Knowledge Cutoff
In one line
Knowledge cutoff is the most recent date covered by an LLM's training data — the reason an AI may not know your current pricing, policy changes or newest products.
Going deeper
Knowledge cutoff is the last date covered by a model's pretraining data. Model cards typically state something like 'as of July 2024'; anything after that is either unknown or hallucinated. Each new model release pushes the cutoff forward, but there is always a several-month-to-a-year lag.
The marketing implication is blunt: if you changed pricing yesterday or launched a new SKU last month, the AI may keep quoting last year's information for a long time. Users asking ChatGPT for 'XX brand pricing' and getting a six-month-old number is a common scenario.
Two ways to close the gap. First, make sure surfaces with live retrieval (ChatGPT Search, Perplexity, AI Overviews) can find your current pages. Second, get accurate, up-to-date facts into the sources that get re-trained on — Wikipedia, press, your own canonical pages.
Related terms
Pretraining
Pretraining is the initial stage where an LLM is trained on huge amounts of text to learn general language capability — the step where the model absorbs most of its 'world knowledge'.
LLMRAG
RAG (Retrieval-Augmented Generation) lets an LLM fetch external documents at answer time and ground its response in them — the technique behind ChatGPT Search, Perplexity and most AI search products.
LLMLLM
A large language model (LLM) is a neural network trained on massive text corpora to understand and generate human language — the engine behind ChatGPT, Claude, Gemini and similar products.
GEO·AEOPerplexity
Perplexity is an answer engine that turns search results into a single cited answer, attaching a numbered source to every sentence — making it a common reference surface for measuring GEO performance.
GEO·AEOAI Overview
Google AI Overviews is the AI-generated summary that appears above the standard results in Google Search — one of the most prominent zero-click surfaces today.