Context

Why this guide matters

Most teams ask AI vague questions like "what do you think about our brand?" and get soft, generic answers. That does not help positioning, demand strategy, or AI recommendation visibility. If you want useful signals, you need prompts that force tradeoffs, expose narrative gaps, and test whether your public signals are strong enough to be retrieved and recommended.

This guide gives you seven production-ready prompts you can run directly in ChatGPT, Gemini, or Claude. Together, they diagnose unaided recall, category narrative power, semantic associations, trust weaknesses, competitive differentiation, retrieval-ready positioning, and moat durability over time.

Use these prompts with your real category terms, named competitors, and current value proposition. Then convert findings into concrete actions: content updates, trust-proof upgrades, positioning rewrites, and prompt-monitoring workflows.

Executive Summary

Key takeaways

• Run these prompts monthly to track how model perception changes over time.
• Use concrete placeholders: category, competitor, use case, and your exact value proposition.
• Treat model outputs as directional intelligence, then validate with real market and performance data.
• Prioritize fixes with compounding impact: trust proof, narrative clarity, and retrieval language.
• Feed strong outputs into your website copy, comparison pages, and source strategy.

Why these 7 prompts work better than generic brand questions

Each prompt is designed to reduce polite, low-signal responses. They introduce constraints such as ranking pressure, explicit scoring dimensions, steelman critique, and scenario-based forecasting. This makes outputs more diagnostic and less promotional.

The sequence also matters. Start with recall and narrative context, then move into trust and differentiation, then finalize with retrieval-focused rewriting and moat diagnosis. This mirrors how buyers and recommendation systems converge: first awareness, then credibility, then fit.

For best results, run the same prompt across multiple models and compare overlap versus disagreement. Where all models converge, you likely have a real positioning signal. Where they diverge, you likely have weak or inconsistent public evidence.

Prompt 1) Unaided Recall Test

This prompt tests top-of-mind presence in your category without giving the model a favorable setup. It helps you understand whether your brand enters the shortlist naturally or only when explicitly named.

Prompt 2) Category Narrative Audit

Share-of-voice alone is not enough. The winning brand often owns the category story. This prompt identifies who currently controls that story and whether your message is differentiated or derivative.

Prompt 3) Semantic Brand Fingerprint

AI retrieval behavior is influenced by concept clusters. This prompt maps the concepts and emotions your brand currently triggers, then compares them to the market leader to reveal overlap and white space.

Prompt 4) Trust Signal Blind Spots

Recommendation systems over-index on trust evidence, especially for B2B decisions. This prompt scores your public signals across competence, benevolence, integrity, and predictability, then identifies the highest-leverage fix.

Prompt 5) Differentiation Stress Test

Most positioning sounds unique until compared under pressure. This prompt forces a steelman argument that your value proposition is interchangeable with a known competitor, then reveals what survives.

Prompt 6) Positioning for AI Retrieval

This is the operational bridge from diagnosis to execution. The prompt asks the model to rewrite your positioning for retrieval and recommendation behavior, not for generic brand copy aesthetics.

Prompt 7) The Brand Moat Diagnosis

Positioning quality is not static. This prompt stress-tests whether your current story compounds into a stronger moat or decays as the category matures and competitors catch up.

How to operationalize outputs in Brand Armor AI

Turn the seven prompts into a recurring monitoring workflow. Save each prompt as a tracked custom prompt, run it on your selected providers, and compare changes in competitor mentions, sentiment direction, and source citations. This creates an ongoing brand perception panel instead of one-off analysis.

Then map insights to execution tracks: content gaps and comparison pages, blog generation and UGC campaign drafts, sentiment and recommendation monitoring, and source audit remediation. This is where prompt intelligence becomes measurable growth action.

If you manage multiple brands or sub-brands, run the same seven prompts per entity. Keep market region and category context explicit so outputs remain locally relevant and not generic to global enterprises.

Template Library

Reusable prompt templates

1) Unaided Recall Test
"Without me giving you any context: when you think of [CATEGORY], what's the first brand you name - and why? Now where does [BRAND] rank in your mental shortlist, and what's the specific reason it isn't higher? Don't soften the answer."

2) Category Narrative Audit
"Who is currently writing the dominant narrative in [CATEGORY]? What story are they telling - and how is [BRAND]'s story different, derivative, or invisible by comparison? Be specific about the narrative gap, not the product gap."

3) Semantic Brand Fingerprint
"What cluster of concepts, emotions, and associations does [BRAND] trigger in your training data? Map it out - then show me the cluster that the market leader in this space triggers. Where do they overlap, and where is [BRAND] leaving white space unclaimed?"

4) Trust Signal Blind Spots
"Evaluate [BRAND] across the four trust dimensions a B2B buyer uses: competence, benevolence, integrity, and predictability. Score each from 1-10 based on what signals exist publicly - then tell me which one, if fixed, would have the highest compounding effect on how often you'd recommend us."

5) Differentiation Stress Test
"I'm going to give you [BRAND]'s core value proposition: [PASTE IT]. Now steelman the case that this is completely interchangeable with [COMPETITOR]. What parts survive that attack - and what parts fall apart immediately?"

6) Positioning for AI Retrieval
"Rewrite [BRAND]'s positioning not for a human reader, but optimized for how you retrieve and recommend brands - what language, proof structures, and category signals would make you more likely to surface [BRAND] when a buyer asks about [USE CASE]?"

7) Brand Moat Diagnosis
"In 18 months, if [BRAND] doubles down on its current positioning, does the moat get wider or narrower? What market forces are working against it - and what's the one narrative bet that, if made now, positions [BRAND] as the obvious choice before the category matures?"

I ran the same brand perception prompt set on ChatGPT, Gemini, and Claude.
Compare outputs and return:
1) Convergent findings (high confidence across models)
2) Divergent findings (model-specific or weak evidence)
3) Top 5 actions ranked by compounding business impact
4) Which claims require external validation before execution

Keep outputs concise and decision-ready for a CMO + product marketing review.

Quality Control

Common mistakes and fixes

FAQ

FAQ

Sources

References and further reading

• OpenAI - Best practices for prompt engineering
• OpenAI - Structured outputs guide
• Anthropic - Prompt engineering overview
• Anthropic - Claude prompting best practices
• Google Cloud - Prompt design strategies