The five support KB article types define the different jobs customer-facing content needs to perform.
The most common reason AI retrieves the wrong answer is not bad writing. It is an article trying to do too many things at once.
AI agents retrieve the content they believe is most relevant to a customer’s question or intent. If an article contains multiple intents, retrieval becomes less reliable. The AI may retrieve the wrong section, miss the actual answer, or assemble a response from pieces that do not belong together.
The fix is simple in principle and harder in practice: one article, one intent.
An intent is something a customer is trying to accomplish, understand, fix, or look up. When each article has one clear job, customers can follow the content more easily, support teams can maintain it more consistently, and AI systems can retrieve the right information more reliably.
This framework uses five article types:
Each article type answers the question:
Before choosing an article type, decide how the information should reach the customer.
That delivery decision determines whether the content should be public or private. Once you know how the information should reach the customer, choose the article type based on the job the content needs to do.
For the delivery decision, see:
A task article helps someone do something.
The user intent is “how do I.” The article is action-oriented, step-by-step, and built around a single outcome. A customer should be able to complete the task without leaving the page or reading anything else.
Examples:
- Set up two-factor authentication for your account
- Connect your Stripe account to your workspace
- Add an admin user to your account
A concept article helps someone understand something or make a decision.
The user intent is “what is,” “what happens if,” or “what’s the difference between.” There are no steps. The article explains behavior, clarifies a feature, or compares two options so a customer can decide which one applies to them.
Examples:
- What is a webhook and when should you use one?
- What happens when you downgrade to the free plan?
- User roles vs admin roles: what’s the difference?
A guide helps someone achieve a broader outcome that spans multiple features or tasks.
The user intent is “how do I accomplish X” where X requires more than one step and touches more than one part of the product. A guide does not contain those detailed steps itself. It organizes them and links to the task articles that do.
Examples:
- Set up your account for your first team
- Configure billing and invoicing for your workspace
- Migrate customer data from your previous platform
A troubleshooting article helps someone fix something that is not working.
The user intent is “why isn’t this working” or “how do I fix this.” The article leads with the problem, covers the most likely causes, and walks through solutions. It is diagnostic first, instructional second.
Examples:
- Why isn’t my Salesforce integration syncing contacts?
- Emails sent from my account are not being delivered
- Can’t log in after resetting your password
A reference article helps someone look something up.
The user intent is “what are the limits,” “what are my options,” or “what are the specs.” There is no narrative, no steps, and no explanation beyond what is needed to make the information usable. Structure and precision matter more than anything else here.
Examples:
- Feature availability by plan
- API authentication requirements
- Supported file types for file uploads
The five article types are organized around customer intent. Start by asking:
Complete a specific action
Understand something or make a decision
Diagnose and fix a problem
If the answer points to one specific action, concept, problem, or piece of information, you have probably found your type.
If you find yourself writing “this article covers X and also explains Y,” you probably have more than one article. Split them.
The discipline of choosing a type before you write is what keeps a knowledge base usable at scale. It is also what makes AI retrieval more reliable. An AI agent cannot always sort out mixed intent after the article is published. That work has to happen before the content goes live.
Once you know what type of article you are writing, the next step is understanding how the content should be structured.