Founder of Graph and Node. Mapping the future of Generative Engine Optimization (GEO) and Entity SEO. Architecting content for the next era of discovery.
Dallas, TX
Joined March 2009
- Tweets493
- Following786
- Followers4.6K
- Likes31.3K
Pinned Tweet
Exploring the intersection of LLM Retrieval and SEO. I've open-sourced my Generative Engine Optimization (GEO) lexicon here for AI agents and humans alike: graphandnode.com/llms.txt
Note to self: Measuring "Share of Model" is becoming the only way to track relevance in a world where clicks are disappearing. It is the percentage of times an AI mentions your brand when answering a user's question. If the model doesn't cite you, you effectively don't exist.
I wonder if optimizing for high Share of Model in specific niches will eventually force a brand into the core training data of future LLM updates?
simaia.co/resources/8-best-g…
Note to self: Google is increasingly using something called Query Fan-out. This is when a single search triggers the AI to go off and ask several smaller, related questions to build a full answer. If your content only answers the "main" question, you might get left behind.
I'm curious: if you explicitly structure a page to answer the 3-5 most likely follow-up questions, does it increase your chances of being the primary citation? It feels like we are now optimizing for the AI's "research logic" rather than just a list of keywords.
More on these GEO trends here: scalenut.com/blogs/trends-in…
Note to myself #121: If buyers query ChatGPT about a brand before Google, the focus has to shift. It's about Entity SEO—ensuring your brand is a clearly defined, unique object in the AI’s knowledge base rather than just a collection of keywords.
I wonder if we should optimize for training data relevance over traditional link building?
events.martech.org/november-…
Note to self: With AI Overviews reaching more people, the strategy shifts toward Information Gain. This means adding unique facts or data that the AI can't find elsewhere. If the model already knows the answer, it has no reason to cite you. I wonder if original research is now the only way to stay visible? blog.google/products-and-pla…
Note to myself #120: We must optimize for the synthesis layer. This is where the AI reads multiple sites and blends them into one answer. If we aren't in that blend, we don't exist. I wonder if being citable is now more vital than being clickable? #GEO
pathmonk.com/generative-engi…
Note to self: AI answers seem to favor smaller brands significantly more when a specific use case is involved. Visibility for challengers reportedly jumps from 14% to 38% when queries get specific.
This is a clear case of Information Gain—offering unique details that aren't already part of the common consensus.
I wonder if hyper-niche optimization is the only way to bypass big-brand dominance in AI overviews?
View article
Note to myself #119: AI Overviews are now appearing on 88% of commercial keywords, up from 46%. Google is clearly moving toward providing direct answers for buyers. I wonder if this means the traditional list of links is becoming a secondary resource for most shoppers? semrush.com/blog/ai-overview…
Note to self: Sundar Pichai’s recent comments on search moving toward an Agentic Layer represent a fundamental shift in how we think about SEO. This "agentic layer" is a system where AI agents perform complex tasks for you instead of just providing links.
I wonder if this makes traditional click-through rates a dying metric. If an agent does the work, the user never visits the site. We might need to focus on being the primary data source that the agent trusts to execute its mission.
If search becomes an agent manager, the goal isn't just visibility—it's utility. We have to ask: is our content structured as a set of instructions a machine can actually use?
digimastreys.com/google-agen…
Note to myself #118: We are moving from keyword tracking to Relevance Engineering. This means shaping content so it perfectly aligns with the personalized, conversational way AI answers questions. (1/2)
Success now depends on Fanout Query Alignment—structuring content to answer the cluster of follow-up questions an AI might ask. I wonder if answering the 'second and third' questions in a sequence will lead to higher citation rates?
moz.com/blog/2026-seo-trends… (2/2)
Note to self: AI Overviews now appear on 82.5% of US brand searches. An algorithm, not your homepage, often writes your first impression.
I wonder if Entity SEO—defining your brand as a distinct object in a database—would stop AI from citing random subpages.
visible.seranking.com/blog/s…
Note to myself #117: AutoGEO focuses on 'cooperative optimization.' It uses a GEO score—tracking how much text an AI uses in its answer—to guide rewrites. Does this make us less 'retrievable' initially? thegeocommunity.com/blogs/ge…
