Key Takeaways
- GEO has no unique ranking signals – AI answer engines search the web, read top-ranking pages, and generate answers via next-token prediction
- 87% of ChatGPT citations match Bing’s top organic results – AI search is traditional search with a chatbot on top
- 92% of Google AI Overviews cite domains already in the top 10 organic results
- The “GEO tactics” that work – citing sources, adding statistics, writing clearly – are just good SEO rebranded
- Ranking well in traditional search is the GEO strategy – if you’re in the search results, you’re in the LLM’s context window
The Biggest Lie in SEO Right Now
GEO – Generative Engine Optimization – is the hottest acronym in SEO right now.
It’s also completely made up.
Not the acronym. Agencies love acronyms. The lie is what’s underneath it: this idea that AI answer engines use some secret set of ranking signals, and if you just optimize for THOSE signals, you’ll show up in ChatGPT and Perplexity and Google AI Overviews.
That’s the pitch. And it sounds great on a slide deck.
The problem? There are no secret signals. There’s no new ranking algorithm to reverse-engineer. And fewer than 12% of marketing teams have a documented GEO strategy – which tells you even the industry buying this stuff isn’t buying it that hard.
Meanwhile, 40–60% of sources cited by AI answer engines change month to month. One EMARKETER analyst put it bluntly: “Almost every GEO response is different from every other GEO response.”
That’s not a discipline. That’s a coin flip with a retainer attached.
How Every LLM Answer Engine Actually Works
Here’s what’s actually happening when you ask ChatGPT or Perplexity or Google AI a question. It’s not complicated.
The system takes your query. The LLM breaks it into sub-queries – a query fan-out – plus grounding queries to verify facts. Then it runs a real-time web search. ChatGPT uses Bing. Google uses Google. Perplexity uses multiple engines.
The LLM opens and reads the pages that rank for those queries. Sometimes a handful. Sometimes dozens.
Then it generates an answer using next-token prediction – the fundamental mechanism of every large language model. Which word is most probable given everything in the context window?
That’s it. The whole thing.
The technical name is Retrieval-Augmented Generation – RAG. It’s the standard architecture for every grounded LLM response on the market. The LLM doesn’t have its own authority algorithm. It doesn’t “rank” sources. It searches the web, reads what’s already ranking, and predicts the next token based on that context.
No secret signals. No magic. Search → context → prediction.
87% of AI Citations Come From Search Results
If the process is “search the web, read what ranks,” then the citations should reflect that.
They do.
Seer Interactive found that 87% of SearchGPT citations match Bing’s top organic results. Most from the top 10 positions. ChatGPT isn’t surfacing hidden gems from the depths of the internet. It’s reading Bing’s first page and rephrasing it.
On Google’s side, 92.36% of AI Overviews link to at least one domain already ranking in the top 10 organic results.
Read that again. 92%.
Google’s AI is citing Google’s search results. ChatGPT is citing Bing’s search results. The entire promise of GEO is that AI search is some new frontier with its own rules. The data says it’s the SAME frontier with a chatbot strapped to the front.
GEO “Tactics” Are Just Good Content With a New Invoice
Here’s where it gets funny.
The most-cited academic paper on GEO – out of Princeton – tested a bunch of optimization methods. The winners? Citing sources. Adding statistics. Including quotations.
That’s just good content. That’s what competent writers and SEOs have been doing for years.
“Fluency optimization” made the list too. As in – write clearly. Use real sentences. Don’t produce garbage.
These aren’t GEO tactics. They’re the bare minimum for content that deserves to rank anywhere, in any search engine, powered by any technology.
But slap “GEO” on it, build a service page, and suddenly you can charge for it separately. Same game the industry plays every time something new comes along. New acronym, same work, bigger invoice.
You Don’t Need a GEO Strategy
If you rank well in traditional search, you’re already in the LLM’s context window. That’s the whole game.
The LLM is gonna search the web. It’s gonna open the pages that rank. It’s gonna read YOUR content if you’re ranking for the queries it cares about. And it’s gonna generate answers based on what it read.
There’s nothing to “optimize for” beyond what you should already be doing. Strong content. Real authority. Technical fundamentals. The stuff that works in SEO works in AI search – because AI search IS search with a language model on top.
The folks selling you GEO as a separate strategy are selling you SEO with extra steps and a new line item.
Don’t buy the fan fiction. Read the source code.
Reach Out
Tired of paying for made-up disciplines? Reach out on LinkedIn or shoot me an email at tomislav@tomislavhorvat.com.
