What AI search actually changed, and what it did not
Every few months someone announces that SEO is dead. It was dead when voice search arrived, dead when featured snippets arrived, and dead again when AI answers arrived. The honest position is that something real has changed, and it is narrower than the announcements suggest.
Informational queries lose clicks
Queries with a single factual answer were always weak traffic. A user asking for a conversion rate, a definition, or an opening time was never going to become a customer, and now they are less likely to click at all.
If a site's traffic was built on that kind of query, the decline is real and it is not coming back. If a site's traffic was built on commercial and comparison queries, the effect is much smaller, because those queries have an evaluation step that an answer box does not replace.
Being the source still matters
AI answers are assembled from sources, and those sources get cited. The practical work of being citable looks remarkably like the practical work of ranking: clear structure, direct answers, verifiable claims, and topical depth that makes a page an obvious source on its subject.
The tactics have not been replaced. Their payoff has partly shifted from a click to a citation, which is harder to measure and easier to undervalue.
The measurement problem is real
The uncomfortable part is that attribution has got worse. A user who reads an AI answer, remembers a brand, and searches for that brand three days later shows up in your data as branded direct traffic with no visible cause.
Rather than inventing a metric to cover this gap, the more defensible approach is to track branded search volume and direct traffic as a trend, state plainly that some portion of it originates upstream, and stop pretending to a precision that does not exist.
What has not changed
A business still needs to be findable when someone in its city needs its service. A page still needs to load, be crawlable, and answer the question the visitor came with. A local business still needs reviews.
The fundamentals were never the interesting part, which is exactly why they keep surviving the announcements.