🗺️ Overview
When using the Digital Knowledge Network (DKN) to generate articles at scale, you may need to apply consistent content refinements to every article — such as adding an About the Author section, inserting contact information, or stripping unwanted formatting tags before articles deploy to GoHighLevel. Instead of applying these changes manually, you can automate them using DKN's custom prompt refinement settings.
This article explains how to configure a reusable refinement prompt that runs automatically on every DKN-generated article before deployment.
⚙️ What Is a Custom Prompt Refinement?
A custom prompt refinement is an instruction set you define once inside DKN. After an article is generated, DKN passes it through your refinement prompt as a post-processing step — adjusting content, structure, or formatting without you needing to intervene manually each time.
Common use cases include:
- Appending a standardised About the Author section to every article
- Inserting location-specific contact details or calls to action
- Removing unwanted HTML or formatting tags produced during generation
- Enforcing a consistent tone, disclaimer, or footer block across all sites
🛠️ How to Set Up an Automated Refinement Prompt
- Open DKN from your Search Atlas dashboard and navigate to the Content Generation settings for your target site or network.
- Locate the Custom Prompt or Post-Processing Prompt field within the article generation configuration. This is where you define the refinement instructions that will apply to every article.
- Write your refinement prompt. Be specific and use clear instructions. For example:
- "After the final paragraph, add a section titled 'About the Author' followed by [your author bio text]."
- "Insert the following contact block at the end of the article: [phone, address, website]."
- "Remove all instances of markdown formatting symbols such as **, ##, and — from the article body."
- Save your prompt configuration. DKN will now apply this prompt automatically to each article generated for this site during the refinement pass.
- Run a test generation on a single article to confirm the refinement is applied as expected before enabling it across your full network.
- Once validated, enable the setting for all sites in your DKN network if you want the same refinement applied universally, or configure it per site for tailored outputs.
✅ Best Practices for Reliable Refinement Prompts
- Keep instructions atomic. Break each refinement into a clear, single action. Avoid combining too many instructions in one sentence — this reduces the chance of any step being skipped.
- Use delimiters for inserted content. When adding blocks like author bios or contact sections, wrap the placeholder text in quotes or brackets so the AI can distinguish your instruction from the content itself.
- Avoid conflicting instructions. If your base generation prompt and your refinement prompt give opposing direction (e.g., one adds a section, the other removes all footers), the output may be unpredictable. Review both prompts together.
- Test on a staging site first. Before deploying refinements across live GoHighLevel sites, validate outputs on a test site to catch formatting or content issues early.
- Version your prompts. Keep a copy of each refinement prompt version in a document. If results change after an update, you can quickly reference what changed.
🚀 Deploying Refined Articles to GoHighLevel
Once your refinement prompt is configured and tested, DKN will automatically apply it to every article in the generation queue. Refined articles are then passed to your GoHighLevel deployment pipeline without any additional manual steps required.
If you are deploying to multiple GoHighLevel sites, confirm that each DKN site node has the correct refinement prompt assigned — especially if different sites require different author bios or contact details.
⚠️ Troubleshooting Common Issues
- Refinement is not applying: Confirm the custom prompt field is saved and that post-processing is enabled for the site. Unsaved changes are a common cause.
- Formatting tags still appearing: Make your removal instruction more explicit, specifying the exact tags or symbols to strip (e.g., "Remove all instances of the characters ** and __ from the article text").
- About the Author section is missing: Check that the prompt instruction references the correct position (e.g., "at the end of the article" vs. "after the conclusion heading") and that the bio text is included in the prompt itself.
- Generation timeout during refinement: If a refinement step causes a timeout, simplify the prompt. Overly complex or lengthy instructions can exceed processing limits. Split refinements into smaller, focused prompts where possible.
💬 Need More Help?
If you need further assistance, open the chat widget in the bottom-right corner of the platform and type human teammate to be connected with a member of our team.