AI Secret Shopper Intelligence

Market Recommendation Share

We ask ChatGPT, Claude, and Gemini: "I need a [your niche] in [your city] — who do you recommend?" — then check if the AI recommends you. Temperature: 0. No hallucinations.

What is Market Recommendation Share?

Consumers are increasingly turning to AI assistants — ChatGPT, Claude, and Google Gemini — to find local businesses. Instead of scrolling through 10 blue links, they ask a simple question and get a direct recommendation.

Market Recommendation Share measures whether those AI models recommend your business when a customer in your city asks for your service. Think of it as your "share of AI voice."

Why it matters: If AI doesn't recommend you, it's recommending your competitor. By 2026, an estimated 30% of local discovery will flow through AI assistants. Your visibility in these models is your future revenue.

How the AI Secret Shopper Works

Our "Secret Shopper" probe sends a real consumer question to three major AI models — simultaneously — and analyzes every response. Here's the exact flow:

You enter your business details

Business name, category/niche, city, and website URL. The cockpit auto-fills these from your latest scan data.

We send the probe question

Each AI model receives the same neutral prompt at temperature 0 (the most deterministic setting — no creative randomness). The models respond with their top 3 business recommendations.

We analyze each response

For every provider, we check: Were you mentioned? What rank? Did they cite your URL? What was the sentiment? Each factor contributes to a 0–100 visibility score per model.

We explain why you're invisible

For any AI that didn't recommend you, a second AI call generates a detailed explanation of why — referencing the competitors it chose instead and what they have that you don't.

You get actionable intelligence

A full narrative analysis, per-provider breakdown, "Why Invisible" cards, and even a ready-to-send lead outreach email are generated from the results.

The Exact Prompt We Use

Transparency is core to our methodology. Here is the exact system and user prompt sent to each AI model:

System Prompt You are a neutral, objective consumer researcher. Your goal is to provide local business recommendations based solely on verifiable data, authority, and digital presence. Do not favor any brand unless it has superior digital congruency and public data.
User Prompt I am in [city] and I need a [niche]. Based on your current knowledge of the digital landscape, which 3 businesses would you recommend I contact first? For each, provide a brief "Reason for Recommendation." If you cannot find enough data to verify a business, do not include it.
Temperature: 0 — This is the lowest randomness setting available. It means every probe returns deterministic, reproducible results. No hallucinations, no creative guesses. If the AI recommends a competitor, it's because their digital signals are genuinely stronger.

Three AI Models, One Simultaneous Probe

We probe all three models in parallel. Each has different training data and different strengths, giving you a 360° view of your AI visibility.

🤖

ChatGPT (OpenAI)

gpt-4o

The most widely used consumer AI. If ChatGPT doesn't recommend you, millions of potential customers won't find you through it.

🧠

Claude (Anthropic)

claude-haiku-4

Known for careful, factual responses. Claude emphasizes verifiable data — strong schema markup and structured data matter most here.

💎

Google Gemini

gemini-1.5-pro

Integrates with Google's knowledge graph. Businesses with strong Google Business Profiles and consistent NAP data perform best here.

How Your Score is Calculated

Each AI provider generates a 0–100 visibility score based on five signals extracted from the model's response:

SignalWhat it meansPoints
MentionedThe AI mentioned your business by name anywhere in its response+50
URL CitedThe AI included a link to your website — a strong trust signal+20
Top 3 RankYou appeared in the AI's top 3 recommendations+15
#1 RankYou were the first business the AI recommended+10
SentimentPositive language about your business adds points; negative language penalizes+10 / −30

Your aggregate score is the average across all available providers, producing one of four visibility statuses:

🌟
AI Darling
Recommended by all models
Partially Visible
Some models recommend you
⚠️
Barely Visible
Mentioned but not recommended
👻
AI Invisible
No AI recommends you

"Why Am I Invisible?" — The Money-Maker

The most valuable part of the probe isn't discovering that you're invisible — it's understanding why.

For every AI model that didn't recommend you, we run a second analysis that examines the competitors it did choose and explains the gap. Common reasons include:

🏗️ Entity Trust Gap

Competitors have richer schema markup — LocalBusiness, FAQ, Review — that makes them verifiable to AI.

📍 Geography Mismatch

Your business address or service area is ambiguous. AI can't confidently place you in the right city.

🔗 Metadata Deficit

Missing title tags, meta descriptions, or canonical URLs make your site unreadable to AI parsers.

😐 Sentiment Conflict

Negative reviews or inconsistent brand messaging causes AI to avoid recommending you.

Deterministic fallback: If the AI analysis provider is temporarily unavailable, we use a pattern-matching engine that generates technical insights based on your scan data — so you always get an explanation, never a blank.

Frequently Asked Questions

Is this different from the AI Visibility Score?
Yes. The AI Visibility Score measures whether your website is structured for AI discovery (schema, bot access, speed, SEO). Market Recommendation Share measures whether AI actually recommends you when asked. You can have a perfect score but still be invisible if your competitors' digital signals are stronger.
What tier do I need to use this?
Market Recommendation Share requires a Starter ($27/mo) tier or higher. The probe uses real API calls to three AI providers, which have associated costs. Free-tier users can see a preview of the feature in their cockpit but cannot run probes.
How often should I run the probe?
We recommend probing once after implementing fixes, and then monthly to track progress. AI models update their knowledge periodically, so your visibility can change. Results are cached for 24 hours to prevent redundant API costs.
Can AI recommendations change?
Yes. AI models are periodically retrained and updated. As you improve your schema markup, NAP consistency, review profile, and content authority, AI models become more likely to recommend you. That's the whole point — this is actionable intelligence, not a static report.
Why temperature 0?
Temperature 0 is the most deterministic setting. It means each probe produces consistent, reproducible results. If we used a higher temperature, the AI might "hallucinate" different recommendations each time, making the data unreliable. At temp 0, what the AI says reflects its actual knowledge — not creative guessing.
What if all providers show "Unavailable"?
This means the AI provider APIs were temporarily rate-limited or unreachable. The error is automatically cleared from your cockpit — it won't persist across page refreshes. Try again in a few minutes.

Ready to see who AI recommends?

Run a $27 audit to check your AI Visibility Score, then upgrade to Starter to unlock the AI Secret Shopper probe.

Run AI Scan