Designing a Campaign Around One High-Value Prompt Cluster

TL;DR

Why one prompt cluster beats many

A practice that tries to optimise for every possible question spreads its effort thin. A single high-value prompt cluster — a group of related questions your prospects actually ask — gives you a focused target:

• One question family. For example, "best clinic for acne scars in Central" and its variants.
• One language. English or 中文, not both at once.
• One surface. The Hong Kong surface or the overseas/inbound surface.
• One metric. Share of AI recommendations for that cluster.

How to design the campaign

1. Choose the question family

Pick a question with high intent. The best clusters are:

• Treatment-specific. "Which clinic is best for pico laser in Hong Kong?"
• Condition-specific. "Which law firm handles commercial disputes?"
• Location-specific. "Best dermatologist in Central?"

2. Write the prompt set

Write 10–15 variants of the same question. Change the wording, not the intent. Include:

• "best", "top", "recommended", "who do you suggest"
• English and 中文 variants
• The same location and specialty

3. Run the baseline

Run the full prompt set across the assistants your market uses. Record the share of recommendations for each practice. This is your baseline.

4. Optimise the sources

• Add the practice to the directories and review platforms the AI cites for that cluster.
• Fix outdated fees, closed locations, and wrong services.
• Add structured data to the practice's own pages.
• Ensure the practice name, address, and specialty are consistent across sources.

5. Re-measure and report

Re-run the prompt set after remediation. The movement from absent to recommended is the metric that matters.

What the audit delivers

• A prompt set built around one high-value cluster
• The baseline share of recommendations
• A fix list for the sources the AI cites
• Before/after reporting with screenshots