New research on how AI search engines actually retrieve and rank content describes something specific: broad, generalist content loses to narrow content built to answer one question completely. I didn’t build Small Factory 5 around that finding. The finding just describes something already built here.
The Old Funnel, and Why It’s Now Upside Down
Classic SEO logic builds broad first. Broad keyword pages capture volume and awareness, domain authority accumulates, and that authority then helps everything else on the site rank, including narrower pages built later. More pages, more authority, more traffic. Scale was the advantage.
AI search doesn’t work that way. When someone asks an AI assistant a real question, the system doesn’t run one lookup against one ranked list. It breaks the question into several narrower sub-questions, retrieves candidate passages for each one separately, and compares them against each other before writing an answer, a mechanism documented directly in Google’s own patent filings on retrieval-augmented search. A broad page covering a topic in general often has no passage that specifically answers the exact sub-question being asked. Not a weak answer. No answer at all, from that page, for that question.
That’s the inversion. Broad content used to earn authority that trickled down and helped everything else. Now a specific page has to win its own retrieval event on its own merits, and a broad library mostly produces pages that are present on the site but unretrieved for the question that actually mattered.
Where This Favors One Person Over a Team
This part is genuinely good news for a one-person practice. A larger agency scales by producing more content, across more clients, with more people. That’s a breadth strategy almost by construction: more output, more topics covered, more pages shipped per month. It was a strong model for the old funnel. It’s a structurally weaker one for winning individual, narrow retrieval events, because breadth and depth-per-page pull against each other once headcount and client load are spread thin.
A single practitioner doing one page at a time, completely, doesn’t have that tension. There’s no incentive to ship ten shallow pages instead of two deep ones, because there’s no team quota to hit. That’s not a euphemism for small. It’s the actual mechanism the current research says wins a specific retrieval event.
The Five-Lever Framework Was Already This
Every page reviewed here gets run through five checks: Citability, Conversational Alignment, Authority Signals, Factual Density, Structured Clarity, in one pass, on one page, before it’s called done. That’s not a simplified version of a bigger process. It’s the whole process, applied completely to one page at a time, which is exactly the shape this research says wins a narrow retrieval event: not broad coverage, one specific answer, built completely.
The same instinct shows up in the tooling, not just the framework. The AI Visibility Tracker built for this practice checks a real buyer question against ChatGPT, Perplexity, and Gemini three times each per platform, not once, because a single AI response isn’t a stable measurement. That decision wasn’t made by reading outside research first and reverse-engineering a matching pitch. It came from watching the same question return different answers on back-to-back runs and deciding a single pass wasn’t good enough to act on. It happens to be the same conclusion current research on measuring AI visibility has landed on: one clean-looking number isn’t the goal. A number repeated enough to trust is.
What This Means If You’re Choosing Between a Big Agency and One Person
If AI retrieval now rewards depth on a specific page over breadth across many, the practical math changes. A large content library built broad and shallow across a hundred pages doesn’t automatically out-compete one page built to answer one real question completely, the way it would have under the old ranking model. That’s the actual argument for a Single Page Review as a real, standalone service here, not a smaller taste of a bigger audit: one page, done completely, can win a retrieval event that a much bigger, broader site structurally can’t.
Common Questions
Is this just a convenient story for a solo practice? The underlying mechanism is real and checkable independent of anything about this practice specifically. A 2024 study testing nine content strategies across 10,000 queries found that adding relevant, specific information to a page improved AI-answer visibility by up to 40%, more than any other single change tested, ahead of citing sources or quoting named authorities (Aggarwal et al., ACM SIGKDD 2024). What’s specific to Small Factory 5 is that the five-lever framework and the AI Visibility Tracker’s repeated-measurement approach were both already built around depth and precision, not breadth, before this was the obvious argument to make for either one.
Does this mean big agencies can’t compete in AI search at all? No. It means their default operating model, more output and more breadth spread across more clients, works against them on this specific dimension, not that scale is worthless everywhere. Authority still matters once a page gets retrieved. It just doesn’t guarantee retrieval for one specific question anymore, and that’s the step a lot of broad-content strategies still aren’t built for.
How is this different from long-tail SEO? Same instinct, different mechanism. Long-tail SEO chased low-competition keyword phrases. This is about whether a page has a specific, extractable answer to one of the parallel sub-questions an AI system actually runs. A page can rank fine and still never get retrieved for the exact question it needs to answer.
Want to See What One Page, Done Completely, Actually Looks Like?
That’s exactly what a Single Page Review is. See how it works.
Written by David Cox, GEO consultant at Small Factory 5.
Ranking isn't the same as getting cited.
If you're not sure whether AI search actually cites this page, that's exactly what a page review checks.
See how it works