The skills this work needs are not the ones a classic SEO job description asks for. The shortage is in people who can design a test and read a result, not people who can run an audit.
What to look for
Three capabilities carry most of the value. Someone who can turn a vague question into a measurable one. Someone comfortable enough with data to resist a chart that flatters. And someone who can brief an agent properly, which is closer to management than to prompting.
Classic technical SEO still matters and is the easiest of the three to hire for.
Train rather than hire, mostly
Most teams are better off training existing people than hunting for an AI SEO specialist. The underlying discipline is familiar; what is new is the measurement and the agent work, and both are learnable in weeks by someone who already understands search.
Table stakes
- One person accountable, with time protected for it.
- A documented method, so the work survives that person leaving.
- A ladder that rewards running experiments, not shipping volume.
- Budget for model spend treated as a running cost, not a project.
- A written standard for what good looks like, so review is not taste.