Site Speed and AEO – An Unanswered Question

The Enduring Paradox of Site Speed in the Age of Conversational Search

The digital landscape is undergoing a seismic shift, one that fundamentally challenges the long-held dogma of Search Engine Optimization (SEO). For years, the metric of site speed—epitomized by Google's performance scores and the mantra of Core Web Vitals (CWV)—was treated as an ultimate goal. While achieving a perfect "green score" once dominated tech team roadmaps, the relevance of this singular focus is now being called into question as we pivot toward an era dominated by Answer Engine Optimization (AEO) and Large Language Models (LLMs).

How weird does the focus on a simple 100-point Google score sound now? Google itself, the historic arbiter of web performance, currently lists only about 10 organic items per search results page. More importantly, search and engagement are changing in almost every possible way—right now. The conventional organic ranking model, while still present, accounts for a diminishing share of overall traffic, leading to fragmentation across user acquisition channels.

The Conversational Shift: Atlas, Comet, and the New Crawl

The launch of major, commercially powerful generative AI search engines, such as OpenAI's Atlas this week and Perplexity's Comet a few weeks prior, marks a definitive inflection point. These platforms are not merely indexing pages; they are consuming, synthesizing, and answering queries, driving a massive shift in how information is accessed and consumed.

In this evolving environment, the reliance on traditional organic rankings is waning. However, this does not mean site speed is becoming irrelevant. On the contrary, it may be more critical than ever, but for different reasons.

Our current thinking is that while the traditional "organic ranking signal" of speed isn't overtly important today, the new breed of LLM-driven answer engines will likely rank for speed AND context as they crawl and process data far more frequently and deeply than Google's legacy infrastructure currently does.

Why LLMs Care About Speed

LLMs operate under different economic and computational constraints than a traditional search engine indexer. They require vast amounts of data to provide real-time, synthesized answers. A slower site imposes a greater cost on the LLM provider in terms of:

1. Computational Overhead: More time spent waiting for a page to load translates to higher operational costs (CPU/GPU time) for the LLM infrastructure.

2. Crawl Efficiency and Freshness: Faster sites allow LLM agents to crawl a much larger volume of the web more frequently, ensuring the freshness and comprehensiveness of their answers—a core competitive advantage in AEO.

3. User Experience (Indirect): Even if the LLM provides the answer directly, it is scraping the source page behind the scenes. If a user clicks through to the source for verification, a slow load time still degrades the overall ecosystem experience.

The speed of a page load is clearly going to remain important, but the idea that a green "google score" is the ultimate aim of your tech team seems a little outdated.

Redefining "Good Enough": The Amber Standard

Given the economic shift in crawl priorities, the focus needs to move from chasing diminishing returns on site speed (i.e., optimizing the last 5% for a perfect green score) to achieving functional excellence that facilitates efficient machine consumption.

Here at Spike towers we reckon a decent Amber score is "fine" for most. This is not a license for technical debt; it is a pragmatic assessment of resource allocation. The resources previously dedicated to shaving milliseconds off the Largest Contentful Paint (LCP) could be better deployed to:

1. Improving Content Semantics and Structure: Making content easily parseable by LLMs (e.g., clear headings, structured data, concise summaries).

2. Ensuring Data Freshness and Accuracy: The true value signal for an LLM is up-to-date, verifiable information.

3. Optimizing for Context and Authority: Building trust signals that LLMs can clearly identify, such as clear authorship and references.

This shift represents a move from metric-chasing to utility-optimization. The goal is no longer pleasing a Google algorithm for organic placement; it's about facilitating efficient, cost-effective data ingestion by the dominant AI models that are shaping the future of search.

Let us know what you think!!

References and Further Reading