If you've been paying attention to the SEO world lately, you've probably noticed a shift in the conversation. People are starting to talk about "AEO" (Answer Engine Optimization) or "GEO" (Generative Engine Optimization) instead of just SEO. The logic goes: why optimize for Google when people are increasingly getting their answers from ChatGPT, Perplexity, and other AI tools?
Traditional search is declining for many queries. AI gives you direct answers. The future is clearly moving away from the old model of scrolling through a page of search results. I heard a podcast with SEO expert Mark Williams Cook and he said "the link graph is rotting." I had no idea what that meant or why I should care. So I went down a rabbit hole to understand what a link graph is, why it might be rotting, and whether all this excitement about AEO is actually missing something important.
What Even Is a Link Graph?
The link graph is the actual network of all the links between websites on the internet - essentially a map of who links to whom across the entire web. PageRank is the algorithm that Google created to analyze that link graph and assign authority scores to pages. When Mark talks about the link graph "rotting," he means the underlying network itself is degrading because fewer links are being created and fewer user signals are being generated. This matters because PageRank and other ranking algorithms depend on having good link graph data to work with. If the data degrades, even the smartest algorithms can't produce reliable results.
Think of it this way: the link graph is the raw data, and PageRank is one method of interpreting that data to determine which sites are most authoritative.
Okay, But Why Would It Be Rotting?
The traditional web flow:
You search for something
You click through results
You read websites
You link to helpful sources
You return to search if unsatisfied
This journey generated invaluable data: clicks, time on page, bounce rates, natural linking patterns, and user satisfaction signals.
The emerging AI search flow:
You ask AI a question
You get a synthesized answer
You're done
No website visits. No links created. No journey data.
But AI Search actually needs an accurate Link Graph?
You probably know that ChatGPT has a knowledge cutoff. When you ask it something recent, it searches Bing in the background. Perplexity does the same thing, but you can see it searching in real-time. Google's AI Overviews use Google's own search results.
Here's the important part: Those search results are ranked using the link graph. When ChatGPT searches Bing for current information, Bing returns results ranked partly based on which sites have accumulated authority through links. The AI trusts those top results more because the link graph has vouched for them.
The problem: By giving you a direct answer, AI tools prevent you from visiting those websites and potentially linking to them. The AI is consuming the link graph's authority signals while simultaneously preventing new signals from being created.
Here is another thing that is changing
Google doesn't just use links. It watches:
What you scroll past
What you hover on
What you click
When you bounce back to search
When AI agents do the searching instead of humans, these signals disappear. An AI might check 30 pages per second with zero indication of human preference or satisfaction. It's not hovering thoughtfully over results or bouncing back when disappointed. It's just grabbing data.
So all this talk about AEO replacing SEO? AI answer engines are still fundamentally dependent on the same infrastructure that traditional search engines built. We're not moving to a new paradigm. We are adding a layer on top of the old one. But are we potentially breaking what makes it work?
So the question is - are we watching a critical system slowly fail, or is this something that will sort itself out?