Primary Research · AI Visibility Study
Where AI Platforms Actually Source Their Answers
A primary research study analyzing 302 AI prompts across ChatGPT, Perplexity, and Google AI Overviews to map citation source patterns by industry and buyer decision complexity.
Central Finding
The brands winning AI citations are not the ones with the best content. They are the ones with presence across the most trusted platforms in their category.
Domain authority, word count, publication frequency, and backlink counts showed weak predictive value for AI citation likelihood across all four verticals studied. Source diversity, platform authority, and community engagement signals explained significantly more of the variance.
The Industry Citation Map
Not all industries are cited the same way. The table below shows where AI platforms pulled citations for each vertical, scored by concentration intensity across 302 prompts. Industries with complex buyer decisions showed markedly different patterns than those with simpler trust models.
| Industry | Brand Site | Review Platforms | Forums & Community | Gov / Credentialed | |
|---|---|---|---|---|---|
| SaaS | Minimal | High | High | Medium | — |
| Insurance & Real Estate | Medium | High | Low | High | Low |
| Finance | Medium | Medium | Low | Low | High |
| Healthcare & Legal | Low | Minimal | — | Minimal | Very High |
Concentration levels are directional based on citation frequency across 302 prompts. Industries with complex buyer decisions — SaaS, financial services, insurance and real estate — showed the highest variation from what traditional SEO authority metrics would predict.
What the Data Showed by Industry
Brand sites were nearly invisible
- Official company sites appeared rarely in AI-generated answers, even for branded queries
- Reddit, G2, and review aggregators drove the majority of citations
- Community consensus and peer validation carried more weight than official product content
- AI platforms treated review aggregators as more authoritative than brand documentation
Community filled institutional gaps
- Long consideration cycles created space for community validation that institutional content left open
- Reddit threads appeared consistently where official sources gave vague or marketing-heavy answers
- Forum discussions about real experiences outranked official explainers in citation frequency
- Trust complexity pushed buyers toward peer validation before they engaged any brand
Two citation tracks co-existed
- Institutional sources carried authority for regulatory and product questions
- Reddit personal finance threads appeared consistently alongside traditional financial sites
- The co-existence reflected the dual nature of financial decision-making: official for facts, community for validation
- Neither track dominated exclusively — both carried citation weight
Authority hierarchy was rigid
- Citations clustered heavily around government portals, credentialed organizations, and peer-reviewed content
- Community influence was minimal compared to every other vertical studied
- Regulatory and credentialing requirements created a trust bar that community content rarely cleared
- Even high-engagement Reddit threads showed low citation frequency for medical and legal queries
How the Study Was Conducted
302 prompts were written to simulate buyer research behavior at different stages of the decision process, then run across three major AI platforms to track citation source patterns.
Platforms Tested
ChatGPT (GPT-4), Perplexity AI, and Google AI Overviews. Selected to represent the three dominant AI platforms receiving significant user query volume in 2026.
Prompt Construction
Queries simulated real buyer research behavior. Four prompt types covered the decision journey:
Citation Tracking
Each AI response was scored for citation source, platform origin, content type, and authority signals present. Citations were categorized by platform type, not individual domain.
Vertical Selection
Four verticals were chosen to represent a spectrum of buyer decision complexity and institutional trust requirements, from low-trust-bar (SaaS) to high-trust-bar (healthcare, legal).
Study Limitations
- Findings represent a point-in-time snapshot. AI citation behavior evolves with model updates and training cycles.
- 302 prompts provide directional findings, not statistically definitive conclusions. Patterns are consistent but not exhaustive.
- Citation behavior varies by specific query phrasing. Small prompt variations can produce different source selections.
- Industry verticals are broad categories. Sub-vertical behavior may differ from the patterns reported here.
What This Means for Visibility Strategy
Your citation strategy should match your industry's trust model
SaaS companies need community presence and review platform authority. Healthcare and legal need credentialed content and institutional signals. The same playbook does not apply across verticals. Knowing where your industry falls on the trust complexity spectrum determines the right investment, not a one-size-fits-all checklist.
Source diversity is the most predictive variable for AI visibility
Domain authority, backlink counts, and content volume showed weak predictive value across all four verticals. Source diversity — appearing across multiple trusted platform types — explained significantly more variance. This represents a fundamental shift from how organic search authority has been built for the last decade.
Industries with complex buyer decisions see the highest community citation activity
The simpler the trust model, the less urgency to diversify beyond your own site. For SaaS, insurance, real estate, and financial services, community platform presence is not optional for AI visibility. It is the primary citation surface.
The measurement gap is where the real competitive advantage lives
AI-referred traffic converts at 4.4x the rate of traditional organic search. Most marketing teams have zero visibility into whether AI is citing their brand or a competitor's. The brands that solve this measurement problem first will have a durable advantage over those still optimizing for last-click attribution models.
The measurement framework is the next step
Understanding where citations originate is the first problem. Connecting that visibility to revenue is the one nobody has solved yet. That is what the Residual Attribution framework addresses.
Read the Residual Attribution Framework →