Skip to content

So you've fed ChatGPT* your prompt...

What happens next? Unlike traditional search engines, LLMs work with full context, not just keywords.

By Chris Pedersen, Technology Director

01

First, the LLM agent decides what it already knows

It draws on its model knowledge, i.e. info absorbed during training.

2a

The question is generic

It may answer straight away.

2b

The question is more specific

It searches websites, documents, data and connected systems for content, running multiple, detailed searches.

3a

Information is vague, jargonistic, unverified or appears irrelevant

The agent discards it and moves on.

3b

Information is clear, consistent, credible and relevant to the user's problem

The agent uses it as part of its response.

04

The agent provides its final answer

This comprises model knowledge + user context + retrieved content, built from the information the agent considers most useful for the question.

REMEMBER:

Humans search. Agents investigate.

Agents don't just ask “Who ranks highest?”, they ask “Who has the clearest, most relevant and verifiable answer?”

Appearing in agentic search relies on your brand showing up via information that is so structured, accessible, consistent and credible that an agent can’t help but trust and therefore recommend it.

And happily, we can show you how to do that.

 

*Or Gemini, Claude, etc.

The Flip

A new way of marketing so people get the emotional connection and agents get the evidence they need - all in one go.
Ready to flip your thinking?
the flip

Have a problem to solve?
Let’s chat!