Hook: Clients will ask how this works. These answers make you sound like the expert in the room.
What You Will Learn
- How the AI Visibility Score is calculated and what each sub-metric measures
- What the Dice Roller tests and how to explain its methodology to clients
- Data freshness, scan frequency, and caching behavior
- Supported languages and why locale matters for AI recommendations
- Security, privacy, and compliance posture
- Which AI models power the platform
Prerequisites
- Familiarity with the Platform Overview playbook
- Basic understanding of the four AI visibility metrics
How the AI Visibility Score Is Calculated
The AI Visibility Score is a composite number from 0 to 100 that measures how well AI models know, mention, and explain a brand.
Four sub-metrics, weighted equally:
| Metric | What It Measures | Client-Friendly Explanation |
|---|---|---|
| Brand Recall | How often AI mentions the brand when answering category questions | "When someone asks AI about your category, does your name come up?" |
| Message Alignment | How well the brand's content language matches buyer query language | "Does your website speak the same language as your customers?" |
| Value Understanding | How well AI can explain the brand's value proposition | "When AI mentions you, can it explain why you matter?" |
| Content Structure | How well the brand's content is organized for AI to reference and cite | "Can AI point to specific pages on your site to back up its claims?" |
How the score is generated:
- The platform creates 30 synthetic prompts based on the brand's category, industry, and target audience. These prompts mirror real buyer questions.
- Each prompt is sent to AI models. The responses are analyzed for brand mentions, message accuracy, value articulation, and source citations.
- The brand's results are compared against every other brand mentioned in those responses. This creates a competitive landscape baseline.
- Each sub-metric is scored from 0 to 100. The overall score is the average of all four.
Score classifications:
| Range | Classification | What It Means |
|---|---|---|
| 90 to 100 | Leader | AI consistently recommends the brand and explains why it matters |
| 70 to 89 | Challenger | AI knows the brand well; specific gaps remain that can be closed |
| 40 to 69 | Foundation Builder | AI mentions the brand occasionally; significant gaps in coverage |
| Below 40 | Unknown | AI does not meaningfully associate the brand with its category |
Pro Tip: When clients ask "What is a good score?", tell them: "Most brands we scan for the first time land between 30 and 60. The important thing is the trajectory, not the starting point."
What the Dice Roller Tests
The Dice Roller is a hallucination verification tool. It measures how consistently AI responds to the same question.
How it works:
The platform asks the same question to AI five or more times. Each response is independently generated. The tool then compares all responses to measure consistency and identify patterns.
Why repetition matters:
AI models are probabilistic. The same question can produce different answers each time. A brand that appears in one response and disappears in the next has an unstable AI presence. The Dice Roller quantifies this instability.
Two result types:
Brand Stability answers: "How reliably does AI mention this brand?" If the brand appears in three out of five responses, its Brand Stability is 60%. If a competitor appears in five out of five, their stability is 100%. This tells clients whether their AI presence is consistent or random.
Message Stability answers: "What does AI always say versus what changes?" Messages that appear in every response are core stable messages. These form the backbone of how AI represents the category. Messages that appear in some responses reveal where AI is uncertain. Variable messages represent opportunities: the brand can influence what AI says in those gaps.
Confidence methodology:
The Dice Roller achieves approximately 90% confidence with five iterations. This means the patterns identified in the results are statistically reliable. If a brand appears in zero out of five responses, there is high confidence that AI does not currently associate that brand with the query topic.
Checkpoint: You can explain Brand Stability and Message Stability to a client without using the words "probabilistic," "statistical," or "iteration."
Data Freshness and Scan Frequency
Scan tiers and what they cover:
| Tier | Pages Analyzed | Typical Duration | When to Use |
|---|---|---|---|
| Free Magnet Scan | 5 pages | 60 seconds | Quick check during a meeting |
| Standard | up to 100 pages | 30-60 minutes | Default for most scans up to Strategist Tier |
| Deep | up to 500 pages | 60-180 minutes | E-commerce sites, large content libraries - Strategist Tier |
| Enterprise | custom pages | 60-600 minuters | Enterprise clients with complex site architectures |
Caching behavior:
Scan results are cached for performance. When you rescan the same URL, the platform returns cached data unless you explicitly request a fresh scan. For demos and monitoring, cached data is usually sufficient. After publishing new content, request a fresh scan to verify the impact.
Recommended scan frequency:
- Monthly: Standard monitoring cadence. Catches shifts in AI model behavior and competitor changes.
- After major content updates: Enterprise Tier. When the client publishes new pages, restructures their site, or launches a new product, scan again to verify impact.
AI model update cycles:
Different AI models update their training data at different frequencies. Some models incorporate new content within weeks; others take months. Monthly rescans help clients understand which models are responding to their content changes first.
Supported Languages
The platform supports 12 locales + multiple world-spoken languages. Natively support languages are:
| Language | Locale Code |
|---|---|
| English | en |
| German | de |
| Polish | pl |
| French | fr |
| Spanish | es |
| Dutch | nl |
| Italian | it |
| Portuguese | pt |
| Finnish | fi |
| Swedish | sv |
| Danish | da |
| Estonian | et |
Why language matters for AI visibility:
Our research demonstrates that the language of a buyer's query significantly changes which brands AI recommends. A buyer asking in German receives different brand recommendations than the same buyer asking the same question in English. This means brands operating in multiple markets need to monitor their AI visibility in each language independently.
Locale-aware scanning:
The scanner auto-detects locale from the URL structure. URLs with locale segments (such as /de-de/, /pl-pl/, or /en-us/) are recognized automatically. The scan locks to that locale, ensuring the analysis reflects how AI perceives the brand in that specific market.
Client talking point:
"If your clients sell in Germany and Poland, they need separate scans for each market. AI treats each language as a different competitive landscape."
Security and Privacy
Partners need confident answers about data handling, especially when selling to enterprise clients.
Data residency:
All infrastructure is hosted in europe-north1 (Finland). Data processing stays within the EU. This satisfies GDPR data residency requirements for European clients.
What we store:
- Scan results (aggregated analysis, not raw page content)
- AI Visibility Scores and sub-metric breakdowns
- Generated personas and content recommendations
- User account information (email, organization)
What we do not store:
- Raw page content from scanned websites (content is analyzed in memory and discarded)
- Client access tokens or credentials in logs
- Personally identifiable information from scanned pages
Access controls:
- Access tokens are stored server-side only; they never appear in client-side error messages or browser storage
- External scans are rate-limited per origin to prevent abuse
- All API communication uses HTTPS with TLS encryption
Compliance posture:
- GDPR compliant (EU data residency, data minimization, right to deletion)
- SOC 2 compliance is on the product roadmap
- Security-first architecture reviewed against OWASP standards
When enterprise clients ask about security audits:
"We host entirely within the EU on Google Cloud infrastructure in Finland. We follow GDPR data minimization principles: we analyze content during the scan and store only the aggregated results, not the raw content. We are working toward SOC 2 certification and can provide our security documentation on request."
Supported AI Models
For analysis (scanning and scoring):
The platform runs analysis on Google Gemini by default. OpenAI is available as an alternative. The choice of analysis model affects processing speed and cost but produces comparable results. Gemini is recommended for most scans due to its speed and cost efficiency.
For the Dice Roller (consistency testing):
The Dice Roller tests against multiple AI models to give a comprehensive view of brand stability. This is important because a brand might appear consistently in one model's responses and never in another's.
For content generation (Content Plan):
Content recommendations are generated using Gemini 2.5 Flash. This model balances speed, cost, and output quality for generating content strategies, article outlines, and keyword recommendations.
Client talking point:
"We test your brand's visibility across multiple AI models because each one has different training data and recommends different brands. What works in ChatGPT might not work in Gemini or Claude. Our platform gives you the full picture."
Pro Tip: If a client asks which specific AI models are tested, the honest answer is that models are updated as new versions release. The platform abstracts this complexity so clients focus on results, not model versions.
Common Partner Questions
"Can clients access the platform directly?"
Yes. Each client account has Playground access where they can view their scanned projects, AI Visibility Scores, personas, and content plans. Partners can manage client access and set up shared projects.
"What happens if a client's website blocks our scanner?"
The scanner respects robots.txt and standard access controls. If a site blocks automated access, the scan will return partial results based on accessible pages. In most cases, standard websites allow full scanning. If a client's security team needs to whitelist our scanner, contact partner support for the IP ranges.
"Can we white-label the reports?"
The reporting system supports co-branding. Partners can add their agency logo and customize report headers. Full white-label options are available at the Enterprise partner tier.
"How do scores compare across industries?"
Score distributions vary by industry. SaaS companies tend to have higher baseline scores because their content is naturally structured for AI consumption. Manufacturing and professional services firms typically start lower because their content is designed for human readers, not AI models. The platform's nine vertical specializations account for these differences in its recommendations.
"What if a client's score drops after content changes?"
Score fluctuations are normal, especially after significant content changes. AI models need time to incorporate new content. A temporary dip followed by improvement is a common pattern. If a score drops and stays down, it usually indicates that the new content is less AI-readable than what it replaced. The Content Structure sub-metric will highlight the specific issue.
Where to Go From Here
- Platform Overview for Partners: Full platform walkthrough for new partners
- Running a Live Demo: Step-by-step demo guide with timing and objection handling
- Closing and Pricing: Pricing tier positioning and ROI talking points