The Experiment That Should Change How You Think About GEO
A small but significant AI visibility study by EchoWi tested closely related queries across ChatGPT, Gemini, Google AI Overview and Google AI Mode. The results were striking.
When researchers asked a buying-intent question - "What is the best email platform for ecommerce?" - ChatGPT cited eight external domains and Gemini cited five. When they asked the definitional equivalent - "What is email marketing automation?" - neither model cited a single external source.
Across the wider test, ChatGPT cited a source in only one of six definitional query runs but in all nine buying-query runs.
Important caveat: This is a small study produced by an AI-visibility vendor with a commercial interest in the finding. It should be treated as a hypothesis-generating experiment, not evidence of a universal ranking rule. Integrated.Social is planning to reproduce this independently across 50-100 B2B buying queries.
Why AI Systems May Not Need to Retrieve Definitions
The finding makes intuitive sense once you understand how large language models work.
A model trained on billions of documents already has a strong internal representation of "what email marketing automation is." It does not need to retrieve an external source to answer that question - it can answer from memory.
A buying question is fundamentally different. "Which CRM is best for a 100-person B2B SaaS company?" requires:
- Current vendor information (pricing, features, recent updates)
- Third-party comparisons and reviews
- Use-case-specific evidence
- Purchase context and buyer objections
The model has strong incentives to retrieve external evidence because its internal knowledge may be incomplete, outdated or insufficiently specific. That retrieval creates the citation opportunity.
The Distinction That Matters for GEO Strategy
The most AI-readable content may not be the most AI-citable content.
Generic informational content - "What is X?", "How does Y work?", "What are the benefits of Z?" - can train or inform the AI ecosystem without earning attribution. The model absorbs it and answers from memory.
Commercial decision content gives the model a reason to retrieve evidence:
- Comparisons: "X vs Y for [specific use case]"
- Alternatives: "Best alternatives to [market leader] for [ICP]"
- Proof: "[Claim] with verified data and methodology"
- Use cases: "[Product] for [specific company type/size/industry]"
- Buyer objections: "Why [common concern] is or isn't a real problem"
If this pattern holds at scale, the GEO priority becomes less "answer every possible definition" and more "own the evidence needed to justify a decision."
What This Means for Your Content Portfolio
Traditional SEO content strategy was heavily weighted toward informational queries because they had high search volume and were easier to rank for. GEO may require a different weighting.
Consider auditing your existing content against three categories:
Category 1 - Model memory: Definitional and explanatory content. Low citation probability. Still valuable for brand awareness and training data, but should not be the primary GEO investment.
Category 2 - Retrieval-triggering: Comparison, alternatives, proof, use cases and buyer objections. High citation probability. Should be the primary GEO investment.
Category 3 - Hybrid: Content that starts with a definition but pivots quickly to commercial evidence. Medium citation probability. Useful for capturing both informational and commercial queries.
For B2B brands, the highest-value GEO content likely sits in Category 2: specific, evidence-rich, decision-stage content that an AI system cannot answer reliably from memory alone.
A Query Portfolio Framework for AEO/GEO
Based on the EchoWi finding and supporting data from Rankmax (53.5% of commercial prompts trigger a live web search in ChatGPT vs 18.7% for informational queries), here is a practical framework for building a citation-optimised content portfolio:
Tier 1 - Decision evidence (highest citation priority):
- "[Product category] for [specific ICP]"
- "[Your brand] vs [competitor] for [use case]"
- "Best [category] for [company size/industry/budget]"
- "[Claim] - the data behind it"
Tier 2 - Consideration content (medium citation priority):
- "How [category] works in practice for [ICP]"
- "What to look for in [category] - a buyer's checklist"
- "[Common objection] - is it a real concern?"
Tier 3 - Awareness content (low citation priority, still valuable):
- "What is [category]?"
- "How does [technology] work?"
- "[Category] explained"
The insight is not that Tier 3 content is worthless. It is that it should not dominate your GEO investment if citation and recommendation are the primary goals.
The Integrated.Social Research Programme
We are planning to reproduce this experiment at scale across 50-100 commercially meaningful B2B buying queries, tested across ChatGPT, Gemini, Google AI Mode, Google AI Overviews and Perplexity.
The methodology will classify queries by intent type (informational, consideration, purchase), measure retrieval rate, citation count, source type and brand inclusion, and track results over time as models update.
If the EchoWi pattern survives at scale, the strategic implication is significant:
Traditional SEO optimised content around search volume. GEO may need to optimise content around retrieval necessity.
That is a fundamentally different content strategy - and one that favours brands willing to invest in specific, evidence-rich, decision-stage content rather than high-volume generic definitions.
Modi Elnadi is the founder of Integrated.Social, a B2B AI marketing agency in London specialising in AEO/GEO, agentic AI lead generation and performance marketing. Follow the Integrated.Social GEO research programme [blocked] for updates as we publish results.
