Investigation Report // 09
Does SEO Still Work? An Empirical Investigation into the 90% GEO Overlap
Analyzing the structural dependency of Generative Engine Optimization (GEO) on traditional organic search signals.
Introduction
The perennial question "Is SEO dead?" has resurfaced with renewed vigor in the age of Large Language Models (LLMs) and Generative Search. However, the data suggests a more nuanced reality: SEO isn't dying; it is being subsumed into a broader framework of Generative Engine Optimization (GEO).
Industry leaders, including Wil Reynolds of Seer Interactive, have posited that approximately 90% of GEO is effectively traditional SEO. This investigation seeks to measure that claim empirically, analyzing the correlation between top-tier organic rankings and citations in AI search environments like Google AI Overviews (AIO), SearchGPT, and Perplexity.
Core Thesis: While AI search reduces friction for users, it creates a "Verification Gap" for brands. SEO fundamentals—authority, technical health, and content relevance—remain the primary signals LLMs use to verify and trust information before citing it in a generative response.
Key Takeaways
- 61% CTR Collapse: Organic click-through rates for queries with Google AI Overviews plummeted from 1.76% to 0.61% between 2024 and 2025.
- 93.67% AIO-SERP Correlation: Nearly all Google AI Overview citations link to at least one domain ranking in the top 10 organic results.
- 87% SearchGPT-Bing Overlap: SearchGPT citations match Bing’s top organic results 87% of the time, confirming a heavy reliance on Microsoft's index.
- The 12% Parametric Gap: Only 12% of URLs cited by pure parametric LLMs (non-retrieval) rank in Google's top ten, highlighting the difference between training data and live RAG.
The Data Behind this Study
This report synthesizes data from three primary longitudinal studies conducted between late 2024 and mid-2026:
- Seer Interactive (2025-2026): An analysis of 500+ SearchGPT citations and a separate study on 80 million ChatGPT queries via SEMrush data.
- Search Atlas (2025-2026): A comparative analysis of 5.5 million responses across Perplexity, Gemini, and OpenAI to measure citation behavior and domain overlap.
- The Digital Bloom / Ahrefs (2025): A study of AI Overview (AIO) citation patterns and their correlation with traditional SERP rankings.
Definitions
- Retrieval-Augmented Generation (RAG): A search architecture where an LLM fetches live external data to ground its response, rather than relying solely on pre-trained weights.
- Generative Engine Optimization (GEO): The practice of optimizing content so that it is retrieved, trusted, and cited by generative AI systems.
- Citation Density: The number of unique external sources cited per character of generated text.
- Parametric Model: An LLM answering from its internal training data (parameters) without a live web search.
Findings
1. The Impact of AI Overviews on Organic Traffic
The introduction of AI Overviews has fundamentally altered the economics of organic search. Data from Seer Interactive shows a catastrophic decline in CTR for informational queries.
| Metric | Pre-AIO (June 2024) | Post-AIO (Sept 2025) | Change |
|---|---|---|---|
| Organic CTR (Informational) | 1.76% | 0.61% | -65.3% |
| Average Organic Clicks | Baseline | -40% to -58% | Significant Loss |
2. Citation Overlap: Traditional SEO vs. AI Search
Despite the traffic loss, the selection of sources remains tethered to traditional search rankings. The correlation is strongest in systems with integrated search indices (Google AIO and SearchGPT).
| Platform | Source Index | Overlap with Top 10 Results |
|---|---|---|
| Google AI Overviews | Google Search | 93.67% |
| SearchGPT | Bing Search | 87.00% |
| Perplexity (Sonar) | Multi-Index RAG | ~60-65% (Shared) |
| Pure LLM (Parametric) | Training Data | 12.00% |
The Search Atlas study highlights that Perplexity achieves a citation density 2-3x higher than parametric models, prioritizing source attribution as a core architectural feature.
What this means for your brand
The data confirms that SEO still works, but its objective has shifted. It is no longer enough to rank; you must be verifiable.
- Diversify Beyond Google: With SearchGPT's 87% reliance on Bing, optimizing for Microsoft's index is no longer optional for AI visibility.
- Optimize for "Trust Signals": LLMs verify information across third-party sources. If your brand's claims aren't corroborated by independent aggregators, news sites, or directories, the LLM will likely omit you in favor of a "verified" competitor.
- Focus on Authority over Volume: In a world where AI can generate infinite content, "Information Gain" and primary-source evidence are the only non-fungible assets.
- Prepare for the Frictionless Search: Users are talking to machines. Your technical SEO must ensure that machines can parse your brand's capabilities as easily as humans can.
For a deeper dive into how machines resolve brand identity, see our briefing on Entity Resolution.