Your AI Visibility Score (0–100) measures one thing: how likely AI assistants are to recommend your product when someone asks about your category.
We believe scoring should be transparent, not a black box. Here's exactly how it works.
Two Layers Working Together
Most AI visibility tools only do one thing — they search the web and count mentions. We do that and something no one else does:
Layer 1 — Real-time Web Scan: We search the live web for your product and analyze what signals AI systems can find about you.
Layer 2 — AI Probe (Direct Query): We literally ask Perplexity and Gemini: "Do you know [your product]? Would you recommend it?" — and show you the raw response. This is what real users see.
Layer 1: The 4 Dimensions (0–100 points)
Your score comes from 4 weighted dimensions. Each measures a different signal that AI systems use when deciding what to recommend:
| DIMENSION | POINTS | WHAT IT MEASURES | HOW TO IMPROVE |
|---|---|---|---|
| Recognition | 35 | Do AI systems know your product exists by name? This is the foundation — AI can't recommend what it doesn't recognize. | Get mentioned on directories, high-authority publications, and third-party review sites. |
| Category Ranking | 35 | Does your product appear in "best of" lists, roundups, or category comparisons in AI responses? | Get included in "Top 10 [category]" articles. Create comparison content. Pitch reviewers. |
| Co-Recommendation | 20 | Do real people and trusted sources recommend your product alongside other known products? Community signals matter. | Build genuine presence on Reddit, Product Hunt, and G2. Real discussions beat manufactured mentions. |
| Web Signals | 10 | Does your site have the structured signals (llms.txt, schema markup, clean metadata) that help AI systems index and understand you? | Add an llms.txt file. Add schema.org SoftwareApplication markup. Ensure your metadata is accurate. |
Why these weights?
Recognition and Category Ranking each receive 35 points — together they determine 70% of your score — because being known and being recommended are the two core outcomes that matter. Co-Recommendation (20 pts) reflects community trust, and Web Signals (10 pts) captures the technical signals that help AI systems discover and understand you.
These weights are based on our analysis of what correlates with actual AI recommendations across hundreds of product checks. We adjust them as we learn more.
Layer 2: AI Probe — We Ask AI Directly
This is what makes us different from every other GEO tool.
After calculating your web-based score, we send real queries to AI systems and ask them about your product directly:
The AI Probe doesn't add to your numeric score — it gives you a separate KNOWN / UNKNOWN status for each AI system. Combined with your 4-dimension score, you get the full picture.
Data Sources
Key technical details:
- Real-time data — We query the live web every time you check. No cached results from weeks ago.
- Deterministic scoring — Same product name → same score. We use temperature=0 and pattern matching to eliminate randomness.
- Self-reference filtering — We automatically exclude your own website from scoring. Getting mentioned on your own site doesn't count.
- 3-layer fallback — If our primary data source is unavailable, we fall back to secondary and tertiary sources to ensure you always get a result.
What We Don't Do
- We don't guarantee higher AI visibility. We measure it. Improving it requires real work: getting mentioned, building community presence, earning reviews.
- We don't manipulate AI systems. Your score reflects what AI sees right now. We don't inject data or game the system.
- We don't share your data. Your product checks are private. We don't publish scores without your consent.
Known Limitations (We're Honest About These)
Sample size: Our web scan analyzes the top 7+ search results. This is a representative sample, not an exhaustive crawl of the entire internet. Products with very niche naming may get fewer relevant results.
AI visibility is a lag indicator. When you publish a blog post or get mentioned on Reddit today, it takes 2–6 weeks for that signal to fully propagate through AI systems. Your score today reflects actions taken weeks ago.
We're early. pickedby.ai launched April 5, 2026. Our methodology improves with every product we scan. We update our scoring algorithm based on what actually correlates with real AI recommendations.
Our Own Score: 12 out of 100
We don't hide behind our own tool. We ran pickedby.ai through our own scoring engine:
- Score: 12/100 — AI has almost no idea we exist
- Perplexity: UNKNOWN — has never heard of us
- ChatGPT: UNKNOWN — does not recognize us
We're documenting the journey from 12 to (hopefully) much higher. Read the Day 0 story →
Questions?
If you have questions about our methodology, data sources, or scoring — reach out at hello@pickedby.ai.
We believe in transparency. If you think our scoring should work differently, we want to hear about it.