If your site sits under DR-20, this is why competitors get quoted by AI engines while you get summarized, or skipped entirely. Every page an AI engine ingests gets assigned a role. Either the engine will summarize it (extracting a fact, paraphrasing, attributing loosely or not at all) or the engine will quote it, with attribution, by name. Which role you get assigned is not random. It is a function of originality signals, and sub-DR-20 brands can engineer those signals into every post they publish. The broader AEO framework this slots into is Answer Engine Optimization Explained.
This is the cluster’s most counter-intuitive post. The instinct for a small brand is to write comprehensive content, covering every subtopic the competitors cover, mirroring the shape of what already ranks. That instinct produces recap content. Recap content lands in the summary layer. It does not get quoted.
§1. What Do AI Engines Actually Score for Citations?
Before a page can be cited, it is scored on how much it overlaps with everything else the engine has seen on the topic. Near-duplicates are collapsed. High-overlap pages are used as summary fodder. Low-overlap pages are used as quote sources. The overlap score is not exposed, but its effects are. A cross-posted copy of your own article is a near-duplicate too. The Dev.to cross-posting guide covers how a canonical URL tells Google which copy is the original.
§2. What Are the Five Originality Signals?
Original data is the ceiling. Distinctive framing is the floor. In between: unique entities you have named and no other site has yet indexed, first-to-name status on an emergent pattern, and personal incidents engines treat as primary source material. You do not need all five on every post. You need at least one on every post that matters.
§3, The overlap distribution
Most pages on any topic cluster at high overlap. The quote layer is a thin tail. For a sub-DR-20 brand, this is leverage. You are competing against hundreds of near-identical recap pages. The cost to enter the quote tail is the cost of collecting one piece of original data or committing to one distinctive frame.
§4. Why Is Recap Content the New Thin Content?
A side-by-side example. Recap paragraph: says true things, says them the way every other source says them. Original paragraph: cites a specific sample, names a specific number, pinpoints which engine saw the effect first. The recap gets absorbed into the engine’s summary model. The original gets pulled out and attributed.
§5, How to engineer originality into a cadence
Originality does not require a research budget. A small site can run its own audits on its own corpus and publish the results. Citability ran against twelve sites this month. Six cities compared on a specific metric. One keyword tracked across three answer engines over eight weeks. Each of those is original data. Each lives in the quote tail. Once the data is yours, the markup that ties a page to its author and publication date is the next step. Schema.org for answer engines lists which properties carry that signal.
§6, Measurement
Originality shows up as citation count and quote length. A page in the summary layer earns occasional partial citations. A page in the quote layer earns repeated, direct, attributed quotes with longer excerpts. Engines show very different numbers of sources per answer, so read how Perplexity and ChatGPT pick citations before you compare counts across them.
Bridge
If you do not know whether your pages are landing in the summary layer or the quote layer, the free citability.dev scan is the quickest first read: it checks crawl access, structure and freshness signals on your own pages in about a minute.
