The Commodity Test: What Predicts AI Citations?

What is the commodity test and how does it work?

The commodity test is a binary evaluation framework used by measurement teams to determine if a piece of text contains non-derivable information. When an LLM evaluates a query, it first assesses whether it can construct a confident answer from its parametric memory. If your page simply restates widely available consensus without adding proprietary research, structural models, or first-party data, it is flagged as a commodity. Content failing this filter is bypassed in favor of sources offering fresh utility. To pass, every 200–400 word section must introduce specific data points, proprietary workflows, or unique analytical observations that cannot be synthesized via general web scraping. Our tracking data shows that ~84% of AI citations trace to third-party sources that successfully pass this exact differentiation filter, proving that vanilla summaries no longer earn visibility in generative engines.

Why does classic SEO rank fail to predict AI citations?

Classic search ranking measures link equity and keyword relevance, but these metrics correlate poorly with generative engine inclusion. Traditional ranking algorithms reward domain authority and keyword density, whereas AI systems evaluate informational density and factual uniqueness. Empirical analysis reveals that a top-10 classic rank predicts only ~38% of AI citations. An enterprise can hold position one for a high-volume head term yet receive zero visibility in the accompanying AI Overview if the target page fails the commodity test. Generative engines prioritize distinct insights and concise, self-contained modular answers over generalized long-form prose. Furthermore, keyword density above ~1.5–2% actively works against you, as it signals low-effort optimization rather than high-value information. Measurement teams must decouple traditional rank tracking from generative visibility metrics to accurately diagnose performance.

Where do AI models pull their citation snippets from?

Understanding citation placement mechanics requires analyzing on-page architecture and content layout. Generative models do not read pages uniformly; our research demonstrates that ~55% of AI citations pull directly from the top 30% of the page. If your unique data points, definitions, or proprietary frameworks are buried below three thousand words of introductory fluff, the retrieval system often misses them entirely. Self-contained 200–400 word sections function as the fundamental unit of citation, meaning each distinct subtopic must act as an independent information asset. Additionally, technical performance heavily influences retrieval success. Core Web Vitals targets—such as LCP under 1.8s, INP under 150ms, CLS under 0.05, and TTFB under 400ms—ensure that crawler requests and browser control mechanisms render content reliably. When rendering fails, citation opportunities vanish instantly.

How do you audit your content for the commodity test?

Auditing your digital assets for the commodity test requires a systematic review of informational uniqueness and crawler accessibility. Begin by verifying that your site permits all major discovery systems; CRO9 checks 6 AI crawlers including GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot, CCBot, and Google-Extended against a browser control to ensure blocked pages and empty renders are accurately diagnosed. Next, analyze your prose structure against the self-contained 200–400 word section rule. Strip away generic definitions that any foundational model can generate natively. Inject primary research metrics, field notes, or proprietary frameworks that require real-world execution. Because ~88% of AI Overview appearances occur on informational-intent queries, your content must satisfy the user's specific context immediately with verifiable evidence rather than speculative commentary.

How does AI citation churn impact long-term measurement?

Generative search results exhibit extreme volatility compared to classic blue-link rankings, requiring adaptive measurement frameworks. Our longitudinal tracking indicates that 40–60% of AI citations churn monthly as models update their retrieval indexes and test alternative synthesis pathways. This rapid fluctuation means static monthly reporting is obsolete; teams need high-frequency telemetry to capture algorithmic shifts. Interestingly, establishing brand authority provides a crucial stabilizing effect. Branded queries earn ~+18% higher CTR under AI Overviews because users exhibit stronger trust signals toward recognized entities within generative panels. By combining rigorous commodity testing with lightweight, event-driven tracking scripts, search marketers can build resilient content strategies that withstand monthly algorithmic churn.