2163072013 alex@ortheon.com

Research

Controlled Experiments in AI Search Visibility

AI Search Authority is not built on theory. It is built on structured experimentation.

We test how Large Language Models interpret, compress, and surface information. Below are selected high-level experiments from our internal research lab. 

Each experiment isolates a structural variable that influences AI inclusion.

Experiment 01

How LLMs Select Definitions

Determine how AI systems choose which brand definition to surface when multiple competitors describe the same concept.

Hypothesis

LLMs prioritize definitions that are:

  • Concise
  • Structurally formatted
  • Repeated across surfaces
  • Reinforced with schema

Method

We analyzed multiple competing brand definitions within the same industry. Variables tested:

  • Definition length (50–120 words)
  • Presence of structured “What Is” blocks
  • FAQ schema implementation
  • Cross-surface consistency
  • Entity clarity signals

Prompts were run across multiple AI interfaces and tracked for inclusion frequency.

Findings

  • Explicit “What Is” definition blocks increased inclusion probability.
  • Definitions under 90 words were cited more frequently than long-form narrative descriptions.
  • Schema-supported definitions had stronger compression survival.
  • Inconsistent brand positioning reduced citation frequency.

Conclusion

LLMs favor structured certainty. Brands that clearly define themselves, in consistent, extractable formats are surfaced more frequently. Definition clarity directly increases citation probability.

Experiment 02

Structured Summaries vs Narrative Content

Evaluate whether LLMs prefer structured formatting over long narrative explanations when generating answers.

Hypothesis

Content with hierarchical structure and bullet compression is more likely to survive AI summarization.

Method

We created two versions of identical topical content:

Version A:

  • Long-form narrative paragraphs
  • No structured summaries
  • Minimal headings

Version B:

  • Definition blocks
  • Bulleted summaries
  • Clear H2/H3 hierarchy
  • Structured FAQ reinforcement

Both versions were indexed and tested across repeated AI prompts.

Findings

  • Structured versions were cited significantly more often.
  • Bullet compression increased extractability.
  • Clear subheading structure improved inclusion stability.
  • Narrative-only formats were more likely to be summarized without brand attribution.

Conclusion

Formatting influences visibility. LLMs extract structure before detail. Content engineered for extraction increases citation consistency.

Experiment 03

Topic Depth vs Backlink Volume

Assess whether backlink authority outweighs topic coverage depth in AI inclusion.

Hypothesis

Semantic depth and entity reinforcement influence AI inclusion more than backlink count alone.

Method

We compared brands with:

  • High backlink profiles but shallow topic clusters
  • Moderate backlink profiles but deep semantic coverage

Variables measured:

  • AI answer inclusion frequency
  • Co-occurrence density
  • Citation stability across prompts

Findings

  • Deep topic clusters increased AI inclusion frequency.
  • Backlinks influenced rankings but did not guarantee AI citation.
  • Brands with comprehensive topic mapping were surfaced more often, even with fewer backlinks.

Conclusion

AI systems evaluate structural knowledge dominance, not just link authority. Backlinks alone do not secure citation. Semantic depth drives compression survival.

Experiment 04

Entity Density and Citation Frequency

Measure how repeated, consistent entity reinforcement impacts AI citation patterns.

Hypothesis

Higher entity density across structured content increases citation probability.

Method

We measured:

  • Frequency of entity mentions within topic clusters
  • Structured reinforcement via schema
  • Cross-surface co-occurrence signals
  • Definition consistency

AI prompts were tested for:

  • Inclusion frequency
  • Competitive displacement
  • Brand recall stability

Findings

  • Controlled entity density increased citation frequency.
  • Over-optimization reduced clarity and weakened inclusion.
  • Reinforced entity-to-topic mapping improved stability across query variations.

Conclusion

Entity reinforcement must be precise, not excessive. Strategic density strengthens AI confidence. Ambiguity weakens inclusion.

Experiment 05

AI Overview Source Patterns

Analyze which types of sources are most frequently referenced inside AI Overviews and synthesized search results.

Hypothesis

AI systems prefer sources with structured authority, not just high traffic.

Method

We analyzed AI Overview responses across multiple industries, categorizing cited sources by:

  • Domain authority
  • Structured formatting
  • Topic depth
  • Schema implementation
  • Entity clarity

Patterns were tracked across recurring query types.

Findings

AI Overviews frequently referenced:

  • Structured definition pages
  • Deep topical guides
  • Brands with consistent entity positioning
  • Content with clean formatting and extractable summaries

High-traffic domains without structured clarity were less stable in citation frequency.

Conclusion

AI selection patterns favor structural authority over popularity alone. Surface inclusion reflects clarity, not volume.

Research Implication

Across experiments, a consistent pattern emerges:

AI systems reward:

  • Definition clarity
  • Structural formatting
  • Semantic depth
  • Entity reinforcement
  • Cross-surface consistency

They deprioritize:

  • Thin content
  • Narrative-only formats
  • Backlink dependence
  • Ambiguous positioning

AI visibility is not accidental. It is engineered.

Ongoing Research

AI models evolve. We continuously test:

  • Inclusion frequency shifts
  • Citation stability changes
  • Emerging compression behaviors
  • Multi-surface reinforcement impact

AI Search Authority operates as a lab, not a trend follower. We document what works before others recognize the shift.