GEO & AI Search Engines: How AI Sees Content Quality

When we were researching our 6 Generative Engine Optimization Strategies, some natural questions that came up are: Can AI Search Engines give a score for sites? What were the metrics?

Apparently we were doing something right, seeing our pages as search results with marketing or Artificial Intelligence (AI) topics in Perplexity, ChatGPT, or Gemini. The result? I think initially it was just dumb luck, but we were on the right track.

What is Generative Engine Optimization?

Generative Engine Optimization (GEO) is all about making your content more visible and effective on AI-driven platforms. The goal? ensure your content appears authoritative, trustworthy, and capable of answering questions conversationally.

GEO is different from SEO and while sharing similarities, they serve different purposes. SEO, or Search Engine Optimization, focuses on optimizing websites to rank higher in traditional search engine results pages (SERPs).

"Generative models, particularly those fine-tuned for specific tasks, play a crucial role in optimizing content visibility on AI-driven platforms by aligning content with user expectations" (Wang, Zhou, Wang, & Peng, 2024).

GEO, on the other hand, is focused on optimizing your content so it shows up in AI Search Engine outputs. The timeline for seeing results can vary. For example, non-internet access tools like ChatGPT relies on periodic knowledge updates. Search based tools, like Perplexity, update more frequently, sometimes even daily.

What is the Criteria for GEO?

We broke down the criteria for GEO into 10 main categories. High-Quality Content, User Intent and Experience, Natural Language Processing (NLP) Optimization, the Structured Data and Schema Markups, Voice Search Optimization, Mobile Optimization, Content Structure and Readability, Authority and Trustworthiness, Technical SEO, and finally Engagement & Social Signals. Each category does have overlap with the other, but given how detailed the equation is, this way it is more digestible.

Now here’s the kicker. In the other article, we focused on traditional SEO strategies - measuring bounce rates, impressions, and engagement. But AI Search Engines don't measure these the same way.

"Generative Engines, in contrast to traditional search engines, remove the need to navigate to websites by directly providing a precise and comprehensive response" (Aggarwal et al., 2023).

Instead, AI Search Engine models rely on alternative metrics like Position-Adjusted Word Count and Subjective Impressions to assess visibility and relevance of your content. These metrics are specifically designed for GEO, not traditional SEO metrics like bounce rate or CTR. Radford et al. (2019) highlighted the potential of language models to perform a variety of tasks without explicit supervision, making them valuable tools for GEO. Which in this context, involves creating the pathways needed for success.

Below is an example of how AI Search Engines, like Perplexity, can calculate engagement score instead of having direct access to SEO tools.

Engagement Score_{AI Search} = ω₁ × Content Quality + ω₂ × Query Refinement Rate + ω₃ × Link Reference Frequency + ω₄ × Sentiment Analysis

For this article, we are zeroing on the High-Quality Content aspect of GEO. What makes your content high-quality or low-quality in the eyes of AI? What are the equations that AI Search Engines use to calculate those metrics? And most importantly, how can we, as marketers, gamify that process?

High-Quality Content: AI Perception and Evaluation

To understand how AI Search Engines perceive content for GEO, we need to break down the criteria used by AI-driven models and algorithms. AI grades high-quality content based the following metrics:

1. Relevance and User Intent
2. Comprehensiveness
3. Accuracy and Freshness
4. Readability and User Experience

Since there is a lot of information, we'll break each part down into an explanation of what it means in context, examine the relevant equations and math, and finally discuss what makes it a good score. At the very end of the article, we'll take a look at a live example to see the score, and what takeaways we can walk away with.

"The quality of content is increasingly being determined by how well it integrates natural language processing techniques, particularly in how it engages with user queries." (Vinutha & Padma, 2023).

Things to Keep in Mind:

🔍 You’ll see mentions of “the keyword”. The AI Search Engine grabs it from your article title, the meta title and the meta description. It identifies the primary keywords and checks if the article aligns with those keywords. Finally, it verifies that the body content uses these keywords in a relevant and user-focused manner.

📓 A “good” score is different between each of the metrics because each metric takes a look at a different aspect of content quality. Some metrics, like accuracy, need to be more precise in the context of user intent, while others, like relevance, allow for more flexibility.

Relevance and User Intent

Relevance is about how closely content aligns with the search query or topic at hand. It’s all about making sure that the content directly addresses what the user is looking for. Intent, as the name suggests, refers to what the user hopes to achieve with their search. We can generally break down intent into three categories:

• Informational: The user is looking for information, like “How to write a blog post.”
• Navigational: The user is looking to find a specific website or page, for instance “Amazon customer support.”
• Transactional: The user is looking to take a specific action, such as “buy dancing shoes for large feet online.”

As highlighted by Liang-Ching and Kuei-Hu (2023), "[It] allows for a structured assessment of keyword relevance, which is critical for aligning content with user intent in GEO." Separately, Asai et al. (2021) explores how different query types influence the effectiveness of content retrieval, a crucial consideration in optimizing for generative engines.