The AI visibility field has a hype problem. Vendors promise secret techniques; gurus sell “one weird trick” for ChatGPT rankings. A small business owner deserves better: a sober account of what published research and large-scale studies actually indicate about how AI systems choose which businesses to cite. Here is that account, hedged where honesty requires it.
Finding One: Traditional Rankings Still Correlate — Partially
Multiple analyses have found that a large share of citations in Google’s AI summaries come from pages already ranking well in traditional search. One widely cited study put over 90% of AI Overview citations within the top ten organic results. Interpretation requires care: correlation isn’t a rulebook, and the pattern is weaker on conversational platforms like ChatGPT. But the practical takeaway is solid — foundational SEO isn’t obsolete; it’s the floor, not the ceiling.
Finding Two: Authority Signals Diverge From Old-School SEO
Studies of what content large language models cite have found weak correlation with classic factors like raw backlink counts and keyword optimization, and stronger association with comprehensiveness, clarity, readability, and third-party corroboration. In plain terms: machines appear to favor sources that fully answer the question and are vouched for elsewhere — which is why analysts increasingly emphasize brand authority over traffic tactics.
Finding Three: Behavior Data Shows Answers Replace Visits
User studies of AI-augmented search consistently find shallow engagement with sources: most users read only the top of an AI answer, and external clicks for informational tasks approach zero. The evidence that visibility inside the answer matters more than the click is, at this point, robust.
What the Research Does Not Support
Equally important. No credible study supports guaranteed placement in AI answers — the systems are probabilistic and change frequently. No evidence supports paying for inclusion in organic AI recommendations. And single-tactic promises (“just add this markup”) consistently fail replication. Anyone selling certainty in this field is selling past the evidence.
The Evidence-Based Playbook
What survives scrutiny is unglamorous: consistent entity information, genuinely comprehensive answers to real customer questions, structured data as reinforcement, and accumulating third-party validation. Direction, not magic. Probabilities, not promises.
How to Evaluate Any Vendor Claim
The research literacy that protects you: when any vendor claims a technique improves AI visibility, ask three questions. What data supports this — a controlled study, or three cherry-picked screenshots? What was the sample — a hundred sites, or one client in an easy niche? Has anyone replicated it — or does the finding exist only in the vendor’s own marketing? You needn’t become a scientist; you need only refuse to buy conclusions that can’t survive those questions. In an immature field, epistemics are a purchasing skill. The businesses that avoid the snake oil have more budget left for the boring work the evidence actually supports.
That evidence-first posture is how LeadSupport.net operates. We build small business AI visibility on what studies and our own testing support — and we’ll tell you plainly where the evidence is thin. If you’re tired of hype and want the version of this work that survives scrutiny, contact LeadSupport.net.