Skip to content
Browse courses

Learning Paradigms

Reinforcement Learning

In shortA machine learning approach where an agent learns by trial and error, taking actions and adjusting its behaviour based on rewards or penalties.

Updated 24 Oct 2026· 2 min read

Reinforcement learning (RL) is a branch of machine learning in which an agent learns how to act through trial and error. It takes actions in an environment, receives rewards or penalties, and gradually learns a strategy (a policy) that maximises its total reward. It sits alongside supervised and unsupervised learning as the third main learning paradigm.

How it works

  • Agent and environment: the agent observes the current state (for example a game board or a user’s context).
  • Actions: it chooses an action, such as a move or a recommendation.
  • Reward: the environment returns feedback, which may be delayed (a game is won only at the end).
  • Policy update: the agent updates its strategy so that actions leading to higher long-term reward become more likely.
  • Exploration and exploitation: it balances trying new actions with repeating ones that already work.

Unlike supervised learning, there is no labelled “right answer” for each example, only a reward signal. Unlike unsupervised learning, the goal is not to find structure in data but to act well. Combined with deep neural networks it becomes deep RL, behind systems such as AlphaGo.

How it’s used in SEO and marketing

Few SEOs will ever train an RL model. It still helps to know where it appears:

  • LLMs: reinforcement learning from human feedback (RLHF) and related methods tune chatbots to follow instructions and give preferred answers.
  • Autocomplete and predictive text: MLforSEO’s autocomplete lessons describe how these systems improve from user interactions and feedback loops.
  • Advertising and personalisation: bid strategies, recommendation systems and multi-armed bandit testing (a simple form of RL) allocate traffic towards what performs.
  • Agents: AI agents that browse and complete tasks are often trained with RL, which is part of why agentic SEO cares about machine-readable, unambiguous information.

Example

A multi-armed bandit test on a landing page shows three headline variants. Instead of splitting traffic evenly for weeks, the algorithm shifts more visitors to whichever headline is converting better while still occasionally showing the others. It learns from the reward (conversions) as it goes, which is the core idea of reinforcement learning.

Limitations and things to watch

  • It needs a well-defined reward; a badly chosen reward is optimised literally (for example clicks at the expense of satisfaction).
  • It usually needs many interactions, so it suits high-traffic systems, not one-off analyses.
  • MLforSEO’s guidance is that knowing when RL applies is mostly about knowing when it does not: most SEO tasks are better served by classification, clustering or entity extraction.

Learn it on MLforSEO

Where Reinforcement Learning shows up on MLforSEO

Learn it in a course

Read about it on the blog

Sources and further reading

Other terms from Learning Paradigms