No. There is no product on sale anywhere that makes ChatGPT recommend your business inside an answer. Ads are a different matter, and they are real. OpenAI announced in January 2026 that it would test ads for US users on the free and Go tiers, placed at the bottom of answers, clearly labeled and separated from the response. It also said those ads will not influence the answers ChatGPT gives. So there is inventory next to the answer, sold like any other media. There is none inside it.
This question usually arrives with an apology in front of it. It does not need one. You are not trying to rig anything. You watched a competitor's name come back from an assistant your own clients use, and you have bought media for twenty years. So you asked the sane commercial question. Is this a channel with a rate card, or is it something else?
It is something else, and the distinction is worth holding onto. An ad is you talking. A recommendation is the machine talking about you. Those cannot be priced the same way, because only one of them belongs to you to sell.
Here is what is genuinely for sale on the AI surfaces right now, what is not for sale at any price, how to read the pay-to-play offers already landing in your inbox, and where the same budget does move the thing you actually wanted.
What You Can Actually Buy
Start with the money that has somewhere legitimate to go. Advertising has arrived on assistant surfaces, and it arrived recently enough that most owners have not seen it yet. In January 2026 OpenAI said it would begin testing ads in ChatGPT for logged-in US adults on the free and Go tiers, shown at the bottom of an answer when there is a relevant sponsored product or service in the conversation. Plus, Pro, Business and Enterprise subscriptions stay ad-free. Ads are labeled, and they are held back from users under 18 and from sensitive contexts such as health and politics. OpenAI also stated that ads will not influence the answers ChatGPT gives.
The other engines each made their own call, which is why one platform's behavior never describes the category. Perplexity started testing ads in November 2024 in the form of sponsored follow-up questions, and said at the time that the answers to those questions would be generated by Perplexity rather than written by the brand. In February 2026 Search Engine Land reported that Perplexity had phased that experiment out, on the reasoning that ads risked the trust that makes an answer worth reading at all. Google's advertising sits in a third place again. Its help documentation says text and Shopping ads from existing Search, Shopping and Performance Max campaigns can be eligible to show inside AI Overviews. It also says advertisers cannot directly target that placement, and that no segmented reporting for it exists yet. In May 2026 Google went further and announced it was testing new ad formats built with Gemini in AI Mode, labeled as sponsored.
Three companies, three answers. Read the list twice and the pattern shows up anyway. Every one of these products sells proximity: a labeled unit above, below or beside an answer, bought through an auction, priced by impression or click. None of them sells the sentence in the middle, the one where an assistant names a business as the right one for a specific person to hire. No platform has published a product that does. That sentence is not inventory, not even at the companies with every commercial reason to make it so.
Whether an ad is worth buying is a separate question, and a reasonable one. It behaves like the rest of your paid media: attention while the money runs, measured by clicks and enquiries. Just do not file it under recommendation, because your buyer will not.
What Nobody Is Selling
The recommendation is the assistant's own credibility, spent on your behalf, to a stranger who is about to act on it. That is why it is not on the price list. An answer that could be bought would be worth nothing to the person reading it, and worth nothing is not a business model. Perplexity walking away from a working ad product is the clearest evidence anyone has offered on that point: labeled inventory was still judged too close to the answer.
So the shortlist gets assembled the slow way, out of what the machine can find, check and repeat about the businesses in your category. The decision an assistant runs before it names anyone is worth watching once in slow motion, because it removes most of the mystery. Nobody outside the model companies can describe the internals, so treat this as our working model rather than a leaked mechanism: it is looking for one real, consistent business, for somebody other than you saying so, and for material that carries knowledge it could not get from the nine other firms in your category.
None of that is a verdict on your work. As we put it in the copy that sits under everything we build: AI can't feel your reputation in a room. It can only work with what it can verify. The room is where your reputation is real. The record is the only part a machine ever meets, and when a competitor's name comes back instead of yours, the gap is almost always in the record rather than the work.
One more boundary belongs here, because it is the same honesty in the other direction. Nobody controls what a model says, and any supplier who promises you a mention, a position or a timeframe is selling something they do not own. What can be promised is readiness: that the evidence exists, agrees with itself and says something worth repeating. Then you measure what happens, and you keep the receipts either way.
The Lists That Will Take Your Money
Meanwhile there is a whole second market that will absolutely take your money, and it is the one your inbox already knows about. Paid directory listings. Best-in-your-city roundups with a placement fee. Guest posts on sites you have never read. Packages of reviews. These are sold on the same promise, that a payment turns into a mention, and they are the reason the pay-to-play question feels unresolved.
Some of it is ordinary advertising with a clean conscience. Sponsoring a page in a publication your buyers genuinely read is a media buy, and there is nothing wrong with one. Where it turns is when the entire purpose of the page is to borrow someone else's authority. Google's spam policies name that directly. Buying or selling links for ranking purposes counts as link spam, and the policy lists advertorials that pass ranking credit, optimized anchor text in guest posts, and low-quality directory links among its examples. Paid links are fine when they carry a rel="sponsored" or rel="nofollow" attribute. That attribute is the whole difference between an ad and an attempt at ranking credit. Google also has a name for the bigger version of the tactic, site reputation abuse, defined as third-party content published on a host site mainly because of ranking signals that host earned from its own work.
Reviews have a harder line still, and it is federal. The FTC rule that took effect on October 21, 2024 prohibits buying or selling fake consumer reviews and testimonials, and prohibits incentives conditioned on the sentiment of a review. Asking every client for an honest review is good practice, and paying a service to produce them is a different act entirely. Compensation that depends on the review being positive is a legal exposure with your name on it, not a visibility tactic.
The practical filter is simpler than the policy language. Ask whether the source would still say this about you if nobody had paid. That is the question the corroboration is supposed to answer, and it is the one a buyer is silently asking too. Paying to be on a list does not make you the reason the list exists.
Why No Is the Better Answer for You
Imagine the version where the answer was yes. A recommendation slot with a price on it goes to whoever can pay for it every month forever, in every category, in every city. A ten-person specialist practice does not win that auction in 2027 and does not win it in 2037. The absence of a buy button is the only reason the shortlist is still open. That is the good news, and it is worth saying out loud.
Being smaller does show up in the current numbers, and it is worth seeing plainly rather than being told it does not matter. In one large vendor study (Ranqo, 100,000+ AI answers across 100+ brands, March to May 2026), household names appeared in 73% of relevant unbranded AI answers, mid-market brands in 44%, and small or niche brands in 11%. That is one vendor's study of its own sample, so hold it loosely. Read as a ladder rather than a verdict, it is also a map. The questions where a national name has thin, generic material are the ones a specialist can answer better than anybody. Those are the questions to claim first.
The economics of the two purchases run in opposite directions, which is the part worth taking to your next budget meeting. An ad stops the day you stop paying. Nothing carries over. A page that answers a real buyer question, corroborated somewhere you do not own, can still be doing its job in two years, and it is still there to be checked by whichever assistant arrives next. Budget rents attention. Evidence accumulates, and accumulated evidence is the part of this a bigger competitor cannot simply outspend you for.
That is also why this work is not the same purchase as your rankings, even though it borrows the same vocabulary. If your first instinct was to ask your search agency, that instinct is right, and the overlap between AI visibility and SEO is worth reading before the next quarterly review, because the two disciplines cover different ground.
Where the Money Actually Works
So point the budget at the three things worth being checkable on. This is the frame we build to: AI recommends businesses it can verify (foundation), that others vouch for (validation), and that know something it doesn't (information gain). Every honest job in this category is one of those three, and the sequence matters more than the spend.
- Make your identity agree with itself. Same name, same address, same people, same credentials, everywhere a machine looks. Unglamorous, mostly free, and nothing above it works without it.
- Put your judgment on the record. Answer the questions buyers actually ask you on the phone, in your own words, with your reasoning visible. The generic version of your category's advice is already in the training data. Yours is not, which is the part that earns the recommendation.
- Get corroborated off your own pages. Independent mentions, accurate listings, real reviews, press that tells the same story your site tells. Self-testimony has a ceiling.
- Measure before and after. Ask the same buyer questions across the engines your clients use, on a date, and save the answers word for word.
That last one costs nothing and settles most arguments, including this one. Running the check yourself takes an afternoon: the same questions, several engines, recorded so you can compare them next quarter. We test five, because ChatGPT, Gemini, Perplexity, Claude and Grok answer the same question differently about the same business, and one screenshot from one engine only ever tells you about one afternoon.
If you want the fuller version of the build rather than the myth-busting version, the guide to getting recommended by ChatGPT walks the same ground in order, including where to spend a first ten hours. The order is the useful part. Skip identity and start publishing, and a year of good work can land somewhere nothing connects it to you.
Want to Learn More?
Probably Genius builds and operates the recommendation layer around expert-led businesses, so that what an assistant can verify finally matches the work you actually do. That is the done-for-you program: your expertise on the record, structured so machines can read it, corroborated beyond your own website, and measured across five engines every month. Every piece is scored 0-100 through the Integrity Gate, and nothing publishes under 80. The published methodologies show how it runs, including the 109-Point Diagnostic behind the free check. No discovery calls. The diagnostic does the talking, an hour, your real score, and you decide.
You're probably a genius at what you do. We make sure AI gets the memo.