LLM (Large Language Model)
Quick answer
An LLM (large language model) is the AI system behind products such as ChatGPT, Claude, Gemini, Perplexity, Grok and DeepSeek that reads a question and generates an answer word by word, drawing on patterns learned from huge amounts of text plus, for some products, live web results it retrieves at the moment someone asks.
Why it matters
A brand's buyers increasingly ask an LLM-based product to compare tools before they open a search engine or read a comparison page themselves, so how an LLM describes a brand now shapes a sale before a human visitor ever shows up in analytics. Because an LLM predicts likely answers from training data and, for several products, pages it retrieves at answer time, the same question can get a different answer on different engines, and the gap between engines usually traces back to which sources each one actually reads rather than anything a brand did wrong. This is true even for well-established brands with strong classic search rankings, since an LLM's answer draws on a different, often narrower set of sources than a search results page does.
How to measure it
A brand cannot inspect an LLM's internal weights, but it can measure how one treats a brand's name. Run the buying questions a prospect would actually ask through each AI-based product a brand's buyers use, record whether the brand is named, how accurately, and which competitors appear beside it, then repeat the same questions on a schedule. Comparing results across engines and over time shows whether a fix, such as a new comparison page, a corrected doc, or a new review, changed the answer, rather than guessing from a single check.
Example
A support-software team notices one AI-based product describes their pricing accurately while a second gets it wrong, naming a plan the vendor stopped selling a while back. Checking which pages each product seems to be reading from shows the accurate one cites the vendor's own pricing page, while the inaccurate one leans on an old review-site writeup that was never updated. The fix is getting the review site corrected or outranked, not touching the pricing page that was already right.
Frequently asked questions
Is an LLM the same product as ChatGPT?
No. ChatGPT is one product built on an LLM. Claude, Gemini, Perplexity, Grok and DeepSeek are separate products built on their own LLMs, and each can describe the same brand differently.
Why does an LLM-based product change its answer over time?
Several of these products retrieve fresh web pages at the moment someone asks, so a changed page, a new review, or a new community thread can shift the answer even when the underlying model itself hasn't changed.
Can a brand influence what an LLM says about it?
Not directly, but a brand can influence the pages, docs and third-party sites these products read from, which is what AI-visibility tracking and fixes actually target.
Does a bigger LLM always describe a brand more accurately?
Not necessarily. A larger model can still repeat an outdated or wrong fact if that is what the pages it was trained on or retrieves say, so accuracy depends more on what the model reads than on its size.
Related terms
Related
MarketHQ tracks brand mentions across communities, news, blogs, social and AI answers, and turns them into gap analysis and action plans.