Skip to Main Content
SEOSeb

Article · AI search · May 2025

Getting Your Brand Mentioned by LLMs Starts With Association

LLMs predict the next word based on patterns in their training data. What that means for your brand, and what the temperature parameter tells us about it.

By Sebastien Edgar · 3 min read

I recently discussed the inner workings of LLMs like ChatGPT. At their core, these models predict the next word (or token) based on patterns in their training data, essentially generating a “reasonable continuation of the input text.”

Image 1 from the original LinkedIn post

What does this mean for your brand?

To increase the chances of your brand being mentioned in LLM responses, associate it closely with key terms or concepts in relevant content. For example, if your brand frequently appears alongside a specific term in high-quality training data (especially Wikipedia!), an LLM is more likely to reference it when prompted about that topic.

As has been mentioned multiple times on LinkedIn: high quality citations are the new backlinks. Since LLMs are prediction machines, when building out a response they choose terms with the highest association (i.e. probability) with the previous words and content.

Image 2 from the original LinkedIn post

The temperature parameter

You can see this in action with the temperature parameter in ChatGPT. By appending “(temperature=0.3)” to your prompt (any value between 0 and 1), you can control the randomness – i.e. the word probability – of the response:

  • 0: more predictable, sticking to high-probability outputs.
  • 0.7: ChatGPT’s default, balancing creativity and coherence.
  • 1: highly creative, but responses can be a bit whacky.
Image 3 from the original LinkedIn post

Of course, getting your brand or content mentioned in LLM responses is more complicated than that. Tactical factors related to content structure (statistics, citations, quotes, HTML tables, Flesch-Kincaid readability, etc.) appear to have a massive impact on your content being mentioned.

Image 4 from the original LinkedIn post

Sources

Highly recommended to read or watch:

Stephen Wolfram on how ChatGPT works

The ChatGPT temperature parameter

How ChatGPT works, by Andrej Karpathy