Parametric memory
Also called Training knowledge, Model memory
Parametric memory is the knowledge a language model holds in its trained parameters, as distinct from information the model retrieves when it answers.
Why it matters
A model answering from parametric memory relies on what it learned in training, so a fact that changed after training can be missing or out of date unless the system retrieves a page that states it.
Example (illustrative)
An assistant without web access describes a company's old pricing because the new pricing was published after the model was trained.
What to check
Ask an assistant about your company with web search turned off and again with web search turned on, then compare the two answers.
What this does not tell you
An answer rarely shows which parts came from training and which came from retrieval, so the comparison is an estimate.
Sources and scope
Documented concept · Sources checked
Find Out How AI Sees Your Brand
AI is already shaping how buyers discover and evaluate your business. Start with a free AEO audit to see how ChatGPT, Perplexity, and Gemini represent your brand today.
Get Your Free AI Visibility Check
HubSpot Solutions Partner. Zero fluff.
