As AI summaries compress traditional organic space, referral traffic to publishers is reducedcreating a world where search is used more than ever, but where the search interface also retains users longer, where it has historically acted as a standard.
The outlook for publishers is pessimistic, with traffic should decrease by half in the next three years.
Contrary to this reality, Google recently reported that search queries hit an all-time highsuggesting a golden age for digital visibility.
AI has given search superpowers, and as a result, people are searching on Google more than ever. Last quarter, we saw an all-time high in search queries.
Google also updates announced to try and send traffic to websitesit remains to be seen whether this is a public relations move aimed at fending off antitrust cases.
What all this highlights is that despite opinions that SEO and search engines have been replaced by GEO/AEO and LLMs, this couldn’t be further from the truth. Search engine optimization remains relevant and technical SEO is the basis of AI search.
LLM Prediction and Grounding
Large language models are probabilistic text generation engines, not databases or reasoning engines. They do not retrieve stored facts; they calculate the statistical likelihood of word sequences.
To make these responses current and well-founded, recovery augmented generation (RAG) retrieves documents from a search index and passes them to the model before it writes its response. A very good explanation of this was posted on YouTube in December 2024 by Jess Peck: “Oh my god, ChatGPT is not a search engine.“
For an AI search engine to answer a query using RAG, it relies on a high-quality data pipeline. It needs an organized, easily navigable and authoritative data source, which is only possible through semantic HTML, logical site hierarchy and clean indexing provided by SEO professionals.
Who creates, structures and manages this data source? The SEO community. We’re the ones who label the data, clean up the mess, and ensure machines can actually read what humans write.
Without the fundamental SEO architecture (semantic HTML, logical site hierarchy, and clean indexing path), AI search engines are left with inefficient website paths and structures. SEOs are not passive victims of the AI revolution.
Modern SEO now encompasses both the traditional work of maintaining site health and specific AI-readiness strategies, optimization for RAG extractionand reinforcing brand entity signals across the knowledge graph.
By structure data so machines can interpret contextSEO professionals provide the exact signals that AI search engines use to verify facts and attribute sources. Technical SEO ensures that the “information gain” of a page is accessible to models who need to cite it. If you want an AI to recommend your product, your digital footprint must support it.
Optimization is not disappearing in the AI era; this becomes the basis of trust.
Can you optimize LLMs without an SEO program?
The smartest brands aren’t abandoning SEO for AI search; they are aggressively using SEO to fuel their AI readiness.
They understand that AI search does not replace the need for traditional SEO and information retrieval practices; this highlights the need to think even more about SEO and information systems.
It’s easy to think of SEO as a relic, but the reality is that SEO built the product that LLMs now produce and charge for.
Jamie Indigo I summed it up perfectly on LinkedIn:
… We need to be clear-eyed about what happened – and intentional about what we build next.
SEO manages the engine room that powers the ship. As AI changes the digital landscape, the engine room is the best place to prepare and prepare for the future. As we look toward an era of GEOs and AEOs, we have to ask ourselves: can you be a GEO/AEO and optimize for LLMs if you don’t have that SEO knowledge and expertise base?
More resources:
Featured image: Salomi art/Shutterstock
