AI search engines amplify household-name brands and favor corporate pages and listicles: top brands appear in 73% of relevant AI answers versus 44% for mid-market and 11% for niche brands. Corporate websites receive about 78% of citations and 'best-of' listicles are the most-cited page type, while sentiment about brands flips far more often than mention/no-mention.

Main Finding

AI search engines produce a reproducible, platform-dependent “share of voice” for brands that strongly favors already‑established, authoritative names. Ranqo’s production measurement across five engines shows a clear three‑tier visibility ladder (household/global brands ≈73% mention rate; mid-market/regional ≈44%; niche/small ≈11%), concentrated citation patterns (≈78% of citations go to corporate websites), and a small set of high‑leverage content formats (notably “best‑of” listicles ≈21% of citations). Mentions are relatively stable and deterministic; sentiment is much noisier (it flips ~6.7× more often than whether a brand is mentioned).

Key Points

• Three‑tier brand‑stature ladder (day‑1 recognition):
  • Tier 1 (global household names): ~73% of relevant AI answers mention them on first run.
  • Tier 2 (established mid‑market/regional): ~44%.
  • Tier 3 (niche/small brands): ~11%.
  • Tiers are separated by roughly 30 percentage points; differences are statistically large.
• Citation composition:
  • ~78% of cited sources are corporate websites (the brand’s own site + third‑party corporate domains).
  • Among non‑corporate sources, YouTube is cited most, followed by Reddit, editorial media, and Wikipedia.
  • “Best‑of” listicles are the most‑cited page format (~21% of all citations), giving single pages outsized leverage.
• Platform behavior and grounding:
  • Engines differ in grounding policies and in which sources they favor; cross‑engine source overlap is low.
  • Gemini, Perplexity, and Grok are configured as always‑grounded; ChatGPT and Claude use a weekly web‑search cooldown per brand.
  • Ranqo weights engines for an aggregate visibility score (example weights: ChatGPT 0.30, Gemini 0.20, Perplexity 0.20, Claude 0.20, Grok 0.10).
• Metrics and dynamics:
  • Ranqo reports three separate primary signals: mention/visibility, average ordinal position when mentioned, and sentiment score (positive/neutral/negative).
  • Mentions are largely deterministic and flat absent intervention; sentiment fluctuates substantially.
• High‑leverage content signals:
  • Page features boosting GEO: machine‑extractable provenance (quotations, stats, citations), clear entity structure, FAQ/structured blocks, author bylines and E‑E‑A‑T markers.
• Practical tooling and protocols:
  • Ranqo’s platform issues controlled prompts across six prompt categories (discovery, problem_solution, comparison, use_case, expert, brand_research), extracts citations and relations, runs a six‑dimension page audit, and produces closed‑loop improvement recommendations.
  • The paper proposes seven v1.1 test protocols (incl. an RCT protocol) to measure causal effects of interventions on AI visibility rather than relying on correlational lifts.

Data & Methods

• Dataset:
  • Production dataset: 100K+ prompt responses across 100+ brands, measured March–May 2026.
  • Reference CRM study: 50 unbranded prompts × 5 engines × 10 runs = 2,500 responses across 10 CRM brands (Jan 2026), yielding 9,600+ brand mentions.
• Prompting and coverage:
  • Prompts are generated automatically per brand and theme; default prompts are unbranded to measure discovery visibility.
  • Six prompt categories: discovery, problem_solution, use_case, comparison, expert, brand_research.
• Engines and grounding:
  • Engines queried: ChatGPT (GPT‑5; weekly search cooldown), Gemini (Gemini‑3; always grounded), Perplexity (Sonar; always grounded), Claude (Claude Sonnet; weekly cooldown), Grok (Grok‑3; always‑on retrieval).
• Signals extracted:
  • Brand mention boolean, ordinal position, sentiment label and score, cited URLs/domains (normalized and typed), source→brand relationship, competitor mentions.
• Page audit (six dimensions, scored 0–100):
  • Crawlability, Content Quality, Page Speed, AI Readiness, Citation Potential, Authority & Trust. GEO‑specific emphasis on AI Readiness, Citation Potential, and Authority & Trust.
• Scoring and aggregation:
  • Visibility = per‑engine mention rate; cross‑engine visibility is a platform‑weighted average.
  • Other metrics: average position when mentioned, sentiment score computed as (positive + 0.5*neutral)/mentions scaled 0–100, share of voice vs qualified competitors.
• Statistical methods:
  • Nonparametric bootstrap (10k resamples) for uncertainty; Kruskal–Wallis omnibus test and pairwise Mann–Whitney U (Bonferroni corrected) for tier comparisons; Cohen’s d for effect size; OLS slopes for per‑engine trends.
• Causal testing plan:
  • The paper proposes v1.1 protocols including randomized controlled tests (Protocol P3/P7 mentioned) that use the platform’s closed loop (recommendation → implementation → re‑audit) to estimate lift; current reported lifts are observational unless run under the RCT protocols.

Implications for AI Economics

• Winner‑take‑more dynamics and incumbent advantage:
  • LLMs and retrieval stacks preferentially cite already‑visible sources (a Matthew effect). That magnifies incumbent brands’ offline and online market power, potentially increasing concentration in many product/service categories.
• Competitive barriers for SMEs and startups:
  • Small brands face a high marginal cost to acquire AI visibility. The observed tier gaps imply diminishing returns from traditional SEO alone; GEO investments (content engineered for quotable provenance, listicle placement, video presence) become necessary inputs to compete.
• Reallocation of marketing budgets and market for GEO services:
  • Firm budgets may shift toward GEO‑specific content production and distribution (e.g., creating quotable research, getting onto high‑value listicles, video content). This creates demand for new intermediaries (GEO agencies, tools like Ranqo) and may raise prices for these services.
• Platform competition and externalities:
  • Cross‑engine variation in citations implies consumers’ exposure to brands depends on which AI they use. This increases platform power over discovery and creates rents for platforms that control retrieval/citation primitives. Regulators may care about how retrieval choices affect competition.
• Information asymmetries and consumer welfare:
  • If AI answers concentrate on a narrow set of corporate sources (including paid or SEO‑optimized listicles), consumer search quality may fall for niche or emergent options. Conversely, high‑quality, well‑audited content could improve answer reliability — the net welfare effect depends on whether citation concentration correlates with true quality.
• Measurability enables market experiments and pricing:
  • Transparent, repeatable GEO measurement (visibility, position, SOV) enables firms to estimate return on GEO spend, supports performance‑based contracting for agencies, and allows economic experiments (RCTs) to identify causal ROI of content interventions.
• Labor and skill demand:
  • Demand shifts toward producing machine‑readable, provenance‑rich content (data, quotes, structured schema, FAQs) and toward expertise in platform‑level strategy (placement on listicles, partnerships with creators). This changes the skill premium in marketing/SEO labor markets.
• Research and policy directions:
  • Empirical questions: elasticities of demand w.r.t. AI visibility; causal effect of GEO on sales/conversions; cross‑platform arbitrage opportunities; long‑run dynamics of brand concentration.
  • Policy concerns: whether platform retrieval designs create anti‑competitive bottlenecks, whether provenance and citation standards should be mandated, and how to ensure discoverability for small firms.
• Practical prescriptions that carry economic logic:
  • High‑leverage, lower‑cost tactics for disadvantaged brands: secure placement on widely‑cited listicles and aggregator pages; produce quotable, structured content with clear provenance; prioritize YouTube and video/content that engines prefer; measure and run controlled experiments to prioritize interventions with the best expected ROI.

Summary: The paper establishes an actionable baseline measurement framework for GEO and demonstrates economically meaningful concentration of AI visibility that advantages large, authoritative brands. For economists and market participants, this implies shifting returns to scale in discovery, rising value for GEO services, testable causal questions about how AI visibility converts to revenue, and potential regulatory interest where AI retrieval materially affects market competition.