Why do great products no longer sell themselves? Two forces arrived together. Building got cheap, so AI-assisted teams ship in weeks what took quarters, a feature advantage decays in months, and sameness is the default state of every category. And discovery compressed: buyers ask an answer engine and get a shortlist of a few names, assembled from what the models absorbed long before your landing page saw a visit.
The web's old middle, where a better product could earn its way up the rankings over time, is thinning out. What replaces it is owned audience, which survives every algorithm change, and distribution manufactured inside the product itself through artifacts users send, share surfaces placed where value peaks, and original data the market wants to cite.
Product people were raised on a comfortable belief: build something great and the product sells itself. It was never fully true, but for a decade the gap was covered by discovery infrastructure that rewarded quality. Search rankings, app store charts, review sites, word of mouth moving through team chats. That infrastructure is being rebuilt around AI, and the cover is gone.
Why did great stop being enough?
2 forces, arriving together. First, building got cheap. AI-assisted teams now ship in weeks what took quarters, so a feature advantage decays in months and sameness is the default state of every category. Second, discovery compressed. Buyers increasingly ask an answer engine and receive a shortlist of a few names, assembled from what the models absorbed long before your landing page got a visit. The web's old middle, where a better product could earn its way up the rankings over time, is thinning out. I wrote about measuring this shift in Attribution After the Click. This piece is about what to do.
Owned audience is a balance-sheet asset
The channels that survive every algorithm change are the ones you own: the list, the community, the operator's own voice. Audience compounds the way code compounds, slowly and then suddenly, and companies that never built it end up renting it back at ad-market prices. For a growth team this has a practical meaning: audience metrics deserve the same weekly review the funnel gets, because audience is now upstream of the funnel.
The product is a distribution engine
The strongest distribution is manufactured inside the product. Artifacts users create and send, which carry the product into rooms sales never enters. Share surfaces built where the value peaks, not where the design has spare space. Usage data turned into original insight the market wants to cite, which is also exactly what answer engines reward. A product team that ships features without asking what each one exports is leaving its cheapest channel unbuilt.
There is a machine-readable version of the same idea. Agents discover tools through registries and executable documentation rather than through a homepage, so the integration surface is a distribution surface, which is the argument in When Your Next User Is an Agent.
The personal version
The same logic applies to the people, not just the companies. A product person who publishes what they learn owns a channel no employer controls, and in a market where recruiters, investors, and language models all read the public record, operators who write compound while equally good operators who do not stay invisible. It is uncomfortable advice for builders who would rather let the work speak. The work no longer gets the chance to speak unless something carries it.
The order of operations reversed
The product still has to be excellent, because distribution without substance burns the trust it borrowed. But distribution stopped being what you bolt on after the build. It is a design input: chosen with the ICP, shaped with the roadmap, measured with the funnel. Product people who treat it as another department's job are handing away the half of the craft that now decides the outcome.