Episode 12: What Does Machine Learning Do in AE Today - and What Might It Do Tomorrow?
Episode 12: What Does Machine Learning Do in AE Today - and What Might It Do Tomorrow?
Nobody in 2018 was willing to say precisely how much machine learning was affecting search results. That reticence was itself revealing. The honest answer was that nobody outside Google, Amazon, or Microsoft knew with confidence - and even inside those organisations, the answer was shifting faster than anyone could document.
Episode 12 of the #SEOisAEO series was built around that uncertainty. Patrick Stox, Dawn Anderson, and Jan Willem-Bobbink joined Jason Barnard to do what few were willing to do publicly: stick their necks out about how Google was applying ML to answer engine results in 2018, and where that application was heading.
The conversation covered three distinct ML applications that were already active at the time. The first was natural language processing - how ML was enabling search engines to interpret the meaning of a query rather than matching its keywords, directly enabling the kind of intent-based retrieval that AEO depends on. The second was Knowledge Graph construction - how ML was being used to extract entity relationships from unstructured text at scale, populating the graph that answer engines draw on when selecting answers. The third was spam and quality analysis - how ML was being used to evaluate the credibility of sources, identifying patterns of manipulation that rule-based systems had missed.
What made the episode valuable then, and makes it worth revisiting now, is that the three guests were willing to hypothesise about where each application was heading. The trajectory they described - ML moving from query interpretation toward full semantic understanding, from Knowledge Graph assistance toward Knowledge Graph construction, from spam detection toward holistic quality evaluation - is the trajectory that played out. The generative AI platforms that emerged after 2022 are the downstream result of exactly the ML capability expansion the episode was projecting.
For practitioners, the implication was the same in 2018 as it is in 2026: optimise for the machine’s ability to understand you, not for the machine’s current rules. Rules change. Understanding compounds.
Published by: Semrush. Host: Jason Barnard. Guests: Patrick Stox, Dawn Anderson, Jan Willem-Bobbink. November 20, 2018
