There's a new pitch landing in marketing inboxes right now, and it's beautiful in the way that only truly dangerous sales decks can be. Publishers — major ones, the kind with mastheads and journalism awards — are selling AI visibility as a product. Place your brand on our pages, the pitch goes, and you'll show up when an AI assistant answers your customer's question.

It's slick. It's well-timed. And if you squint at the data the right way, it almost makes sense.

Almost.

Here's the setup. When an AI assistant builds an answer — whether that's ChatGPT, Perplexity, or Google's AI Overviews — it overwhelmingly reaches for third-party sources rather than a brand's own site. One analysis of more than 25 million cited links across the major assistants found earned media accounts for roughly 84% of all citations. A brand's own domain? A small single-digit share. Mentions of a brand scattered across trusted third-party sites predict whether an assistant will cite that brand far more reliably than links pointing back to its own pages.

So the logic writes itself. If AI assistants pull from authoritative outside domains regardless of how the content arrived there, then the shortest path to being cited is to buy your way onto one of those domains. Pay for the placement, inherit the domain's authority, appear in the answer.

And publishers are not being subtle about this. A joint venture owned by three German publishing houses now runs prompt-set audits across the main assistants, then produces listicles and comparisons designed — in its own description — with AI systems rather than human readers as the primary audience. One publisher has claimed a roughly 25% rise in citations from its own domains for a single campaign over three months. Programmes are being priced anywhere from a few thousand to the high tens of thousands a year.

This is not theoretical. It is the specific thing being sold to brands right now.

There's just one problem. (There's always one problem. Usually it's hiding in the same data deck that made the pitch look so compelling in the first place.)

That same body of evidence that makes the paid-publisher route look clever? It also demolishes it. The analysis that put earned media at 84% of citations put paid and advertorial content at around 0.3%.

Let that number sit for a second. 0.3%.

The property that makes a publisher page worth citing — its standing as independent editorial — is precisely the property that payment removes the moment it's disclosed. AI assistants demonstrably weight down content they can identify as advertising. So the paid-publisher play only pays off to the degree the payment is invisible to the model.

Which leaves you standing in front of two doors, and neither leads where the pitch implies.

Behind door one: the placement is honestly labelled as sponsored. The assistant largely discounts it. You've bought a page the model treats as an advertisement rather than a source. Congratulations, you've paid premium rates for content the AI equivalent of shrugs at.

Behind door two: the sponsorship is played down or obscured so the page reads as editorial. You're now doing the thing that Google's May 2026 spam policy expansion explicitly reclassified as manipulation of generative answers. That enforcement is retroactive — reaching pages published before the rule changed — and it carries cascade risk, because a penalty in Google's index can pull visibility down across the AI surfaces and downstream tools that draw on that index simultaneously.

Door one wastes the money. Door two turns the placement into a liability that can detonate later. (I'd say "pick your poison," but honestly, both taste about the same.)

It's tempting to reach for the reassurance that the 0.3% figure only captures paid content the analysis could identify as paid, and that cleverly disguised placements slip through uncounted.

That is true. And it is not reassurance. It's the same problem viewed from the other side.

If the value of the channel depends on the disguise holding, then you've bought an asset whose worth evaporates the instant an assistant, a regulator, or a competitor gets better at spotting what it is. And these systems are getting better at spotting what things are. That's their whole deal.

The field is openly divided on whether a disclosed branded placement keeps any citation value at all. Some practitioners argue disclosure changes nothing. The retrieval data suggests it changes a great deal. That disagreement is unresolved, and an unresolved disagreement about whether the thing works is a strange foundation for a budget line item. (I've made some questionable investment decisions in my time, but even I draw the line at "the experts can't agree if this actually does anything.")

The paid-publisher route looks worse still once you place it next to the other ways a brand can reach an AI answer, because each of those alternatives is honest about what it is.

The platform ad slot is the cleanest option. The sponsored cards that began appearing inside ChatGPT in early 2026, and opened to self-serve buyers by the spring, sit below the assistant's answer — clearly labelled, bought through a second-price auction at roughly $25 CPM, with daily budgets starting around the price of a decent lunch. The governing principle the platform has repeated is answer independence: the ad never alters what the model says. The slot buys attention beside the answer, not a place inside it. It's rented visibility, but rented on terms you control.

The revenue-share model points in the opposite direction entirely. One major assistant killed its own advertising experiment because its leadership feared ads sitting near answers would make users doubt the whole environment. It pivoted to paying publishers directly, routing the large majority of a multi-million-dollar pool to the outlets whose content gets cited. The brand buys nothing at all. Where money was allowed to touch the answer, this engine chose to send it to the publisher rather than the advertiser. Which tells you something about how the platforms themselves regard the idea of a purchasable citation.

Genuine earned coverage is the route the paid-publisher pitch is quietly imitating. A journalist's article about a product can't be pulled back onto the brand's balance sheet. But the model trusts it precisely because no one paid for it. The brand controls none of it and can rely on all of it.

Against those three, the paid publisher placement occupies the only square that is bad on every axis. You pay (unlike earned coverage). You don't own (unlike almost nothing else you pay for). The model discounts you if you're honest and penalises you if you're not. And the price sits at a premium while the sellers themselves concede the measurement underneath is inconsistent.

Here's the part the pitch never raises, because it exposes what you're actually purchasing. A sponsored placement is not a purchase. It's a tenancy. And the lease is almost never written down.

Advertorial has run for years on an unspoken assumption that published meant permanent. That assumption held while the page earned its keep through search traffic. Now that traffic has collapsed and the archive has become the one piece of inventory a publisher has already been paid for once. It can be sold twice.

The mechanics are unglamorous. The takedown arrives several months after the campaign closes, framed as routine editorial housekeeping. The link gets stripped while the article stays live. Or it quietly acquires a tag telling crawlers to ignore it. Or the page vanishes, redirected to the homepage so nothing visibly breaks. From a reporting dashboard, all four look identical. The placement just stops working.

For AI visibility this is more damaging than the equivalent loss ever was in traditional search. When a placement died in search, it stopped passing authority. When a placement dies in an AI answer, it stops existing as a source the model can retrieve. The brand's presence degrades with no error thrown, no alert raised, and nothing on any dashboard to mark the moment it happened.

Tracking of citation stability suggests most cited positions hold steady week to week. But among the positions that do move, the clear majority of the movement is loss rather than gain. Decay is the ordinary weather of this channel, not the exception. And the paid-publisher route hands the controls to a counterparty whose commercial incentives have quietly reversed. The party that took your money to publish the page now has a live financial reason to unpublish it and charge again.

That's not a citation strategy. That's a subscription you didn't know you were signing up for.

None of this makes paid distribution worthless. It makes it a media buy — which should be priced, governed, and expected to expire like one, rather than mistaken for something you come to own.

If you still want publisher placement for AI visibility: get the retention term in writing before any money moves. A publisher who won't commit to how long the piece stays up has already answered your question. Audit placements quarterly — a removal caught inside a month is usually recoverable; one caught inside a year usually isn't. Weight spend toward outlets running a handful of sponsored pieces a year over those churning several a week. And confirm the domain is one the assistants actually retrieve from before paying for anything, because a placement on a site no model reads is a media buy wearing a citation strategy's clothes.

Of everything making you visible in AI answers today, how much did you earn and how much are you renting?

The honest reply for most brands is that they don't know. The tools they're being sold measure presence rather than ownership. The harder questions are whether the assistants will ever learn to price a disclosed branded placement as something between advertising and editorial instead of discounting it wholesale. Whether enforcement aimed at manufactured citations tightens fast enough to close the arbitrage before the money chasing it grows larger. And whether a citation you have to keep paying to retain was ever really a citation at all — or was always an advertisement that happened, for a while, to be standing in the right place.

The platform ad slot is revocable by you. Earned coverage is revocable by no one holding a price list. The paid publisher placement is revocable by someone else, on their schedule, for their reasons.

Pick the one you can sleep with. Preferably before the invoice arrives.

Jo Lambadjieva is an entrepreneur and AI expert in the e-commerce industry. She is the founder and CEO of Amazing Wave, an agency specializing in AI-driven solutions for e-commerce businesses. With over 13 years of experience in digital marketing, agency work, and e-commerce, Joanna has established herself as a thought leader in integrating AI technologies for business growth.

😍It was great, all I needed

🙂Good, not great

🙄It sucked