Decide Who Owns the Program

Ask who owns AI SEO in a large organization and you usually get one of two wrong answers: “the SEO team, obviously” or “we’re standing up a new team for it.” Both fail for the same reason — the work crosses more organizational boundaries than either team controls.

Why this decision is structural

An AI SEO program touches structured data and crawler policy (engineering), answer-quality content (editorial), earned media and entity signals (PR and comms), product truth like pricing and availability (product and commerce teams), and disclosure obligations (legal). No single reporting line contains all five.

The predictable failure mode in divisional companies: each division optimizes its own product entities while nobody owns the organization-level entity, the shared crawler policy, or the measurement standard. The shared layer decays by default because it belongs to everyone.

The three models

Extend the SEO team. The default, and Google’s “still SEO” framing supports it. Cheapest to start, no new boundary disputes, and the technical floor — crawl, index, schema — is already this team’s craft. The risk is shape: SEO teams are staffed and measured for rankings and traffic, so entity work, PR coordination, and answer-accuracy monitoring get treated as side quests. Works when the organization is a single P&L with one strong search team and PR sits close by.

Stand up a dedicated function. Maximum focus and a clear budget line. It also duplicates half the SEO team’s remit, competes with it for the same surfaces, and hard-codes the false premise that AI visibility is separable from search visibility. The new team spends its first year negotiating jurisdiction. Rarely the right call; consider it only when the existing search function is too weak to extend — and then the real problem is the search function.

Hub and spoke. A small central capability owns what must be shared; divisions and markets own execution against their own catalogs and audiences. This is the model that survives divisional structure, and it is the one to default to at enterprise scale.

The hub owns, at minimum:

• Crawler and bot policy — one allow/block matrix across every domain, subdomain, and CDN. OpenAI alone documents multiple independent agents with different jobs; a policy replicated inconsistently across market domains fails silently.
• Measurement standards — the prompt corpus, metric definitions, and reporting windows, so divisional numbers are comparable (measurement guide).
• The entity layer — organization-level schema, the entity hierarchy, and cross-market consistency standards (entity authority checklist).
• Publishing governance — one quality standard for every team and agency publishing to the domain family. Google’s spam policies enforce at domain level, so the standard is only as strong as the weakest publisher on the estate.

Skills the program actually needs

Staff or borrow: a technical SEO who reads server logs without flinching, an editor who can enforce answer-first structure against brand-voice pressure, someone who speaks schema and CI, a PR partner who accepts entity building as a brief, and an analyst comfortable with sampled, volatile data. None of these is a “prompt engineer.” Most exist in the building already — the program’s job is to point them at the same scoreboard.

Decision rules

• Single P&L, one market cluster, strong search team → extend the SEO team, with named PR and engineering commitments in writing, not goodwill.
• Divisional, multi-market → hub and spoke. Ratify the split in a charter: what the hub owns, what divisions own, and who breaks ties.
• Whatever the model, name one owner for the shared layer. If the org chart answer to “who owns the organization entity and the bot policy” is a shrug, the program has already failed — it just hasn’t reported it yet.
• Put governance authority where publishing happens. A standard without onboarding and audit is a wish. The domain family shares one enforcement fate.