G
GEO Toolbox

How AI Search Works

Reasoning Model

Also: reasoning models, thinking model, reasoning LLM

A reasoning model is a large language model trained to work through a problem step by step before giving its final answer, rather than responding in one pass. Examples include OpenAI's o-series, DeepSeek-R1, Gemini Deep Think, and Claude's extended thinking. The extra 'thinking' improves accuracy on hard math, coding, and multi-step logic, at the cost of more time and tokens.

Updated

A reasoning model spends compute thinking before it answers, often generating a hidden chain of intermediate steps you do not see. That is why these models are slower and pricier, and why labs reserve them for the hard tier (Pro, Deep Think, Heavy) while a faster model handles everyday questions.

For visibility, the relevant detail is what happens when a reasoning model is paired with web search, which the big assistants increasingly do: the extra deliberation means it can spend more effort planning its searches and comparing what it finds, though which sources it ultimately trusts still depends on the assistant's retrieval system. Clear, consistent, well-structured facts about your brand are easier for a step-by-step model to verify and cite, and harder for it to get wrong, than vague or contradictory information scattered across the web.