AEO / GEO - what "optimizing for AI search" actually means
The acronyms arrived before the discipline did: AEO (answer engine optimization), GEO (generative engine optimization), AIO, LLMO. Vendors are selling all of them. So let's do the honest sorting: what's genuinely new practice, what's classic SEO wearing a new badge, and how to tell either from snake oil.
What's genuinely new
Four things actually differ from classic SEO. The unit of competition: you're competing for mentions in composed answers, not positions on a results page - fewer winners per query, and being the named recommendation matters more than being one of ten links. The reader: pages are parsed for extractable facts (last lesson), which raises the value of explicit, structured, self-contained information beyond what classic ranking ever demanded. The query shape: people ask assistants long, conversational, multi-constraint questions ("linen shirt, under €150, ships to Denmark, sustainable") that nobody would type into Google - meaning your content needs to answer constraint-combinations, not just keywords. The fanout lesson goes deep here. The measurement: no Search Console for ChatGPT exists; visibility has to be sampled and inferred (a dedicated lesson later this phase).
What's classic SEO, rebadged
Most of the rest. Crawlability, indexation, ranking - still the retrieval gate. Structured data - a 2015 recommendation that finally became urgent. Content that actually answers questions - Phase 4's content strategy, verbatim. Brand mentions across the web - public relations with new packaging. When someone sells you "GEO services," the deliverables are usually this list. That's not a scam exactly - the work is real - but you should know you're buying SEO fundamentals, and price accordingly.
How to spot the snake oil
Red flags, from a market currently full of them: guaranteed placement in AI answers (nobody controls this; the systems change monthly), proprietary "AI visibility scores" with no stated methodology, tactics that boil down to keyword-stuffing for robots ("add an FAQ that repeats your brand name eleven times"), and anything premised on tricking the model rather than informing it. The systems are trained to synthesize reliable information; the durable strategy is being reliable information. Tricks that work this quarter get patched next quarter - ask anyone who did SEO in 2011.
The practical posture
For us, the sensible position is neither panic nor dismissal: AI assistants are a real and growing discovery channel, the optimization surface genuinely overlaps 80% with work you should do anyway, and the 20% that's new - fanout-shaped content, extraction-friendly pages, visibility monitoring - is what the rest of this phase covers concretely. Do the fundamentals because they pay on every channel; add the new 20% because it's cheap while your competitors are still arguing about acronyms.