Training for AI search and training for Google are two different races with almost the same training. The industry keeps arguing about what to call the new race. That is the wrong argument.
Think about a 100m sprinter and a 60m sprinter. Different disciplines, maybe even different winners. But properly training for those two is almost the same. Good SEO practice is what it takes to win in AI search as well, and the fight over whether to call it GEO, AEO, or something else is energy spent on the label instead of the training.
I am not the only one making this point. Jono Alderson, a technical SEO consultant who came on the No Hacks podcast, argued that SEO vs GEO is the wrong question and that AI is not a new discipline. Mordy Oberstein, in conversation with Brent Csutoras, put it as "SEO isn't dead, strategy is dead." Ross Hudgens warned against branding yourself as an "SEO / GEO writer." Three practitioners, within days of each other, arriving at the same place from three directions. When that happens, it usually means the naming fight has become a way to avoid a simpler and less flattering conversation.
GET WEEKLY WEB STRATEGY TIPS FOR THE AI AGE
Practical strategies for making your website work for AI agents and the humans using it. Podcast episodes, articles, videos. Plus exclusive tools, free for subscribers. No spam.
The training that wins both races is a decade old
Everything Google has been suggesting, subtly and less subtly, for more than a decade is the training, and it comes down to common sense and no hacks. Be clear. Have a healthy website and a healthy off-site presence. Let your business drive and power your SEO, not the other way around.
None of that was written for AI. It is the unglamorous advice that has been sitting in plain sight for years, the advice that does not sell a course or a new acronym, so most of the industry walked past it in favor of whatever tactic was ranking at the time.
A model answering a question about your product is doing a version of what a search engine has always tried to do. It is trying to find the clearest, most consistent, most trustworthy account of who you are and what you sell, and repeat it. A clear website with a coherent off-site presence made that easy for Google, and it makes it easy for a language model. A website built on tricks made it hard for Google, and it makes it hard for a model. The mechanism changed. The thing being rewarded did not.
What is actually new, and who it feels new to
One thing is genuinely new: more focus on off-site, and not in the sense of trying to build backlinks. The non-glamorous activities that never really went mainstream, entity optimization and consistency across every place your brand shows up, matter more now than they used to, because a model assembles its picture of you from everywhere it can read, not from your homepage alone.
Whether that feels new to you depends entirely on what your SEO was.
If your SEO was content marketing, or publishing shallow content so it temporarily ranks, then yes, a lot of this is new, and it is going to keep feeling new and difficult. The tactics that produced temporary rankings do not produce a trustworthy, consistent entity that a model can rely on. If your SEO was focused on fundamentals, no, this is not new. You have been doing entity work and consistency work the whole time, even if nobody called it that.
The people who feel the ground shifting hardest are the ones who skipped the fundamentals the first time. The naming argument is more comfortable than admitting that.
The one move: check what the machine already believes about you
So what do you actually do about it? Check whether what you are giving the LLMs, and Google's AI Overviews and AI Mode, would make them hallucinate or pass the wrong information to the user. Every key page on your website, and everything off-site as well. Then prompt the LLMs about your brand and your products. See what they know and where they get their information. Reverse engineer that, instead of focusing on "best X for Y" prompts.
That last distinction is the whole move. Most people testing AI visibility type the category prompt, "best CRM for small teams," and check whether they show up. That tells you where you rank on one phrasing today. Asking the model what it knows about you, and watching which sources it reaches for, tells you what it actually believes and where that belief comes from. One is a ranking check. The other is a map of the exact pages and profiles you need to fix. The map is worth more.
And this connects up to something bigger than visibility. The same machine that reads your website to answer a question is the one that will increasingly act on it and buy through it. Getting it to hold an accurate account of you is the groundwork for everything else an agent will try to do with your business. Citations are the smallest part of it.
AI search only feels like a new sport to the people who were never really training. Everyone else kept doing the work.