The "Pay to Get Into ChatGPT" Email Is Just Link-Buying Again
You opened a cold email this week promising to plant your brand inside ChatGPT and Perplexity for a few hundred dollars per mention. The pitch sounds new because the platforms are new. The mechanics are not. I have spent fifteen years in SEO, and this is the exact same paid-link scheme I watched companies get burned by in 2012, repriced at a steep markup and sold under a fresh acronym.
This article proves three things: paying to be inserted into AI answers is link-buying in a new suit, you are paying ten to fifteen times the cost of a backlink for a result that often vanishes inside a month, and as of Google's June 2026 spam update, the tactic now carries penalty risk on top of being a waste. That is three strikes before you have spent a dollar.
You are renting a result that disappears in 30 days, at 10x the price
Gaetano DiNardi laid out the economics in Search Engine Land on June 26, 2026, and the numbers should stop you cold. Paid brand mentions are being sold at ten to fifteen times the cost of a normal backlink. A single mention runs around $250. Vendors package this into pitches like "25 brand placements" sold against a claimed "10 to 15 percent mention-rate lift," which is the kind of precise-sounding promise that should make you suspicious, not reassured.
Now follow the money into what you actually get. DiNardi reports that paid "mentions" placed on Reddit are often scrubbed within 30 days. Picture a roofer in Tampa who pays $250 for a Reddit thread comment naming his company as the best in the area. A month later the moderators remove it, the AI re-crawls, and the citation he paid for is gone. He rented a result. He did not own anything.
There is a second problem buried in the same report. These placements usually go undisclosed, which is a Federal Trade Commission disclosure issue, not a gray area. If you pay for an endorsement and nobody says it was paid, you are the advertiser on the hook, not the vendor who took your money and moved on. You are buying legal exposure and calling it marketing.
I have written before about why being mentioned by an AI is not the same as being believed by it, and a purchased mention is the weakest version of a mention there is. It has no trust behind it. It is a sentence someone was paid to write, sitting on a page built to disappear.
The pitch works because the trick really does work, for a minute
If these placements did nothing, nobody would buy them twice, and the scam would die on its own. The uncomfortable truth is that planted text genuinely sways AI answers in the short term, which is exactly what makes the cold email convincing when a vendor shows you a live demo.
A vendor can run a query in front of you, point at your brand in the response, and say "see, it works." It does work, that afternoon. DiNardi himself frames the timeline plainly.
"This creates a temporary window of effectiveness, perhaps one to two years, before AI platform companies refine their authority and spam signals."
Gaetano DiNardi, Search Engine Land, June 26, 2026
A one-to-two-year window is not a strategy. It is the lifespan of a tactic the platforms have not gotten around to killing yet. Glenn Gabe, cited in the same coverage, frames paid AI mentions as a straight evolution of the old paid-link schemes, and the people who lived through those schemes know how the story ends. The window closes. The bill comes due. Lily Ray put the naivety of betting against the platforms in blunt terms.
"It's naive and irresponsible to assume the search and AI platforms won't eventually catch on too."
Lily Ray, quoted in Search Engine Land, June 26, 2026
She also pointed at who is selling this, and why the inexperience matters.
"…some of the new GEO/AEO companies, the ones selling services to boost visibility in AI search, lack the experience and context to anticipate how their approaches will be treated once Google and the AI platforms build countermeasures…"
Lily Ray, quoted in Search Engine Land, June 26, 2026
That is the part that should land. Many of the vendors emailing you have never weathered a Google update. They are selling you a demo, not a durable result, and they will not be around to explain the cleanup. This is the same pattern as the last over-sold AI tactic, the llms.txt file that vendors pushed as essential. New acronym, same empty promise.
What the research shows: easy to do, fragile to keep, now against the rules
Two things became true in the same week, and they cut in opposite directions for anyone holding a vendor's invoice. The first is that manipulating an AI answer is technically easy. A Cornell Tech preprint (arxiv 2605.24245, and worth flagging that it is not peer-reviewed) found that roughly 13 words of planted text inserted an attacker's chosen entity into a Deep Research report in 38 to 51 percent of sessions. Buried in under 4 percent of a page, the planted text still landed 30 to 53 percent of the time.
Read those numbers the way a vendor reads them, and you understand the sales pitch instantly. A dozen words can swing an AI report half the time. That is cheap to produce and easy to demonstrate. It is also exactly the fragility that should worry you, because anything thirteen words can build, thirteen different words from a competitor can overwrite, and the platform can wipe with a single signal update.
The same research shows how much of this surface is already shrinking. Gemini Deep Research pulled from user-generated content only 12.1 percent of the time, and SE Ranking finds Google's own self-citations now make up roughly 20 percent of AI Mode citations. The space where a bought Reddit comment can even be seen is getting narrower while the vendors are still selling it as wide open.
The second thing that became true is the one that turns a bad investment into a risky one. Matt G. Southern reported in Search Engine Journal on June 27, 2026 that Google's June 2026 spam update now treats attempts to "manipulate generative AI responses" as a policy violation. The tactic the cold email is selling is now, in writing, a spam violation. Enforcement is hard, Southern notes, but the rule exists, which means you are buying something the platform has formally declared against the rules. Southern's larger point reframes the whole job.
"AI visibility has become a surface you actively monitor, not just a channel you passively optimize for."
Matt G. Southern, Search Engine Journal, June 27, 2026
Compare that to the official line on the last shortcut everyone was sold. John Mueller of Google said the llms.txt file is "not done for search" and called it a "temporary crutch." When the platform itself describes your tactic as a temporary crutch, you have your answer about how long it will pay off. A bought AI mention is the same kind of crutch, except this one now sits inside the spam policy.
Three questions that tell you if a pitch is a scam before you pay
You do not need to understand how a language model retrieves a citation to vet one of these emails. You need three plain tests, and any pitch that fails them is the same risk as buying links, only more expensive. Run them in order.
First test: would you be comfortable if a journalist saw exactly what you paid for? If the honest description of the service is "we pay people to mention you without saying we paid them," you already have your answer, and so would the FTC. The DiNardi reporting is clear that these placements usually go undisclosed. A tactic you would have to hide is a tactic that can be exposed.
Second test: does the result survive if the vendor stops billing you? Ask the salesperson directly, in writing: "If I cancel next month, what happens to the mentions I paid for?" If the placements get scrubbed, like the 30-day Reddit removals DiNardi documented, you were never buying an asset. You were renting a billboard on a wall that gets repainted monthly.
Third test: is anyone disclosing that it is paid? If the answer is no, you are carrying the disclosure liability, not the vendor. Three noes in a row, no journalist comfort, no survival without billing, no disclosure, and you are looking at the link-buying scam in a new costume.
Here is how to split the work. The owner runs the three tests, because they are business judgment, not technical calls, and they take ten minutes. Delegate the verification to whoever handles your marketing: have them ask the vendor for a sample placement and check it again in 30 days to see if it still exists. The exact question to put to any vendor before signing is this: "Show me a client placement you made six months ago that is still live and still cited, and tell me whether it was disclosed as paid." A vendor selling something durable can answer that. A vendor selling a demo will change the subject.
If you want to confirm what is actually happening with your own brand instead of taking a salesperson's word, learn how to check whether AI is genuinely recommending your business first. Measure your starting point before anyone sells you a lift on top of a number you never verified.
Spend the $250-a-mention budget on signals that compound
The same money that buys 25 disappearing mentions buys a quarter of real, durable signal, and the difference is that one of them is still working a year from now. My rule after fifteen years and three "SEO is dead" cycles is simple: if a tactic would embarrass you when the platform catches on, do not buy it. Spend on the things that get stronger with age instead of weaker.
Earn genuine mentions and reviews. A real review from a real customer survives a moderation sweep, a model update, and a journalist's scrutiny, because there is nothing to hide. This is not a soft suggestion. In local results, reviews now beat proximity for the local top three, and the same trust signal that wins the map pack feeds the AI answer. One investment, two payoffs, and no invoice that has to keep recurring to keep working.
Keep your business facts consistent everywhere. Your name, hours, services, and claims should match across your site, your profiles, and your listings, because an AI quoting you accurately depends on finding one consistent version of the truth. This is unglamorous and it compounds. This is the real work behind the idea that brand is the new backlink, the strength you build into your name is the thing models lean on when they decide who to cite.
Make your site easy for an AI to quote accurately. Structured, verifiable claims, plain answers to the questions customers actually ask, facts a model can lift without guessing. None of this disappears in 30 days. None of it shows up in a spam report. And when you measure, do not measure the vendor's "mention-rate lift." Measure whether real reviews are growing, whether your facts are consistent across the web, and whether your own checks show AI describing your business correctly. Those numbers do not reset when you stop paying.
Frequently Asked Questions
Is paying to get my brand into ChatGPT against Google's rules?
Yes. Matt G. Southern reported in Search Engine Journal on June 27, 2026 that Google's June 2026 spam update now treats attempts to manipulate generative AI responses as a policy violation. Enforcement is hard and inconsistent right now, but the rule is in writing. That means a paid-placement tactic is no longer just a poor use of money, it carries penalty risk if Google acts on it.
If planted text really does change AI answers, why not just buy it?
Because the effect is real but fragile. A Cornell Tech preprint found that about 13 words of planted text inserted a chosen brand into a Deep Research report 38 to 51 percent of the time, which is exactly why vendors can demo a quick win. The catch is that anything those words build, a competitor's words or a model update can erase. DiNardi documented paid Reddit mentions being scrubbed within 30 days, so you are paying a premium for a result you do not own.
What should I spend the money on instead?
Fund the signals that survive a moderation sweep and a model update. Earn real customer reviews, which already beat proximity for the local top three and feed AI answers with genuine trust. Keep your business facts consistent across your site and listings so an AI can quote you accurately. Make your claims structured and verifiable. None of this disappears in 30 days, and none of it shows up in a spam report.
The cold email in your inbox is not selling you the future. It is selling you a 2012 tactic with a 2026 price tag and a fresh acronym stapled to it. The vendors moving fastest right now are the ones with the least to lose when the platforms close the window, and Google just told you in writing that it intends to. Before you wire anyone $250 a mention, ask yourself the only question that has ever mattered in this work: when the platform catches on, and it always catches on, will you be glad you bought this or ashamed you did?
