Advanced Prompting — Get Better Results from Any AI

In the last post, we covered the basics of prompt engineering. Now let’s level up. These advanced prompting techniques will help you get consistently better results, whether you’re using ChatGPT, Claude, Gemini, or any other AI tool.

Chain of thought prompting - Get Better Results from Any AI

Instead of asking the AI for a final answer, ask it to think through the problem step by step. Add “Think through this step by step before giving your final answer” to your prompt. This forces the model to explain its reasoning, which often leads to better, more accurate results.

This is especially powerful for analytical tasks. “Analyze this network traffic log step by step. First, identify the source and destination IPs. Then determine the protocol. Then assess whether the traffic pattern is normal or suspicious.”

Few-shot prompting

We touched on this in the basics. Few-shot means giving the AI a few examples of the input and the desired output before asking it to perform the task. The more examples you provide, the more consistent the output becomes.

Example: “Here are three examples of how I summarize vulnerability findings: [example 1] [example 2] [example 3]. Now summarize this finding in the same format: [new finding].” The AI picks up on the pattern and replicates it.

Structured output

Tell the AI exactly what format you want the output in. “Return the results as a table with columns for IP address, port, protocol, and risk level.” Or “Format your response as a JSON object with fields for title, severity, description, and remediation.”

This is incredibly useful for cybersecurity tasks that require data in a consistent format for reports, tickets, or further analysis.

System prompts and custom instructions

Most AI tools let you set persistent instructions. In Claude, you can set custom instructions in your profile. In ChatGPT, you can use the custom instructions feature. These act as a baseline for every conversation.

For example, you could set: “I work in cybersecurity. When I ask about security topics, assume I have intermediate knowledge. Format technical responses with clear headers. Always include relevant NIST or CIS references when applicable.” Now every response is tailored to your context.

Prompt chaining

Break complex tasks into smaller prompts. Instead of “Write me a complete incident response plan,” try: First prompt: “List the six phases of incident response according to NIST.” Second prompt: “Now write a detailed procedure for the Containment phase, specific to a ransomware scenario.” Third prompt: “Add a communication template for notifying executive leadership during that phase.”

Each step builds on the last, and you get to review and redirect along the way. This produces much better results than trying to get everything in one shot.

Negative prompting

Tell the AI what you don’t want. “Don’t include generic advice. Don’t use bullet points. Don’t repeat information I already provided.” Sometimes, defining the boundaries is as important as defining the goal.

Bottom line

Advanced prompting is about control. The more deliberately you structure your prompts, the more useful and consistent the output becomes. These aren’t tricks — they’re communication skills adapted for AI. Practice them on real work tasks, and you’ll see the difference immediately.