Research We Cite

Claim

"SEO in AI search has collapsed 60-80% in the past two years."

Methodology

Two independent measurement programs tracked the same phenomenon across overlapping time windows. Ahrefs measured 863,000 keywords and found that the share of Google top-10 pages also appearing in AI-generated results dropped from approximately 76% in early 2024 to 38% in February 2026, a relative decline of roughly 50%. BrightEdge applied a different methodology across the same period and found the overlap dropped to approximately 17% by February 2026, a relative decline of roughly 78%. The synthesis source (almcorp.com) aggregates both measurements and the October 2025 BrightEdge midpoint reading into a continuous timeline. The 60-80% range in our published hook reflects the floor (Ahrefs: ~50% relative collapse) and the ceiling (BrightEdge: ~78% relative collapse), rounded to the nearest 10.

Attribution

almcorp.com. Google AI Overview Citations Drop From Top Ranking Pages 2026. Early 2024 to February 2026.

Modern AI confidence note

Two methodologies produce directionally consistent findings across the same time window. The Ahrefs measurement is based on 863,000 keywords, a sample size large enough to generalize across categories. The BrightEdge methodology is not fully published; we report its finding as approximate. The range we publish accurately captures both readings. We treat this as high-confidence directionally; the exact percentage is time-bound and will move as AI engines continue to update.

Claim

"In early 2024, around 76% of pages ranking in Google's top 10 were also cited by AI engines. By February 2026, that overlap had dropped to between 17% and 38%."

Methodology

Same sources as above. The 76% baseline is the Ahrefs early-2024 measurement. The 17% endpoint is BrightEdge's February 2026 reading. The 38% endpoint is Ahrefs' February 2026 reading across 863,000 keywords.

Attribution

almcorp.com as synthesis source for both Ahrefs and BrightEdge datasets.

Modern AI confidence note

We present both methodological endpoints together rather than choosing one, because they are both credible and the spread between them is itself informative. Presenting only the lower number (17%) would be more dramatic but less representative. Presenting only the higher number (38%) would understate the BrightEdge signal. The honest representation is the range, with the source of each bound named.

Claim

"Pages ranked around position 5 in organic search experienced a 115% AI visibility increase after GEO optimization. Pages already at position 1 saw minimal change."

Methodology

Princeton University, IIT Delhi, Georgia Tech, and the Allen Institute for AI published a benchmark study presented at KDD 2024. The study evaluated generative engine optimization (GEO) across 10,000 queries and measured changes in AI citation visibility as a function of organic search position and content modification strategy.

Attribution

Princeton University GEO paper, presented at KDD 2024. Collaborative research across Princeton, IIT Delhi, Georgia Tech, and Allen Institute.

Modern AI confidence note

Peer-reviewed academic study with named institutional provenance and conference presentation (KDD 2024, a top-tier data mining venue). The 115% figure for position-5 pages is specific enough to trace. We cross-referenced against the HOTH's published summary and found consistent reporting. We apply this finding to illustrate that SEO dominance does not transfer to AI visibility dominance, which is the diagnostic premise of Modern Discovery.

Claim

"89% of LLM citations come from earned sources."

Methodology

Onely analyzed which source types (brand-owned, earned media, forums, review platforms, aggregators) appear as cited sources in LLM responses across a defined query set.

Attribution

Onely. GEO vs. SEO. Published 2025-2026.

Modern AI confidence note

Onely is an established technical SEO research firm with a documented methodology. We apply the 89% figure carefully, noting that 'earned source' includes a range of content types and the threshold depends on how queries are categorized. We do not present this as meaning brand-owned content has no role; we use it to support the finding that third-party source coverage matters for AI citation.

Claim

"Third-party community discussions generate 4x higher AI citation rates compared to brand-owned pages."

Methodology

Evertune tracked AI citation rates by source type across a study of brand visibility in AI-generated responses.

Attribution

Evertune. AI visibility research.

Modern AI confidence note

We apply this as a directional signal, not a precise universal multiplier. The 4x figure is Evertune's finding for their study set. Different categories and query types will produce different ratios. We cite it to illustrate that community sources (Reddit, forums, industry publications) carry weight in AI citation logic that differs from Google's ranking signals.

Claim

"Only 45% of brands in Google's top 3 appear in AI results for the same query."

Methodology

sellm.io analyzed the overlap between Google top-3 organic rankings and AI-generated results for a defined set of brand and category queries.

Attribution

sellm.io. AI Search vs. Google Brand Rankings.

Modern AI confidence note

Consistent with the broader overlap collapse documented in the SEO Collapse section. We apply it specifically when making the point that organic ranking position is no longer a reliable predictor of AI visibility. We acknowledge that the sellm.io study's specific query set is narrower than the Ahrefs 863,000-keyword study.

Claim

"38% of AI-cited brands do not appear in Google's top 10."

Methodology

Same sellm.io analysis, looking at the inverse relationship: brands that appear in AI results without organic search ranking.

Attribution

sellm.io. Same source as above.

Modern AI confidence note

This finding supports the claim that AI search is producing a distinct winner set from Google search, and that brands without organic ranking still have a path to AI visibility. We apply it carefully to avoid implying that SEO investment is harmful; the correct reading is that AI visibility requires its own measurement infrastructure.

Claim

"ChatGPT, Claude, Perplexity, and Gemini each have distinct citation mechanisms, training cutoffs, and audience characteristics. A brand's AI visibility varies significantly across platforms."

Methodology

soar.sh analyzed platform-specific citation patterns, audience composition, and brand visibility behavior across the four major AI search platforms.

Attribution

soar.sh. Platform-specific AI search citation and visibility analysis.

Modern AI confidence note

We apply this finding to support the design premise of Modern Discovery: a brand cannot collapse AI visibility into a single score without losing diagnostic value. Platform-level visibility measurement is necessary because each platform operates differently. We acknowledge this area is evolving rapidly and that platform citation behavior can shift with model updates, which is one reason ongoing measurement matters.

Claim

"AI users phrase queries in natural language that keyword-optimized pages were never built to answer. A consumer asking 'I need a TV that displays artwork, slim, 65 to 75 inches, under $2,000, that looks good when off' activates a retrieval pattern no keyword-SEO strategy anticipated."

Methodology

LinkedIn research post by Swati Gole documenting the shift in consumer query behavior toward conversational, multi-constraint natural language in AI search engines.

Attribution

LinkedIn / Swati Gole. Conversational query research post.

Modern AI confidence note

We treat this as a directional and illustrative source, not a quantitative benchmark. The specific TV example is a useful illustration of the pattern rather than a measured statistic. We apply it to explain why brands need AI-specific measurement alongside traditional SEO tracking.

Claim

"Conversational intent patterns in AI search activate different content than short-form keyword queries. Brands optimized for traditional keyword density are often invisible to AI retrieval on natural-language queries."

Methodology

dabaran.com research on AI-driven keyword research and the gap between conversational intent and keyword-optimized content.

Attribution

dabaran.com. AI-driven keyword research and conversational intent patterns.

Modern AI confidence note

Consistent with the broader finding from the Princeton KDD 2024 GEO study that position-1 SEO performance does not predict AI visibility performance. We apply this as supporting context for the need for AI-specific measurement, not as a standalone quantitative finding.