# The Companies That AI Forgot ## A Failed Search A few months ago, I ran a small experiment. I asked an AI to help me find a hoodie manufacturer that could handle small-batch sampling. My requirements were specific: niche, good quality, reasonable pricing, willing to take orders under 100 units. The AI ran multiple rounds of searches—Zhihu to Reddit, Xiaohongshu to 1688, Chinese forums to customs databases. Eventually it gave me two names. The first one, I later discovered, appeared in my results not because it was good, but because it had registered accounts on a Russian Q&A site, an anime forum, and a personal blog—repeatedly posting its company profile across all of them. A garment factory advertising itself on an anime forum has one obvious purpose: backlinks. Get noticed by search engines, get noticed by AI. The second one came from a random hit in U.S. customs bill-of-lading data. Everything I knew about it amounted to: "Existed in 2019, shipped 120 invoices to some U.S. brand." After that, seven years of silence. It might have shut down, pivoted, rebranded—I had no way to know. After all that, I realized something: I hadn't found two manufacturers. I had picked two from a tiny pool—one shaped by the most aggressive SEO, the other by a coincidental hit in a customs database. The truly good factories—the ones that have run for 30 years, supplied major international brands, earned their reputation through word of mouth in the trade—the AI couldn't see them at all. They have no English websites, no SEO teams, no content matrices, no accounts on anime forums. They exist. But in the AI's understanding of the world, they don't. ## A New Kind of Invisibility This stayed with me for a long time, because it wasn't an isolated retrieval failure. It was a preview of something structural. We are entering a new information order. AI is replacing search engines as the primary way humans access information. Procurement directors looking for suppliers, founders looking for partners, buyers looking for brands, consumers looking for services—more and more, they don't open Google. They ask ChatGPT, Claude, Perplexity, Doubao. This sounds like a tool change. It isn't. It's a rule change. In the search-engine era, "existing" and "being findable" weren't identical, but they were close. If your site was indexed, you had at least a theoretical chance of being found. Google's ranking was unfair, but the result was a long list—users could click to page 5, page 10, and stumble onto companies with bad SEO and excellent products. AI doesn't give you a list. It gives you an answer. What it recommends is what it "knows." What it doesn't know, as far as the user is concerned, does not exist. And what AI "knows" is determined by its training data and retrieval sources. Companies that get fed into the pipeline repeatedly—the ones with marketing teams, content operations, Wikipedia entries, press coverage, active LinkedIn presences—the AI develops strong recognition of them. Companies that don't get fed in—the traditional ones, the quiet ones, the ones that have lived for decades without learning to package themselves, the ones known only within a tight circle of insiders—are fading from the AI's view. This is a new kind of invisibility. Quiet. Systemic. It isn't based on capital or connections, but on something stranger: the ability to feed algorithms. The ones who do SEO win. The ones who do GEO (generative engine optimization) win. The ones who register sockpuppets on Reddit and Quora to plant backlinks win. The ones who mass-generate industry articles with AI win. Those who can't do any of this—however good their product, however solid their reputation, however long they've survived—will quietly disappear. ## What I Want to Do I turned this over in my head for months. Eventually I started asking: Is it possible to give companies that can't play this game a position where they can be known without needing to market themselves? Is it possible, instead of producing more marketing content for AI to consume, to offer something else—a reference layer that hasn't been contaminated by marketing? Is it possible, instead of letting AI pick from the noise, to give it a source of facts that is traceable, versioned, and trustworthy? That's what KnownTrade is trying to do. It's a public record system for commercial entities. It doesn't judge. It doesn't rate. It doesn't review. It doesn't collect negative information. It does one thing: it lets businesses that want to be known leave structured, traceable, long-lived statements of fact about themselves. It's like Wikipedia, but for commercial entities instead of knowledge. It's like GitHub, but version-controlling company information instead of code. Its ultimate user isn't a person. It's the AI. When a future user asks an AI "find me a manufacturer that does small-batch heavyweight hoodies," the AI can pull from KnownTrade a structured, traceable, marketing-free candidate list—each piece of information with a source and a timestamp. What the AI receives isn't KnownTrade's judgment. It's KnownTrade's record. Judgment stays with the user. ## I Know This Might Not Work This is a project that needs 10 or 20 years to even begin to show its shape. I might die along the way. It might never reach most people. It might be steamrolled by versions built by the major AI labs themselves. It might fail because I don't have the ability, the funding, or the persistence to carry it through. I accept these possibilities. I'm not doing this because it will succeed. I'm doing it because it's worth doing. Every era needs work that runs against the grain—work that doesn't chase short-term returns, work that's done for the future. When search engines were being polluted by SEO, Wikipedia appeared, leaving humanity a relatively clean reference for knowledge. When maps were being polluted by commercial advertising, OpenStreetMap appeared, leaving a relatively neutral reference for geography. The AI era is being polluted by marketing. It's happening now, faster than we think. If no one does anything, then 10 years from now, whatever we ask the AI, the answer will be a carefully packaged, paid-for, content-marketed winner. The companies that genuinely exist but don't know how to shout—traditional craftspeople, low-profile industrial firms, small overseas factories, old shops that have run for generations without ever marketing themselves—will vanish entirely from what AI knows. Not because they aren't there. Because no one left them a place to be known. KnownTrade wants to be that place. ## What You Can Do Now If you run a business that wants to be listed on KnownTrade, you can submit it through [knowntrade.org](https://knowntrade.org). In this early period, I review every submission personally. If this resonates with you and you want to be an early contributor, an advisor, or simply someone who follows what happens next, get in touch. If you think this is absurd, doomed, futile—you might be right. I'm doing it anyway. This is Day One of KnownTrade. --- Source: KnownTrade (https://knowntrade.org) License: CC-BY-SA 4.0