Help Centre article format structure | Community
Skip to main content

Any thoughts, advice appreciated!
Using Intercom for our Help Center and wondering what is the best format structure in terms of AI and LLM’s. We currently start each article with a “summary” and “why its helpful” in a callout box but am thinking this style, particularly the colourful callout box does not add a lot of value to humans or AI responses and it looks a little messy. Are H1, 2, 3 headers better than callout boxes?  When would you use callout boxes? 
Thanks !
 

Hi ​@juliez, Intercom has a cool article just for what you’re looking for:
https://www.intercom.com/help/en/articles/10560969-fin-guidance-best-practices

Pro tip: You can also use these guidelines and other best practices you might find to create a custom GPT (or equivalent on any LLM) to help you format your articles per those practices. 
 

Thanks Luke for the quick response!  I read the article and its great for content tips but I was also looking for some tips on the best use of the formatting options that Intercom gives. eg Callout bubble, Headers etc. I have been told that the summaries inside callout bubble are particularly good in regards to reference and quality of AI responses but I am not so sure. Looks like Intercom does not use them this way either🙂

That said I was also curious in the article that it mentioned British English is preferred. Doesn’t that depend on where your main market is? (I will out myself here and say I prefer British English😉) 

In my experience, the quality of answers mainly depends on the quality of the content itself and logical structure formatting. Fin is based on frontier LLMs and is more and capable in that regard, and I found no relevant issues with any of the supported languages. I’d say the ideal formatting also depends on your target audience, but any general best industry practice you find out there should be fine, and a good LLM prompt can really help you in that regard. 

I don’t think there’s one right answer, and the best is probably to have a continuous maintenance - improvement - iteration feedback loop cycle in your processes and test, test, test and see what works best. 

 

Terms & Conditions