I ran 96 YC-backed companies through ChatGPT, Claude, and Gemini. Each company got 16 prompts: 5 category queries (brand-agnostic like "Best X tools in 2025"), 5 brand queries, and 5 competitive queries.
The correlation between ChatGPT and Claude citation rates? 0.39.
That's weak. A company ranking well on ChatGPT has less than 40% statistical relationship with how it ranks on Claude. For Claude vs Gemini, it drops to 0.17—basically noise.
The Numbers
| Platform | Avg Citation Rate | Position Avg | Position Std Dev |
|---|---|---|---|
| Gemini | 71.3% | 1.8 | 2.2 |
| ChatGPT | 69.7% | 6.3 | 19.5 |
| Claude | 67.3% | 2.0 | 6.5 |
The averages look similar. But look at ChatGPT's position standard deviation: 19.5.
Gemini's standard deviation is 2.2. When Gemini cites you, it's consistently at position 1-2. When ChatGPT cites you, it could be position 1 or position 140. Same company, same query type.
The Distribution
| Visibility | ChatGPT | Claude | Gemini |
|---|---|---|---|
| 0% | 0 | 0 | 0 |
| 1-24% | 0 | 4 | 0 |
| 25-49% | 0 | 0 | 0 |
| 50-74% | 92 | 88 | 80 |
| 75-99% | 1 | 2 | 14 |
| 100% | 3 | 2 | 2 |
Three patterns:
-
Claude has a floor. 4 companies get cited in fewer than 25% of queries. ChatGPT and Gemini have zero companies in this bucket.
-
Gemini clusters high. 14 companies reach 75-99% visibility on Gemini. ChatGPT has just 1.
-
ChatGPT is binary. Almost everyone lands at 50-74%. You're either visible or you're not.
The Extremes
Position divergence is where things get insane:
| Company | ChatGPT | Claude | Gemini |
|---|---|---|---|
| CloudHumans | #140 | #1 | #1 |
| Leapingai | #96 | #62 | #1 |
| Newscatcher API | #72 | #1 | #9 |
| AgentHub | #56 | #15 | #1 |
| Momentic Marketing | #37 | #1 | #1 |
CloudHumans does customer support software. Claude and Gemini put them first. ChatGPT buries them at position 140.
That's not a ranking difference. That's invisibility.
Claude's Blind Spots
Four companies get cited well by ChatGPT and Gemini but barely exist on Claude:
| Company | ChatGPT | Claude | Gemini |
|---|---|---|---|
| Leapingai | 69% | 6% | 62% |
| CloudHumans | 69% | 12% | 69% |
| AgentHub | 69% | 19% | 69% |
| Noble Stay | 69% | 19% | 69% |
Leapingai is an AI call center company. ChatGPT cites them in 11 out of 16 queries. Claude cites them in 1.
When Claude does cite them, it's at position 62. When it doesn't, they don't exist.
Gemini's Advantage
Gemini has 14 companies at 75-99% visibility. ChatGPT has 1.
| Company | ChatGPT | Claude | Gemini |
|---|---|---|---|
| Vapi | 69% | 69% | 94% |
| Waydev | 69% | 69% | 94% |
| Mem0 | 69% | 69% | 94% |
| Unsloth | 69% | 69% | 94% |
| Newscatcher API | 69% | 63% | 88% |
| Prosper | 69% | 69% | 88% |
| Abacum | 69% | 69% | 81% |
Why? Gemini uses Google Search. If you rank well on Google, Gemini sees you.
ChatGPT Buries, Claude Elevates
5 companies get buried past position 20 on ChatGPT. Only 1 on Claude. Zero on Gemini.
Companies where Claude and Gemini put first, ChatGPT buries:
| Company | ChatGPT | Claude | Gemini |
|---|---|---|---|
| Hyperbrowser | #12 | #1 | #1 |
| Afterquery | #9 | #1 | #1 |
| Codeviz | #14 | #1 | #1 |
| Mayalabs | #5 | #1 | #1 |
| LLM Data | #13 | #7 | #1 |
Position 1 means "Use Afterquery." Position 9 means "Also consider Afterquery, among others..."
Sentiment Divergence
I tracked sentiment across all responses:
| Platform | Positive | Neutral | Negative |
|---|---|---|---|
| Claude | 124 | 885 | 25 |
| ChatGPT | 85 | 962 | 23 |
| Gemini | 41 | 1045 | 9 |
Claude generates 46% more positive mentions than ChatGPT. And nearly 3x more than Gemini.
Gemini is almost entirely neutral. It cites you, but it doesn't advocate for you.
Technical GEO Doesn't Predict Visibility
I calculated a GEO score for each company based on schema markup, llms.txt, meta tags, and content quality. Then I correlated each factor with actual visibility:
| Factor | Correlation |
|---|---|
| Sentiment Score | 0.68 ✓ |
| GEO Score (composite) | 0.18 |
| llms.txt file | 0.18 |
| Content Quality | 0.12 |
| Schema Markup | 0.10 |
Sentiment is the only factor that actually predicts visibility.
The technical stuff? Basically noise.
I dug into the academic research to understand why. The answer is fascinating: LLMs destroy schema markup during tokenization. When a model processes "@type": "Organization", it breaks this into individual tokens—@, type, Organization—that become indistinguishable from the same words in regular prose.
The markup's semantic meaning literally gets shredded before the model ever sees it.
Same with llms.txt. LLMs use semantic retrieval—they query vector indexes based on meaning, not metadata files. Your llms.txt is invisible to this process unless it happens to contain semantically relevant text.
What the research says actually matters
According to recent empirical studies:
| Factor | Impact |
|---|---|
| Brand search volume | 0.33 correlation (strongest technical signal) |
| Wikipedia presence | 47.9% of ChatGPT's top-10 citations |
| Cross-platform presence | 2.8x more likely to appear if on 4+ platforms |
| Content freshness | 65% of AI bot hits target content from past year |
The pattern is clear: LLMs don't reward technical optimization. They reward market presence and reputation.
This explains why Lightly (GEO score 39) outperforms companies with scores in the 70s-80s. They have stronger market presence and more consistent positive coverage.
Only 3 Companies Win Everywhere
Out of 96 YC companies, only 3 achieve 75%+ visibility on all three platforms:
| Company | ChatGPT | Claude | Gemini | GEO Score |
|---|---|---|---|---|
| Flow.club | 100% | 94% | 94% | 67 |
| Sully | 100% | 100% | 75% | 59 |
| Weave | 75% | 100% | 75% | 72 |
That's 3%. Everyone else has at least one platform where they're significantly weaker.
Flow.club does virtual coworking. Sully is a healthcare AI assistant. Weave is patient communication software.
What do they have in common? All have clear, established categories. All have consistent messaging. None are in crowded AI-tool categories where dozens of competitors fight for the same keywords.
Why This Happens
Three architectural differences explain the divergence:
Different Search Backends
- ChatGPT uses Bing
- Gemini uses Google Search
- Claude uses multiple providers
CloudHumans probably ranks well on Google but poorly on Bing. Gemini cites them first. ChatGPT buries them at position 140.
Different Filtering Logic
Claude applies stricter quality thresholds. That's why 4 companies fall into the 1-24% bucket—they either meet Claude's bar or they don't.
ChatGPT is more inclusive. If you exist somewhere in Bing's index, ChatGPT will cite you. But the position is unpredictable.
Different Position Certainty
ChatGPT's high position variance (std dev 19.5) suggests it's synthesizing from many competing sources with uncertainty about which is most relevant. Gemini's consistency (std dev 2.2) suggests it has strong opinions about source hierarchy—likely because it relies on Google Search rankings.
What This Means
You can't "optimize" your way to AI visibility with technical tricks. The 0.68 sentiment correlation tells you what actually matters: how people talk about you.
What to stop doing
- Don't add schema markup expecting AI visibility gains. It helps Google, not LLMs.
- Don't create llms.txt files. LLMs don't read them.
- Don't chase GEO scores. The composite score has 0.18 correlation—basically noise.
What to start doing
- Earn positive coverage. Get mentioned in publications, communities, and reviews. The 0.68 sentiment correlation is real.
- Build Wikipedia presence. 47.9% of ChatGPT's top citations come from Wikipedia.
- Expand platform presence. Companies on 4+ platforms are 2.8x more likely to appear in ChatGPT.
- Keep content fresh. 65% of AI bot traffic goes to content from the past year.
Platform-specific strategies
ChatGPT: Broad web presence matters. Get mentioned in forums, review sites, aggregators—anything Bing indexes.
Claude: Authoritative positioning matters. Major publications, expert reviews, consistent positive sentiment. Claude's threshold is high but consistent.
Gemini: Google Search rankings translate directly. If you rank well on Google, Gemini will cite you.
The weak cross-platform correlations (0.17-0.39) mean you need different strategies for each.
The Top 3 Summary
| Metric | ChatGPT | Claude | Gemini |
|---|---|---|---|
| Avg visibility | 69.7% | 67.3% | 71.3% |
| Position std dev | 19.5 | 6.5 | 2.2 |
| Positive sentiment | 85 | 124 | 41 |
| Companies at baseline | 91/96 | 85/96 | 74/96 |
Gemini is the most consistent and most generous with visibility. ChatGPT is the most unpredictable. Claude generates the most positive advocacy.
Try It
Want to see how each platform describes your company?
Takes 30 seconds. Shows you exactly what ChatGPT, Claude, and Gemini say about you.
Methodology
Sample: 96 YC-backed companies analyzed January 2026
Platforms:
- ChatGPT (gpt-5-nano + Bing web search)
- Claude (claude-haiku-4-5 + web search)
- Gemini (gemini-2.5-flash + Google Search)
Queries: 16 prompts per company
- 5 category prompts (brand-agnostic: "Best X tools in 2025")
- 5 brand prompts (direct evaluation)
- 5 competitive prompts (comparison queries)
- 1 description prompt
Citation Rate: Percentage of category prompts where brand was mentioned (measures true AI visibility without brand bias)
Position: Average mention position across successful queries (1 = first mentioned)
Correlation: Pearson correlation coefficient between platform citation rates
Related
- GEO Benchmark Report — Full analysis of YC companies
Last updated: January 14, 2026
Stay Updated on AI Visibility
Get weekly insights on GEO, AI traffic trends, and how to optimize for AI search engines.
No spam. Unsubscribe anytime.
Related Articles
The 15 Search Queries Your Competitors Are Not Appearing In (And Why You Should Target Them)
Most companies optimize for 20-40 queries where they already win, missing hundreds of high-value queries where competitors haven't established dominance.
The AI Search Visibility Stack: 7 Essential Components Every Company Needs in 2026
AI search visibility needs 7 components: monitoring, content optimization, entity mapping, audience research, competitive intel, integrations, measurement.
Check Your AI Visibility
See what ChatGPT, Claude, and Perplexity say about your brand. Free, no signup.
Get Free Report