The AEO and GEO Gold Rush Is Repeating SEO's Oldest Sales Pitch

We are in the early years of answer engine optimization (AEO) and generative engine optimization (GEO), and the ground shifts almost weekly. That is exactly the environment where confident claims sell best, and where buyers should be most careful.

We Have Heard This Pitch Before

Anyone who bought search services in the early 2000s remembers it. Every SEO firm understood the Google algorithm. Some knew a guy who knew a guy at Google. All of them had a secret sauce.

Google eventually addressed this in its own documentation. Under the heading "No one can guarantee a #1 ranking on Google," it tells site owners to beware of SEOs that claim to guarantee rankings, allege a "special relationship" with Google, or advertise a "priority submit" (Google Search Central).

Change the nouns and the pitch is back. Monitoring dashboards promise to track your share of voice inside ChatGPT. Content enhancement tools promise to rewrite your pages so AI assistants will cite them. The line works for the same reason it worked twenty years ago: fear of missing out. If a competitor shows up in an AI answer and you don't, someone will call it an emergency, and they will have a product ready.

We recently sat through a demo from one of the larger AI visibility platforms, a tool licensed for upwards of $50,000. We asked a simple question: "How do you get analytics and metrics out of ChatGPT?" After a long pause and a round of glances across the table, one of their team answered with complete confidence: "We just guess." Holy. Shit. At least they were honest.

"We just guess."

That answer was more honest than the pitch around it. Nobody outside the AI labs knows exactly how these systems decide what to say, and that includes the people selling the tools.

What the Monitoring Tools Actually Measure

Start with monitoring, because it looks the most scientific. In early 2026, SparkToro's Rand Fishkin and Gumshoe.ai's Patrick O'Donnell had roughly 600 volunteers run 12 identical prompts through ChatGPT, Claude, and Google's AI nearly 3,000 times. The chance of getting the same brand list twice was under 1 in 100. The chance of the same list in the same order was closer to 1 in 1,000 (Search Engine Land).

That does not make tracking worthless. The same study found that strong brands appeared in 60 to 90 percent of responses for a given intent, so how often you appear across many runs carries real signal. A dashboard that says you "rank third in ChatGPT," however, is reporting noise. If a vendor leads with position, ask how many runs sit behind each number.

Two Systems, Not One

To cut through the claims, separate what an AI assistant knows from what it looks up. They are different mechanisms with different rules.

The Knowledge Frozen in the Model

A large language model (LLM) learns from a training snapshot. What it knows on its own stops at a knowledge cutoff, and vendors publish those dates. Anthropic, for example, lists a training data cutoff for every Claude model (Anthropic). Models are refreshed more often than they once were, but each release is still a photograph of the web at a point in time, usually taken months before you use it.

So yes, being present when that snapshot is taken matters. But you cannot time it, and nobody can sell you a seat at it. The only reliable strategy is to be consistently present, on your own site and across credible third-party sources, so that whenever the next snapshot is taken, you are in it. That is another way of describing a durable reputation on the web, which is what good SEO always aimed for.

The Live Retrieval Layer

The second system is real-time grounding. The model runs a search, reads the results, and answers with citations. Google's Gemini API offers Grounding with Google Search, which connects the model to real-time web content so it can cite sources beyond its knowledge cutoff (Google AI for Developers). Claude's web search draws on Brave Search, which Anthropic lists among its web search subprocessors (TechCrunch). OpenAI runs its own crawler, OAI-SearchBot, to surface sites in ChatGPT search (OpenAI). Perplexity maintains an index covering hundreds of billions of web pages (InfoQ). Developers can even connect DuckDuckGo results to an assistant through community-built Model Context Protocol (MCP) servers, which let a model call a search engine as a tool.

Look closely and this is classic search. A crawler fetches your page, an index ranks it, and the model reads the top results. The signals that decide which pages make that cut are the signals SEO has always worked on.

The Fundamentals Have Not Changed

Google said it plainly in its own guide: from Google Search's perspective, optimizing for generative AI search is optimizing for the search experience, and it is still SEO. The guide says you do not need llms.txt files, Markdown versions of pages, content chunking, special schema, or AI-specific rewriting to appear in its AI features (Google Search Central).

My colleague Marla Quinn has pointed out that Google's guide only governs Google's surfaces, and she is right. But the non-Google systems reward the same basics, often more strictly:

• Server-rendered, fast pages. Vercel's analysis of AI crawler traffic found that none of the major AI crawlers, including those from OpenAI, Anthropic, and Perplexity, rendered JavaScript. If your content appears only after client-side scripts run, those bots see an empty page (Vercel).
• Clean site architecture. The same analysis found ChatGPT and Claude crawlers hitting 404 pages on roughly 34 percent of fetches, compared with about 8 percent for Googlebot. Broken links and redirect chains cost you more with AI crawlers, not less.
• Accurate structured data. Schema markup that matches the visible page still helps indexes understand your products, people, and organization. It is hygiene, not a magic lever.
• The answer near the top. Retrieval systems pull passages, and models work within token budgets. A page that states its main point early and supports it below is easier to quote. That is also just good writing.

What About llms.txt and Markdown Files?

These are the tactics AEO and GEO agencies push hardest. They are cheap to add, and I have no objection to adding them. An llms.txt file can help coding agents and documentation tools navigate a site.

Keep the effort in proportion, though. Originality.ai found llms.txt adoption grew 8.8 times between June 2025 and May 2026. Yet Ahrefs server logs across 137,000 domains showed that 97 percent of those files received zero requests in May 2026, and AI retrieval bots made only about 1 percent of the requests that did arrive (PPC Land). Add the file in an afternoon. Do not pay a retainer for it.

Where the Real Effort Belongs

Most of your effort belongs where it always has: E-E-A-T, Google's shorthand for experience, expertise, authoritativeness, and trustworthiness. Google is clear that E-E-A-T is not a specific ranking factor and that, of the four, trust matters most (Google Search Central). It describes the qualities ranking systems look for, and those are the same qualities any AI system needs to see before it will cite you.

The best independent research points the same way. The paper that coined "generative engine optimization," presented at KDD 2024 by researchers from Princeton and partner institutions, found that citing sources, adding quotations, and including statistics improved visibility in AI-generated answers by up to 40 percent. Keyword stuffing performed worse than doing nothing (arXiv). Credible, well-sourced writing wins. That is not a new discipline. It is editorial standards.

A house built on a sound foundation does not need a new one every time the weather changes. Neither does your content strategy.

Practical Takeaways

Before you sign an AEO or GEO contract, hold the vendor to these questions:

1. Which system does this tactic affect? Training data, a specific retrieval index, or neither? If they cannot name the mechanism, they are guessing.
2. What evidence supports it? Look for platform documentation or independent research, not the vendor's own case study.
3. How is visibility measured? Ask for appearance rates across many runs, not a single "AI ranking."
4. Is the technical foundation sound? Server-rendered content, fast load times, clean links, and accurate schema come first. They pay off in every system.
5. Is your content worth citing? Original data, named experts, clear sourcing, and the answer up front.

The rules that applied to SEO are the rules that apply to AEO and GEO. Anyone who claims a tool that will place you inside ChatGPT or Claude is selling something that is, at best, partially true. Build the foundation every system rewards, add the inexpensive extras if you like, and spend the rest of your budget on content that deserves to be quoted. For a broader view of how the three disciplines fit together, see our 2026 SEO framework for enterprise content teams, or start with an AI content audit of what you already have.