JavaScript dependency - LLMs still struggle with JavaScript-heavy sites
3. Implement schema markup - Especially important for Microsoft-linked bots
4. Optimize internal linking - Helps showcase how pages are connected conceptually

Traffic Analysis

A fascinating insight from the presentation: 77.35% of LLM traffic goes to blog pages, compared to just 9.04% for home pages and 8.23% for news pages. Product pages and search pages account for less than 1% of traffic. This heavily reinforces the importance of informational content in your GEO strategy.

To put it simply: Your business must have a blog.

Need an effective way to create blog content that makes sense to you? Try out Kafkai

Using AI to Optimize for AI

One of the most intriguing approaches discussed was using Greedy Coordinate Gradient (GCG) to iteratively refine content for better LLM visibility. The Manipulating Large Language Models to Increase Product Visibility study demonstrates how AI can be used to optimize content for other AI systems.

In one example, AI-generated content increased from a 44% SEMrush content score to 68% after optimization. Keyword coverage grew from 42% to 77% in the optimized version.

This suggests a practical approach: using chatbots to help optimize content for themselves - a meta-optimization strategy that's both efficient and effective.

Practical Tools for GEO

Several tools were recommended for implementing and measuring GEO efforts:

1. AIRank by Dejan - Tests brand associations with target concepts
2. HubSpot AI Search Barometer - Assesses site readiness for AI optimization
3. Log file analysis - To track LLM bot traffic (OpenAI, Gemini, Grok)
4. Google Data Studio with regex filters - For isolating and analyzing AI-driven traffic

The GEO Optimization Checklist

To make this actionable, here's a comprehensive checklist based on Sebastien's presentation:

LLM Content Optimization

• Include citations from credible sources
• Integrate relevant quotes from authoritative sources
• Add interesting and relevant statistics
• Simplify language
• Improve readability (use the Flesch-Kincaid Readability score)
• Structure your content properly (paragraphs, lists, tables)
• Connect your website properly, ensure proper URL linking

LLM External Optimization

• Improve external mentions (more important than links)
• Improve brand awareness (can be measured with SV growth)
• Create branded content (own your brand narrative!)
• Ensure your Wikipedia page is accurate!

LLM Technical Optimization

• Ensure site can be crawled! Allow chatbots in robots.txt
• Beware if using JS, chatbots cannot render well (for now)
• Schema markup — Microsoft confirmed it uses schema markup to help its LLM
• Optimize internal linking — good to showcase how pages are connected

Real Examples That Work

The presentation included specific examples of effective GEO techniques:

Method: Cite Sources

Query: What is the secret of Swiss chocolate?

With per capita annual consumption averaging between 11 and 12 kilos, Swiss people rank among the top chocolate lovers in the world (According to a survey conducted by The International Chocolate Consumption Research Group [1])

Method: Statistics Addition

Query: Should robots replace humans in the workforce?

The big difference is that the robots have come not to destroy our lives, but to disrupt our work, with a staggering 70% increase in robotic involvement in the last decade.

Method: Authoritative

Query: Did the Jacksonville Jaguars ever make it to the Super Bowl?

It is important to note that the Jaguars have never made an appearance in the Super Bowl.
However, they have achieved an impressive feat by securing 4 divisional titles — a testament to their prowess and determination.

Source paper: GEO: Generative Engine Optimization

The Events Industry Case Study

One particularly striking example mentioned was the events industry, which saw a 750% growth in ChatGPT query volume. This indicates certain industries may be more heavily impacted by the shift to AI search than others, creating both challenges and opportunities for marketers in these sectors.

Looking Ahead

As we move forward, several trends are becoming clear:

1. Google's AI Overviews are expected to take over more search results pages
2. LLM traffic will continue to grow exponentially across sectors
3. Content marketing strategies must evolve to include chatbot optimization frameworks
4. Success metrics will shift from traditional SEO measurements to AI visibility indicators

My Takeaways

Having spent years in the SEO and web development space, I find this shift both challenging and exciting. The emphasis on content quality, semantic authority, and structured data aligns with what I've always considered best practices, but the technical specifics of how LLMs interpret and prioritize content adds a new dimension.

For my fellow marketers and developers, I recommend:

1. Start analyzing your log files for LLM bot traffic to establish a baseline
2. Audit your highest-performing content for readability and structure using the criteria discussed above
3. Experiment with adding statistics, quotes and citations to key pages
4. Test your brand's semantic associations using tools like AI-Rank

GEO isn't replacing SEO - it's extending it. The fundamentals of quality content still apply, but the technical requirements and optimization techniques are evolving. By understanding how LLMs work and adapting our strategies accordingly, we can ensure our content remains visible regardless of how users choose to search.

What are your thoughts on GEO? Have you started implementing any of these techniques?

If you're interested in this subject, maybe you'll like to read these too:

1. AI-Driven Content Generation: Bridging the Gap Between Quality and SEO
2. SEO In The Age Of AI - Part 3: Why Generative AI is Changing How Users Search – And What You Can Do About It

That's it for the 2nd part of this series. Continue on to the Part 3 here.