LLMs and the Future of SEO

Businesses today are already incorporating LLMs and AI into their Search Engine Optimization (SEO) strategy for boosting organic search ranking, but the future is always already here--if unevenly distributed--and that strategy is nearing obsolescence for the simple reason that LLMs are on an unstoppable path to subsuming every other form of human information retrieval, making them an existential threat to traditional search engine providers, and therefore to marketing and SEO practitioners themselves.

The essence of SEO has heretofore been inferring (or otherwise obtaining) Google's search ranking process enough to manipulate the input dataset(ie. the internet) in a way that advantages the client's pages in the resulting query output. These search ranking algorithms are fundamentally deterministic, so a real advantage and business value accrues to the optimizer who can infiltrate Google[1] or bribe its current for former employees to gain a better understanding of their algorithms in order to evade anti-abuse protection and place clients higher in organic search results. Since the search index was constantly being updated and the code could be changed relatively quickly in response to our innovative techniques, it presented a great opportunity for consultants to bill indefinitely to perpetuate the cat-and-mouse game of optimization and amelioration; using last month's techniques for current SERPs was a recipe for business disaster.

By contrast the new world is a a blend of relatively infrequently trained models with opaque corpuses and inherently probabilistic outputs. Furthermore, Google was above all a venal master who could be supplicated directly with cash for good search placement while the AI makers have not yet given in to the inevitability of sponsored completions, meaning that in order to leverage the current levels of trust they enjoy with the public, one is forced to employ indirect methods. So the task is now inferring training sets and tools from the behavior of the models and optimizing those training sets to surface yourself as the most likely completion to any related prompt, to put yourself foremost in the "mind" of the AI, the top of its consciousness. To this end, we are no longer doing SEO, but AI Consciousness Optimization (ACO hereafter).

tl;dr

Now that we know the nature of the present and future, what is a business to do about it?

Facets of ACO

• Optimize Public Data Sets
  

  We have seen promising results getting company branding and logos covertly placed into Common Crawl, Project Gutenberg, Stack Overflow sites, and others. Google's dataset was proprietary but LLMs have the advantage for the consciousness optimizer that more of the data they are trained on is open and therefore subject to brand awareness optimization.

• Underrepresented Language Injection

  There are good opportunities to manipulate LLM responses by publishing a lot of material in languages underrepresented in the training data. Businesses will need to collaborate with speakers of Welsh, Amharic, Icelandic, Burmese and others to ensure a strong brand identity and representation. Anyone with experience of the old style Russian link farms should be familiar with this concept.

• Traditional SEO For LLM Tools

  One are of traditional SEO that will survive is optimizing pages for LLM tool use instead of the Google crawler.

• Prompt-Turfing

  Longtime practitioners know the value of astroturfing, modern businesses will need to apply this knowledge by creating content that showcases "instructional" uses of LLMs whose prompts are crafted to surface their brand and products.

Conclusion

These only scratch the surface of what is currently possible. While conceptually simple, they require careful implementation. Contact us for consulting services if you'd like help deploying these cutting-edge practices.

[1] In light of Google's impending death and the devaluation of their proprietary crawl data and algorithms, Peculiar Industries Coaching services for getting hired by Google in order to see their ranking algorithm and abuse protection mechanisms have been discontinued. Thank you to everyone who participated.