Generative Engine Optimisation

[BakerStreet]

Hey everyone,
So, I've been researching optimisation for ChatGPT and the like. I thought it would be a good idea to start a thread about this, as it is the future. I've been talking to a few Prompt Engineers and we found two good sources. Let me share them with you.

1. https://ig.ft.com/generative-ai/ - This article shows you how analysing the terms that appear on the page (called encoding) allows transformers to learn what words usually appear together. With this, transformers can use beam search technology to figure out whole sentences that belong together. This allow them to write sentences that makes sense, even if the word has multiple meanings. For example, for the sentence "I took a walk next to the bank..." it can be "and they had an ad for 5% interest on a CD" or "and the river was flowing fast". Both independent clauses make sense for the word "bank" and the right would would depend on other words in the prompt.
2. https://searchengineland.com/decoding-llms-generative-ai-search-results-448630 This article goes over how LLMs work and what Retrieval Augmented Generation is. In short, RAG supplements generative AI by having Generative AI read the top 10 results in Bing or Google. This is to reduce hallucinations. If you want to optimise for generative AI, you should look at where they get their training data and their index from. ChatGPT uses Bing and Gemini uses Google. So you want to be indexed in Bing and Google. Other than that, make sure your answer is correct, authoritative, and well cited in other publications on the Internet. For example, if you are in travel, it will be good to have your brand mentioned on TripAdvisor. So "BS Travel" should be occurring next to "Bratislava vacation" for example, if BS travel was a Bratislava travel agency. That way, generative AI knows to associate BS Travel with Bratislava vacations.
   Also, this article says that there's a 50% correlation from the top 10 results of Google/Bing with the results that shows up on Generative AI. What Generative AI does is that it turns the prompt into a query and then runs the query through the search engine. Generative AI is just another layer on top of the regular search engine. In the article, he used an example for the prompt "I am 100kg and want to run 10km a day. what are good running shoes for me?" The LLM figured out that he was overweight and performed the query "best running shoes for overweight person." The generative AI result was a summary of the top 10 pages for that query.

Here's my takeaway from this:

1. You want your brand associated with this product or service. You need to position your brand like that. Brand is more important now than ever and will become even more important in the future.
2. Content is king. Your content needs to be well written and correct. No more $1/100 words content from someone who is not a native speaker.
3. The pie is shrinking. Generative AI will result in lower CTRs. There's no doubt about it. Since RAG summarises the top 10 results for a query, if you are a content site that is based upon ad impressions, you're going to die. Really. Amazon Affiliates? maybe you can survive but good luck.
4. The starting capital requirement went up. You need more capital to enter the search channel. No way around it. RAG will lower traffic. Sucks but true. You need to scale with a lot of content that's expensive to write.

I'm seeing some traffic from ChatGPT already but its tiny. Right now, I'm moving into products and adult, since both are a hedge against Generative AI. But obviously, moving into GEO would be a good move as well. I definitely would not make a content based, amazon affiliate site nowadays.
I'm scared of RAG since it will disrupt things a lot. But you need to adapt and adapt fast if you want to succeed in today's VUCA world.
What about you? What do you think?