For twenty years, search visibility was a gradient. Position one was best, position five was decent, position eleven was disappointing but not fatal. Some people clicked through, some scrolled, some opened six tabs and compared.
A generated answer is not a gradient. It is a cliff.
Two or three names, and then nothing
When an assistant answers a hiring question it typically returns two or three options. Occasionally one. There is no page two, no scroll, no fourth result waiting below the fold for a diligent researcher.
So the distribution of outcomes changed shape. Under ranked search, being the eleventh-best-optimised business in your category earned you a small but real share. Under generated answers, it earns you nothing — and crucially, you never find out. There is no impression, no click, no line in a report. The deal simply happened somewhere else.
Rankings have not stopped mattering
This is where the argument usually overreaches, so let us be precise.
Traditional search still drives enormous volume, and it is the substrate everything else sits on. A page an assistant cannot crawl cannot be retrieved. A site with broken indexing is invisible to both systems simultaneously. Technical SEO has not become less important; it has become the price of entry to a second competition rather than the whole game.
The honest framing is not that rankings are dead. It is that ranking well is now necessary and no longer sufficient.
Three things that behave differently
The mechanics diverge in ways that change what you should actually do.
- Corroboration outweighs self-description. Models weight agreement between independent sources far above anything you publish about yourself. A claim only you make is a weak signal regardless of how well it is written.
- Passages compete, not pages. Assistants retrieve chunks. Your strongest page may be retrieved for one paragraph while the rest is never read — so every paragraph has to survive being read alone.
- Being wrong is worse than being absent. An assistant confidently repeating stale information about you does damage that new content corrects slowly, because the incorrect version is already corroborated somewhere.
What this means for budget
Not that you should move spend from SEO to something new. Most of the work overlaps: entity consistency, structured data, crawlability and clear content serve both systems.
The genuine additions are narrower than the marketing around them suggests. Measure what assistants currently say about you. Restructure key pages so the answer sits at the top rather than at the end. Source your statistics so they can be verified. Build corroboration deliberately rather than hoping it accumulates.
That is a meaningful programme, but it is an extension of good search practice rather than a replacement for it. Anyone selling it as a wholesale reinvention is selling the framing, not the work.
The uncomfortable timeline
Technical fixes register in weeks. Answer capture moves in one to three months. Assistant citations are the slowest thing in the discipline, because they depend on independent sources updating on their own schedule.
Three to six months is realistic. In entrenched categories, longer. That is a worse pitch than most agencies want to make, which is precisely why so few make it — and why the ones who do are usually describing the work accurately.