In the rapidly shifting landscape of digital marketing, a new buzzword has taken center stage: GEO, or Generative Engine Optimization.
If you read the latest marketing newsletters, you’ll see everyone looking into GEO today. It is being pitched as the shiny "next generation" of SEO—the silver bullet to draw traffic in an era where users are talking to chatbots instead of typing into search bars. However, this narrative is misleading. While the intention behind optimizing for AI is valid, the concept of GEO as it stands today is very ill-constructed.
The Simplicity of SEO vs. The Complexity of AI
To understand why GEO is a fractured concept, we first have to look at why SEO worked so well.
SEO (Search Engine Optimization) is actually easy to understand conceptually. Why? Because for the last two decades, Google dominated the search engine market. When you optimized for "search," you were really just optimizing for Google’s specific algorithms. There was a clear set of rules: use these keywords, build these backlinks, structure your schema this way, and improve your Core Web Vitals. If you pleased the Google bot, you won.
GEO is quite different. If SEO was for Google, and GEO is for ChatGPT, Claude, Gemini, Deepseek, Le Chat, and Qwen, and Manus, and Doubao, and Minimax, and Microsoft Copilot, and z.ai, and ... There is no single "AI algorithm" to please. To understand this, let's look at the mechanics of how an LLM model or a chat agent actually finds a product or information about your brand. It is not a straight line; it is a maze.
How an LLM Actually Finds You
When a user asks an AI about a product, the answer can come from a chaotic mix of sources. Here is the breakdown:
- From Training Texts: The model might "know" you because your brand was mentioned in the massive dataset it was trained on (likely data from 2024 or earlier). You cannot "optimize" this retroactively. It is frozen in time until the next training run.
- When 'Search' is Enabled: If the AI browses the web (like ChatGPT with Search or Perplexity), it is pulling data from different search engines and API providers. It might be using Bing, Google, or a proprietary index. You aren't optimizing for the AI; you are optimizing for the underlying search engine it relies on.
- When RAG (Retrieval-Augmented Generation) is Used: In enterprise settings, the AI might be pulling answers from unknown private documents, internal wikis, or vector databases.
- When 'Deep Research' Occurs: Advanced agents don't just look at the top result. They dig much deeper, following links 3 or 4 layers down, synthesizing information from obscure forums or PDFs that traditional SEO might ignore.
- When Using Tools/MCP: With the rise of the Model Context Protocol (MCP) and function calling, the AI might find you through specialized sources—your email history, Slack logs, or a direct API connection to an e-commerce platform like Shopify or Amazon.
- General Chatbots vs. Special Agents: AI agents are the not same. General chatbots rely on broad, pre-trained knowledge from large, diverse datasets, while specially tailored agents use narrower, curated, and sometimes real-time or proprietary data sources.
- Prompt Variance: Even within a single agent, the effective "source" of information is fluid. Different prompts(system prompts) activate different parts within the model's neural network, results bias to steer the model to yield significantly different results from the same user request.
- Keywords search vs. Vector search: The technical retrieval methods vary wildly. Some agents use keyword search (lexical matching), while others use vector search (semantic meaning, yet different embedding/rerank models result different relevancy). Most use a hybrid of both.
The Fractured Reality of GEO
GEO is quite fractured.
GEO is a single term, but requires you to be everywhere for every agents in everywhere formats.
If you really think about it, "how an LLM finds you" is frequently just "how an LLM uses a search engine to find you." If an AI agent uses Bing to answer a user's query, "GEO" is effectively just "Bing SEO." If the AI uses a tool to search Amazon, "GEO" is just "Amazon Optimization."
SEO became a standardized industry discipline pretty much after Google dominated the market and stabilized the rules of engagement. The search engine giants evole their ranking algorithms mostly to fight exploiting of their mechanims, the core principles of SEO don't change much.
In contrast, LLMs are still constantly evolving. Not only the models, the way of agents doing things, the tools/skills agent shall use, the whole system may be completely different next month. GEO is merely an attempt to ride the LLM hype，clumsily stitching together unrelated concepts.
GEO as a concept is pretty much something that neither the seller nor the buyer understands.
Agencies are selling it because it sounds cutting-edge. Clients are buying it because they fear missing out. But currently, you cannot do much specific "GEO work" that isn't just "good marketing," and you certainly cannot measure it with the precision of Google Analytics.