How to build a balanced prompt portfolio in BrandHalo

A balanced prompt portfolio is not one with equal numbers in every category. It is one where the composition is deliberate, documented and matched to the questions your organisation needs AI visibility monitoring to answer.

The practical goal is to prevent the easiest prompts to add, usually Brand and Comparison prompts, from dominating the headline result. At the same time, you still need enough of them to understand representation and competitive positioning.

Step 1: define the decisions the data should support

Before writing prompts, agree on the decisions the monitoring programme should inform. A CMO may need to know whether the brand enters category conversations. A brand manager may need to identify inaccurate descriptions. A content lead may need to see which topics and sources influence citations.

Write down three to five measurement questions, such as:

• Are answer engines recommending us for our priority categories?
• Do our most important audience segments see us as relevant?
• Is our positioning accurate when buyers compare suppliers?
• What does AI get wrong when people ask directly about us?

Step 2: choose an intentional starting mix

There is no accepted industry standard for category weights. External guidance supports prioritising unbranded prompts for category-level discovery, but your exact mix should reflect your market and measurement purpose.

For a broad discovery and consideration programme, an illustrative starting allocation is:

• 40% Generic: unprompted category discovery
• 30% Persona: audience and use-case relevance
• 20% Comparison: competitive positioning
• 10% Brand: direct knowledge and representation

These percentages are a working hypothesis, not a BrandHalo or industry benchmark. A new category entrant may place more weight on Generic discovery. A well-known organisation dealing with misinformation may give Brand prompts more weight. A procurement-led market may need a larger Comparison sample.

Step 3: build coverage before volume

In BrandHalo AI Visibility Monitoring, start with the generated suggestions from your approved brand profile, then review every prompt. Suggestions can draw on your category, personas and competitive landscape, but the final portfolio needs human judgement.

Within each topic, cover the dimensions that matter to your strategy:

• Priority products, services and category language
• Key personas, roles, goals and pain points
• Customer journey stages from discovery to evaluation
• Regions, industries and regulatory contexts
• Named competitors and non-brand alternatives

Five well-chosen prompts across distinct needs are more informative than 25 near-duplicates that ask which platform is "best" with minor wording changes.

Step 4: preserve a stable baseline

Once a prompt set is approved, treat its composition as part of the metric definition. Record the prompt count by topic and the date the baseline began. When you need to expand the set, add prompts in planned review windows instead of continuously changing the portfolio.

If the mix changes materially, do one of two things:

1. Start a new baseline. Use this when the market, product or audience strategy has changed.
2. Report category results separately across the transition. This preserves useful comparisons even when the blended score is no longer directly comparable.

Step 5: check the mix before every executive update

Include a short measurement note beside each headline result:

AI visibility was measured across 100 prompts: 40% Generic, 30% Persona, 20% Comparison and 10% Brand. The prompt mix was unchanged from the previous reporting period.

That one sentence makes the number far easier to trust. It tells readers what was measured and whether movement is likely to reflect answer-engine behaviour or a change in the lens. Part four addresses the other half of trustworthy measurement: whether the prompts are representative of real customer questions.