Last week, nerds from across America and Europe converged in Hamburg for the third-annual Tech SEO Summit. I've spoken at my share of conferences over the years. This was the first that encouraged speakers to tell audiences about the riddles for which they didn't have answers. Curiosity over bravado was the mood. Rather than a multiple day rapid fire, TSS is single day with six speakers given 45 minutes to deep dive into their topic, plus another 15 for questions and discussion. The venue was arthouse cinema. The food was… a surprising amount of fish and mustard sauce?

I had the pleasure of kicking off the morning with a session on the new search conversion funnel in an AI enabled world. (My day/trip/month was made thanks to a surprise video introduction from the inimitable Roxana Stingu!) I talked mechanics and business impact– with a dash of furby humor. Yes, AI is still less than 1% of all traffic but it certainly has changed some things. Users are searching more unique queries and asking more questions. Questions are more likely to trigger AIOs. Arguably, the largest AI surface out there are the increasingly prominent AI features in Google SERPs. I offered an analysis trick to let you see where it's eating your lunch.

Since the line up was arranged in order of most jetlag, Paul Calvano then took the stage to talk fonts and web performance. Are you shipping a megabyte of fonts but only showing two characters from the entire library? Etsy was. Paul showed us how to diagnose and solve the issue! He's a fellow Web Almanac alum and used the experience to interweave freshly pulled stats from the HTTP Archive into the presentation.

Next up, Will Kennard gave those die hard technical masochists in the form of caching as a competitive advantage. At the speaker's dinner, Will shared that his deck was an astonishing 226 slides. By the time he took stage, it was a svelte 160. Honestly? Give me the director's cut. You can supplement the deck with his guide to caching for Next.js.

After lunch, Martin Splitt took us down the HTTP pipeline like a magnificent dev rel Mario. We got a deeper look into Google's egress and DNS systems. Fun fact: Googlebot uses the crawling infrastructure as its networking stack. It uses http keep alive– but not pipelining. Full H/2? Yes. H/2 push? Not today, satan. No websockets or service workers while we're at it.

I can think of few things as daunting as being asked to fill a 45 minute timeslot for an ill speaker. René Dhemant was up to the challenge with a session on architecting content for AI answer engines. The RAG framework is an architecture that combines two separate systems:The RAG pipeline (Retrieval & Augmentation) and the generative LLM (Generation). René hypothesizes that since the retrieval and weighting of data chunks before their transfer to the LLM are deterministic, they are subject to clear mathematical rules. Consequently, this phase can be systematically measured and optimized. He even has a Github ready for you to play along.

Rounding out the event, Estela Franco got practical with catching technical SEO regressions and rendering disasters in the CI/CD pipeline. Tired of devs missing or breaking critical SEO elements? You're going to want to check out her "SEO as code" framework. Estela also shared her open source project on GitHub.

Cheers to the venues where nerds thrive!