Your Brand Is Invisible to AI Search, and the Gap Is Widening Fast

The SEO playbook you spent years building may now be worth less than a single “best-of” listicle you don’t control. A new large-scale study puts hard numbers on who AI search engines surface — and the answer for most startups is: not you.

What happened

Researcher Pratyush Kumar analyzed 100K+ prompt responses across 100+ brands tracked on the Ranqo platform between March and May 2026, building the first large-scale baseline for GEO (Generative Engine Optimization) across ChatGPT, Claude, Perplexity, and Gemini. The findings reveal a brutal three-tier hierarchy: global household names (Stripe, Nike) appear in 73% of relevant AI answers; established mid-market brands (Olipop, Klaviyo) in 44%; niche and small brands in just 11% — a roughly 30-percentage-point drop per tier. When engines cite sources, 78% of citations flow to corporate websites, but among third-party sources, YouTube leads, beating Reddit, editorial media, and Wikipedia. The single highest-leverage content format is the ranked “best-of” listicle, accounting for ~21% of all citations — a format you almost certainly don’t own. Most alarming for brand managers: sentiment is wildly unstable, flipping positive-to-negative 6.7× more often than a brand’s mention status changes at all, meaning your share of voice in AI can look fine while the framing quietly turns hostile.

Cold read

This is descriptive, not causal. The paper measures correlation between brand stature and AI visibility but explicitly defers causal tests to a proposed v1.1 protocol — meaning we don’t yet know whether any specific action actually moves the needle. The 100+ brands tracked are all on a single commercial platform (Ranqo), which introduces selection bias: these are brands already paying attention to GEO, not a representative sample of the startup universe. The March–May 2026 window is three months of data across engines that update their underlying models and retrieval behaviors constantly; what’s true in May 2026 may be meaningless by Q4. And the 11% visibility figure for small brands likely understates the real floor — brands not even monitored on a platform like Ranqo probably hover near zero. The paper acknowledges these limits, which is honest, but the gap between “we can measure this” and “here’s how to fix it” remains wide open.

What it means for you

• Signal maturity: 2/5 — Solid measurement baseline, zero proven interventions yet
• Who gets hurt: D2C founders and early-stage B2B startups who depended on long-tail SEO traffic that is now being eaten by zero-click answers before the user ever reaches their site
• What breaks if this is true: The entire content marketing flywheel — blog posts, owned media, SEO-optimized landing pages — may generate authority for AI engines to cite other sources rather than you, particularly the “best-of” listicles written by third parties
• Why it might not land: The 30-point-per-tier gap may largely reflect brand recognition baked into LLM training data, not something addressable through any near-term content or PR strategy
• Watch for: The v1.1 causal protocols the paper proposes — if those publish with statistically significant lift numbers for specific interventions, that is the moment this field graduates from measurement to playbook

Forecast as of 2026-06-21

By Q2 2027, at least one major GEO/AEO vendor will publish a controlled study claiming a specific tactic (structured schema, third-party review seeding, or YouTube presence) produces measurable visibility lift for sub-50-employee brands — but the effect size will be under 10 percentage points, and the SEO-to-GEO visibility gap between small and large brands will remain above 40 points.

Source: Generative Engine Optimization at Scale: Measuring Brand Visibility Across AI Search Engines — Pratyush Kumar. https://arxiv.org/abs/2606.20065v1