Four engines, four different answers. Two of them sell you a labeled box next to the answer. One of them built that box and tore it out. None of them will sell you the answer itself.
Ad space around AI answers is a real, growing product category in 2026. It's just not the same product as being recommended, and the engines are careful to keep the two visually and functionally separate.
Every engine above draws the same line, whether it currently sells ads or not: the ad unit is a separate object from the answer. You can buy the box. You cannot buy the sentence where the model tells someone which brand to pick.
OpenAI says this outright - ads run on separate systems from the chat model, and advertisers can't shape, rank, or alter ChatGPT's responses. Google labels its AI Mode ads distinctly from the AI-generated list around them. Microsoft's sponsored panels sit alongside Copilot's own synthesis, not inside it. This isn't a policy quirk. It's the thing keeping each engine's answers worth trusting - and worth appearing in. The moment an engine sells the recommendation itself, the recommendation stops meaning anything, and so does every ad sold next to it. We go deeper on why the organic pick moves on retrieval and trust signals, not spend, in why does ChatGPT recommend your competitor.
There's an honest middle ground worth naming, because pretending it doesn't exist makes you easier to sell a lie later. AI engines lean on trusted third-party sources - comparison sites, review platforms, industry roundups - when they build an answer. Some of those sources accept payment for inclusion: affiliate-driven "best of" listicles, sponsored placements dressed as editorial rankings, pay-to-play directories.
If an engine already treats that source as trustworthy, and reads it while answering a buyer's question, money did indirectly reach the answer - just not through the engine. That's a real mechanism, and a fragile one. It depends entirely on the source keeping its reputation as independent. The day an engine's systems (or a journalist, or a competitor) catch that source laundering payment into "editorial," the engine stops trusting it, and every brand that bought its way in loses the placement overnight, with no refund and no warning.
Run both tracks, and don't confuse them. Buy the labeled ad units where the economics work - they're real inventory now, with real reach, especially ChatGPT's since its May 2026 self-serve launch. Separately, earn the organic pick: publish the pages that answer a buyer's actual question, get named on the comparison and review sources engines already trust, and keep your facts consistent everywhere a model might read them.
Then measure the two apart. An ad impression next to an answer and your brand being the answer are different outcomes with different value, and a vendor that reports them as one number is hiding which one you're actually getting.
We don't sell placement, because there's nothing to sell - the organic recommendation isn't for sale anywhere, from any engine, at any price. What we do is measure whether you're earning it: a frozen set of real buyer questions, asked across ChatGPT, Claude, Gemini, and Perplexity every wave, tracking whether your brand is named, where, and against which competitors. Every count re-derives to a specific question and answer, so you always know whether movement came from work that earns the pick or from a paid box someone put next to it.