Where AI Products Buy Attention Now, Beyond Google and Meta

In late August 2026 a one-page website with a payment button and a list went viral among software makers. Submit a product, bid a few dollars, and the highest bid holds the top of a public leaderboard until someone pays more. Secondary coverage put the bids at roughly $130,000 in the first 48 hours and the top bid in the low five figures, and within days there were boards for pixels, maps, apps and social profiles, and an index to list the boards themselves. A week before, an indie developer had sold ten sticker slots on the lid of a MacBook he did not yet own by live auction, raising 111% of its price.

These are stunts, and they are also data. Founders with AI products are paying for placement in places that did not sell attention two years ago, because the places that do sell it, search and social, are crowded and the AI assistants that increasingly answer buying questions do not run on either. This census records each channel, what is actually for sale, what it costs where a price is published, who sees it, and whether it does anything for visibility inside an AI assistant. We do not name the viral boards; the category is the subject and the sites change weekly.

The mechanic is older than the web: the Million Dollar Homepage sold pixels in 2005, and the car-wrap and laptop-lid auctions cite it as an ancestor. What August added was a leaderboard. Instead of a fixed price per pixel, each product’s rank is its cumulative payment, the top spot belongs to whoever has paid most, and every buyer stays on the board permanently in payment order. One index that catalogues the copies describes the genre in a phrase: “Your rank is what you paid.” The variants we read sold pixel grids at two cents each, positions on a world map, ranks for iOS and Android apps, and “featured on the homepage” slots for a dollar more than the current leader.

Two things make the category interesting to a marketer beyond the spectacle. First, it is a pure auction for attention with no targeting, no algorithm and no measurement, which makes it a clean price signal for what makers think a day at the top of a maker-facing page is worth: on the viral board, a few thousand dollars for the top spot at the peak, single dollars for the long tail. Second, the audience is almost entirely other makers, which is the right audience for developer tools and the wrong one for nearly everything else. The half-life is the shortest in the census. Treat it as a launch-week stunt with a known price, and keep the receipt.

An August 2026 survey of twelve AI tool directories found that eleven have a free way in and that paid submissions, “typically $50–$200,” buy a place at the front of queues that can otherwise take months. One directory is paid-only; others sell a followed link, a queue skip, or a premium placement from about $19. The survey’s argument for paying is that the larger directories are crawled by AI assistants and distribute new listings through their own newsletters, so a listing is both a page a model can read and a message to a subscriber list. That matches what our audit of AI citation visibility across 500 software sites found: assistants cite structured, comparative pages, and a directory entry is exactly that shape.

Launch platforms are the same channel with a clock attached. The largest are free and the value is a day of concentrated attention plus a permanent page that comparison traffic finds later. The practical rule the survey states, and we agree with, is to launch on the free platforms and buy queue position only on directories whose domain authority and newsletter reach justify it. Ten curated directories, a launch and a permanent listing on the comparison sites is a week’s work and a few hundred dollars, and it is the part of this census with the longest half-life for the least money.

The pattern across the four is that price transparency runs opposite to durability. The auction channels publish every bid and last a day. The editorial and sponsorship channels publish nothing and last for years. The assistant ads sit in the middle, priced by auction and lasting as long as the budget, which is why our ChatGPT Ads market ledger is the companion page to this one and why Perplexity’s decision to drop advertising still matters: not every assistant will sell the slot.

Read the two right-hand columns together. A channel that feeds AI assistants and lasts is worth a standing budget; a channel that does neither is worth a stunt budget or nothing. For most AI products that sorts the census into three tiers. Standing budget: directories with authority, ads inside assistants where the product has a conversational buying moment, and disclosed editorial placements on publishers that get cited. Launch budget: the launch platforms and one week on the boards, priced as a known, small cost with a screenshot at the end. Opportunistic: creator sponsorships when a comparable rate can be found, and physical-object sponsorship when the object is going somewhere your buyers are.

What the table cannot do is tell you the cost per visit, because almost none of these channels report visits and the ones that do are the auctions, which report visits to the board rather than to you. If you want to compare any of this with search or social, put a tagged link on every placement and count on your own analytics. That is the same discipline our guide to marketing tasks for coding agents recommends automating first, and it is the only way the cheap channels ever get compared with the expensive ones honestly.