Two offers, one week apart, one sender
The first arrives as a standard link sale. A pool of 3,200 websites, expanding toward ten thousand. Niches available: casino, online gaming, betting, football, other sports. The placement is a permanent sidebar link, appearing sitewide on every page of each site. Fifty sites for $120, 350 for $400, and a pre-order tier that works out at a dollar a link. Orders completed within a few hours, URLs shared once the links are live.
The second arrives from the same sender, days later. A new product: strategic AI citations that shape how ChatGPT, Gemini and other models talk about your brand. Fourteen dollars each. The pitch describes it as SEO for the AI era.
It is worth being precise about what is and is not knowable here. What the second product actually consists of is not stated in the offer. What is stated is the price, the promise, and who is selling it. That is enough to work with, because the promise is testable on its own terms.
What the first offer tells you
Sitewide sidebar links across thousands of unrelated sites, in the highest risk niches in the industry, sold by volume and delivered in hours, is the most heavily documented link pattern of the last fifteen years. Every mechanism Google built since 2012 exists to identify it. The footprint is not subtle: identical anchor placement, identical position in the template, no editorial context, no relationship between the linking page and the linked one.
One detail in the pitch does more work than the rest of it. The sites are described as having no spam flags. That is a claim about what a third party tool currently displays, not about whether the sites are read by anyone. It is a quality standard defined entirely by a metric, which is a good short definition of numerical authority: optimising the reading rather than the thing being read.
At a dollar a link, the arithmetic only works if nobody involved expects the links to be read. That is not a moral observation. It is the business model.
Why the second offer cannot deliver what it promises
Set aside intent entirely and look at the mechanism.
A model deciding what to cite is choosing where a claim came from. For a page to be that source it has to be in the model’s reach, and it has to be the place where the claim actually lives. A sidebar link on an unrelated site is neither. It contains no claim, it is attached to no context, and it appears on pages whose subject has nothing to do with yours.
There is a second problem, larger than the first. The offer promises to shape how a model talks about your brand. Nobody selling links controls that. The retrieval and selection behaviour of a commercial model is not an interface anyone can buy into from outside. What a vendor can do is place text somewhere and hope it is ingested. Whether it is ingested, whether it is trusted, and whether it is surfaced in an answer are three separate gates, none of which is for sale at fourteen dollars.
And a third, which is the one that matters longest: citation is not a link. A model can cite a page that never linked to you and ignore a thousand pages that did. Buying links to influence citation applies the currency of the old mechanism to a mechanism that does not accept it.
The pattern this belongs to
This is what a gold rush looks like from inside. A new selection mechanism appears. Nobody knows how it works yet. The existing supply of sellable inventory gets relabelled with the new vocabulary, because relabelling is fast and building something new is slow.
It happened with mobile-first, with E-A-T, with Core Web Vitals, with every named shift of the last decade. The inventory stays the same, the deck changes. AEO is simply the current label, and it is a better one than usual for this purpose, because almost no buyer can yet verify the claim. Nobody can easily check whether a model cites them more than it did last month, and a product whose result cannot be measured is an extremely comfortable product to sell.
The price is the clearest tell. Fourteen dollars buys placement. It does not buy the two things that actually produce citation, which are an audience that would have any reason to read the page, and a claim worth attributing.
What the alternative actually costs
Being cited comes from holding something the corpus does not already hold: a number you measured, a document you produced, a concept you defined and named, an account of something you did. That is slow, it does not scale, and nobody can sell it to you in a package of 350.
Which is precisely why the packages exist.
The distinction between authority that a metric can see and authority that a system will act on is the subject of the Authority Growth cluster. The short version: if it can be bought by volume in an afternoon, it is the kind that shows up in a tool and not the kind that shows up in an answer.
