There are no AI optimization experts. AI SEO, GEO, LEO, AIO, whatever you wanna call it. No one's an expert, not yet.
This industry is brand new and it's changing rapidly. That's not stopping people from pretending like they know the best way to rank in these large language models.
I've spoken with three people just this week who have received some big pitch from some AI optimization company that's asking for a ton of money for some secret system. Here's the thing. There are no secrets in 2025.
Yes, people are constantly finding new ways to rank in large language models, but anyone who's paying attention to this is gonna know about that method within a week.
Now to be clear, I currently call my company an AI optimization agency is that is my primary focus. I'm not against people using that term. I just think we need to be honest about what we actually know.
I think that everything people are calling AI optimization or Geo right now can be broken down into four categories. The first category is just SEO. These are things like keyword optimization and backlinks building.
And yes, it's true anything that helps you with SEO is also gonna help you rank in these large language models because they are using search engines under the hood.
However, because of this, a lot of SEO's are saying it's exactly the same thing, and wouldn't that be great for them? You don't have to change anything. Just keep doing SEO.
It's not true.
Second category is something I'm calling AI focused SEO. These are things that are generally still helpful for SEO, but help even more in ranking in ChatGPT and other large language models. These are things like adding copious amounts of information about your company to your website, often in FAQ's, and making sure your brand is mentioned in all the places that ChatGPT and other large language models are looking, even if there's no link.
The third category I call AI hacks. These are spamy tactics that happen to work right now, but probably won't work long term and could get you penalized. These are things like hiding in prompt injections in your website copy, listicle spam and low quality press release spam.
The fourth category is just pure speculation. These are things that might work, but there just hasn't been enough time or data to say for sure. Of course, that's not stopping people from acting as if they know it's working.
txt file, or adding marked down versions of your webpage copy. Just for large language models.
I've seen hundreds of ideas that fall into this speculation bucket. I would say it makes up about 60% of the ideas I'm seeing hawked. People have no problem passing it off as knowledge, even though there's practically no data to support it.
Now, there is actually a fifth category, and it might be the most important. And those are all the ideas that actually work that no one's even thought of yet. And we haven't thought of them either because we don't understand the technology well enough yet or because the technology that will ultimately enable them doesn't exist yet.
I think that's the most important point to keep in mind. The things that are gonna be working well six months or a year from now are probably gonna be very different than the things we're doing right now. We need more data, we need more time, and it's going to continue evolving.
So if you're looking to hire someone to help you with this, you should not be looking for someone with all the answers. In fact, you should be running away from those people. Look for someone who's learning everyday and changing their systems in real time.
Things are changing faster than ever, and so we have to too.
