What is a Markov Chain?
A Markov chain is a mathematical system that transitions from one state to another based on certain probabilities. Its key feature is the “memoryless” property — the next state depends only on the current state, not on how you got there.
For example:
A Markov chain is like walking through a series of rooms where each door leads to another room with certain probabilities. The choice of the next room depends only on the room you’re currently in, not on the path you took to get there. (This is an oversimplified example, read on as I explain more about this idea!)
How Markov Chains Shaped the Early Internet & SEO
Source: Google Ranking Systems Guide
When Google first revolutionized search in the late 1990s, it used PageRank, which was essentially based on a Markov chain.
It means that the web was modeled as a series of “states” and “transitions.” Each webpage represented a state, while the links connecting one page to another acted as the transitions between states.
How the Markov Chain model worked!
In the classic Markov chain model (like the one Google’s PageRank used for the web), the probabilities don’t come from a person’s preferences. Instead, they’re fixed by the system’s structure. For example:
- If you’re on a webpage (a “state”), and that page has 10 links, a very simple Markov chain would assign each link a 10% probability of being clicked.
- If the page has 3 links, then maybe each has a 33% probability.
- More sophisticated models might “weight” the links differently (e.g., higher probability if a link is more prominent on the page).
So in this sense, the 70% or 30% probability is determined by how the system is modeled — not by the actual choices of individual users.
Over time, the probability of them landing on any given page could be calculated. Pages that were more frequently “landed on” in this model were considered more important, and therefore ranked higher in search results.
Pages with more inbound links (and from highly ranked pages) had higher probabilities, meaning higher rankings in search results.
👉 For early SEO, this meant that backlinks were the golden ticket. The more high-quality links a site had, the higher it would rank.
How Things Changed Today
The Markov chain is one of the main reasons why backlinks were included as one of the key criteria in traditional SEO techniques as well as its central role in EAT (Expertise, Authority and Trustworthiness).
However, while that was significant in the past, search has grown way beyond pure Markov-chain-based PageRank.
With the introduction of AI (and all sorts of ways to identify search patterns and actual intent), that is no longer the only focus! Some of the modern criteria’s include:
- Semantic Search: Today, Google looks at the meaning of queries, not just link patterns.
- User Experience: Mobile-friendliness, page speed, and site usability are now ranking factors.
- AI & Personalization: Search results are influenced by your location, intent, and even past behavior.
- Content Quality: Algorithms like Google’s Helpful Content Updates prioritize articles that genuinely help people, not just link-heavy ones.
So while the Markov chain still underlies link structure models, modern SEO is multi-dimensional — blending probability, semantics, and AI-driven relevance.
Does AI Use Markov Chains?
Yes, but with nuance:
Early AI (like old text generators or chatbots) often used Markov models to predict the next word. But they felt repetitive and limited.
As technology continues to advance, modern AI (like GPT) goes far beyond. It uses transformer models trained on massive datasets.
These models also predict the “next token” — but instead of a simple Markov chain, they consider long-range context across entire sentences, paragraphs, or even documents.
