The promise, and the machine behind it
The pitch has settled into a standard shape. Search has changed. People ask assistants instead of typing keywords. Assistants answer with a cited source instead of a list of links. Content built for 2022 will not be that source. Here is a tool that fixes it.
All of that framing is correct. It is the next sentence that matters, and it is usually printed right there in the same email: the tool reads the pages that currently rank for the topic, and drafts an article from them in your brand voice.
One example, currently in market: AnswerThePublic’s Content Studio, promoted by Neil Patel, which analyses your site, proposes ranked content ideas, then researches top ranking pages and generates the article, publishable straight to your site. The marketing states plainly that content built this way earns citations.
Read the mechanism and the diagnosis and they contradict each other.
What “synthesise the top ten” produces
A model answering a question does not need more copies of what it already holds. If ten pages already say the same thing, an eleventh page saying the same thing adds nothing to the answer, and the model has no reason to prefer it over the ten it has seen for longer, on domains it has more history with.
That is not a quality judgement about the writing. The output can be well organised, correct and pleasant to read. It can be better written than the sources it was built from. It is still, by construction, a restatement of the consensus, and a restatement of the consensus is the exact category of page a citation mechanism is designed to skip.
The tools describe this accurately without noticing. Researching the top ranking pages and rewriting them in your voice is a description of producing derivative content at speed. Applied to rankings, it was arbitrage that worked for a while. Applied to citation, it is self-defeating, because the selection criterion changed.
Why the criterion changed
Ranking and citation reward different things.
Ranking rewards the best match for a query. A page can be the best available match while containing nothing new, if it is clearer, better structured or better linked than the others.
Citation rewards being the source. When a system attributes a claim, it points at where the claim comes from. A page earns that by holding something the other pages do not: original data, a first hand account, a specific number, a defined concept, a document, a method with a name.
Structure helps a model parse you. It does not make you worth pointing at. That distinction is where most of the current AEO advice quietly collapses: it is all about markup, formatting, question headings and summary blocks, all of which are real and none of which answer the question of why anyone would cite this page rather than the source it was assembled from.
We have seen this exact move before
Every Google update of the last fifteen years punished the same thing: content produced at scale from what was already ranking, optimised for the mechanism rather than the reader. Panda punished it. Helpful Content punished it. The tooling changed each time, the pattern never did.
What is happening now is the third act of that same play. The mechanism is new, so the assumption is that the arbitrage resets and the old approach becomes viable again under a new name. It does not reset. The mechanism changed in the direction that makes derivative content less viable, not more.
The tell is in the vocabulary. Tools that have not understood the shift talk about optimising for AI, structuring for AI, formatting for AI. Everything is about how the content is presented to the system. Almost nothing is about whether the content contains anything a system would have a reason to attribute.
What this does not mean
It does not mean the tools are useless or that the people selling them are acting in bad faith. Idea generation from a real site is genuinely useful. Drafting help is genuinely useful. Publishing friction is a real problem and reducing it is a real service.
What it means is narrower and harder: no tool can generate the thing that gets you cited, because that thing is by definition not present in the material any tool has access to. It comes from doing something, measuring something, or defining something, and then writing down what happened. A generator working from the existing corpus cannot produce what is missing from the existing corpus.
It also does not mean volume never works. It means volume works on the axis it always worked on, which is coverage, and does nothing on the axis that produces citation, which is being the source.
The practical version
Two questions, applied to any page before it is published.
What is on this page that is not already on the pages it was built from? If the answer is “better organisation”, the page may rank and will not be cited.
If a system attributed this claim, would pointing here make sense to a reader who followed the link? Being the destination of a citation is a reader-facing promise, not a technical property.
Neither question is answerable by a tool, which is precisely why the tools do not ask them.
Citability is one of the five surfaces authority lives on, and it is the newest. What decides whether a page becomes a source, rather than one more retelling, is covered across the Authority Growth cluster.
