AI visibility fails as a solo marketing project because most of the inputs belong to other teams. The sources models cite are frequently not yours.
Who owns what
PR and comms own third-party corroboration: the trade press, the roundups, the analyst mentions that models read as evidence. Product owns the facts being described, and tells you when they change. Sales hears the real buying questions long before they appear in any keyword tool. Support knows which answers are wrong in the wild.
Marketing owns the pages and the measurement, which is a smaller share of the work than it first appears.
Make it one loop
The habit that makes this work is boring: the buying questions sales hears become tracked prompts, the answers models give become a monthly read-out everyone sees, and anything factually wrong in an answer becomes a job for whoever owns that fact.
Table stakes
- A named owner in each function, not a committee.
- A monthly forum where the actual answers are read aloud.
- A route for sales to add a prompt they keep hearing.
- A correction path when a model states something false about you.
- One shared definition of recommended, so nobody argues about the chart.