How different is SEO from GEO? How differently should we treat Google versus LLMs?
There is a lot more overlap than separation. The fundamentals still matter. Crawlability, clear site structure, meaningful content, consistency across pages, etc. all influence both how search engines rank content and how AI systems interpret and cite your content.
At the same time, there are some very real differences. And if you ignore those, you can create problems in one system while trying to optimize for the other.
A good example is content created specifically for AI citations. In some cases, it can make sense to create pages designed to surface in AI responses. For example, you might have pages that frame the same topic in slightly different ways, tailored to different contexts or use cases. That can help an LLM better understand how your content applies across scenarios.
From a traditional SEO perspective, these practices can start to look like duplicate or low-value content. If those pages don’t meet Google’s quality guidelines, you need to protect traditional SEO while optimizing for AI search. You can block otherwise low-quality or duplicate pages from crawling entirely via robots.txt, or allow crawling but add a noindex directive so they do not end up in Google’s index. Either way, you are protecting the rest of your site from potential duplication or thin content issues, while still making that content available where it could be useful.
There is a similar thing happening with llms.txt files and other AI explanation pages. These often end up as dense lists of entities, URLs, and keyword-heavy summaries. They might be useful for machines, but potentially problematic for Google. Left unchecked, they can look overly optimized or even spam-like. In those cases, adding a noindex is often the right option to make sure that these pages do not harm your regular SEO efforts.
It works the other direction as well. Some controls that work perfectly for Google simply do not translate to AI systems.
Canonical tags are a good example. Google uses them to consolidate signals and manage duplication. Many AI crawlers, at least today, do not appear to use canonical tags. So if you are relying on canonical tags to manage parameterized URLs or duplicate variants, that may be sufficient for Google, but those duplicate URLs can still be crawled, interpreted, and even cited in AI responses.
In that scenario, it may be more effective to control access directly. Prevent certain URL patterns from being crawled by AI bots so they never see the duplicate content in the first place.
The bottom line here is to remember that we are now optimizing for two overlapping systems with different rules. There is a lot in common, but there a lot of nuanced differences that are important to remember.