Do AI search engines favor well-known brands over others?

Ranqo, a GEO (generative engine optimization) measurement platform, reports that on a brand's very first visibility-tracking run, AI search engines mention it at a rate set almost entirely by brand stature rather than anything the brand has done for AI search specifically. Analyzing "100K+ prompt responses across 100+ brands tracked on Ranqo between March and May 2026," the company finds "global household names such as Stripe and Nike appear in 73% of relevant AI answers on their first tracking run; established mid-market and regional brands such as Olipop and Klaviyo appear in 44%; niche and small brands in just 11%" — each tier roughly 30 percentage points below the one above it.

Ranqo attributes the ladder to how the five engines it tracks (it names ChatGPT, Claude, Perplexity, Gemini, and Grok) select sources: citations concentrate on a small set of outlets that are already heavily cited elsewhere, a "Matthew effect" the paper attributes to prior work by Algaba et al. (2025). Because Ranqo's starter prompt set for a new brand is deliberately unbranded — discovery, problem/solution, use-case, and expert prompts that never name the brand — the mention rate it measures is specifically whether an engine surfaces a brand nobody asked for by name, which the authors call "the hard, GEO-relevant case." Ranqo is explicit that this is a baseline reading of production telemetry, not a treatment effect: its own no-intervention trajectory finding (§6.4) shows unbranded visibility holding flat or drifting down over the tracking window, so the ladder describes where a brand starts, not what moves it.

The finding sits alongside other recent measurements of how unstable or skewed AI answers about brands and sources can be. How consistent are AI brand recommendation lists across repeated prompts? measures volatility within a single query's answer set across repeated runs; Ranqo's ladder instead measures a different axis — whether a brand appears at all on its first tracked run — and finds that axis far more deterministic, stratified cleanly by brand stature rather than fluctuating. The wider shift from search rankings to AI answers that gives Ranqo's ladder its stakes is the same one behind Will AI Overviews reduce search referral traffic to publishers?, where publishers brace for lost referral traffic as the same AI engines substitute for blue-link search; Ranqo's paper is the brand-visibility mirror of that shift, measuring who gets named instead of who gets clicked.

Ranqo names its own limits plainly: the cohort is convenience-sampled and skews toward SaaS, retail-execution, fintech, and Indian D2C brands, the Tier-1 cell is only 11 brands, and the tier assignment itself is hand-coded from public stature signals (Wikipedia, press, funding) rather than independently validated — so, as the authors put it, "the ladder should be read as quantifying the size of an expected effect, not as independent proof that prominence drives citation." The paper also states directly that it reports no randomized controlled trial of its own recommendation engine: whether a brand can climb tiers through deliberate action is left to a proposed v1.1 protocol, not demonstrated here. The excerpt therefore supports a baseline fact about AI search today — that visibility on first contact tracks brand stature almost three-to-one between tiers — but not yet a claim about what any brand can do to change its tier.