Research
Research
Published research from AI+Automation. This is where our services come from , every recommendation we make to agencies is grounded in these findings.
Key Findings for Agencies
- 6.8% URL overlap between Google top results and ChatGPT citations , ranking on Google does not mean getting cited by AI.
- Position-controlled citation predictors from 10,293 pages across 250 queries (Experiment M, April 2026). Within position bands, content features (AUC=0.673) match domain identity (AUC=0.687). Top actionable signals: comparison structure (d=0.43), query-term coverage (d=0.42), H3 subheadings, word count (~2,000), statistics density, and absence of blog tone. Page speed confirmed irrelevant (p>0.39 in all bands).
- Query intent varies dramatically by vertical , Agency/Law is 66% Discovery intent, SaaS is 87% Informational. Content strategy must match.
- Reddit shapes AI recommendations through training data, not citations , community consensus predicts brand mentions.
How AI Platforms Search
April 2026 | Preprint v1.3 | DOI: 10.5281/zenodo.19554329
1,323 fan-out queries classified across ChatGPT, Gemini, and Perplexity. Two-layer retrieval model: Layer 1 (search decision) is highly deterministic (Gemini 98.9%, ChatGPT 91.7%). Layer 2 (query strings) shows canonical cores covering 59-76% of fan-out events across platforms. Fan-out types stay stable 55-65% across runs.
Read more →I Rank on Page 1 -- What Gets Me Cited by AI?
April 2026 | Preprint v1 | DOI: 10.5281/zenodo.19398158
Position-controlled analysis of 10,293 pages across 250 queries. Content features match domain identity within the SERP. Top predictors: comparison structure, query-term coverage, subheading depth, statistics density. Replication data publicly available.
Read more →Reddit Doesn’t Get Cited (Through the API)
February 2026 | Preprint v3
Training data influence, access-channel divergence, and the shadow corpus in AI brand recommendations.
Read more →Query Intent, Not Google Rank: What Best Predicts AI Citation Behavior
March 2026 | Preprint v3
Empirical investigation of AI citation behavior. Corrected to 3,251 real websites (UGC excluded) across ChatGPT and Google AI Mode. 8 confirmed predictors across 3 independent models. Methodology correction published.
Read more →Cognitive Redistribution, Not Cognitive Decline
February 2026 | Preprint
A critical review of studies linking AI chatbot use to diminished cognition.
Read more →Apply These Findings to Your Clients
Our weekly intelligence report turns this research into actionable optimization, one week at a time. Read a sample, or send your URL for a Free Video Audit.