Who Owns What in an AI Search Programme
Give each recurring job one accountable owner, and keep the person measuring separate from the person whose work is measured. The ten jobs: the prompt set, sampling runs, the surface register and platform change log, first-party data archives, the prompt-to-page map, content changes, technical fixes such as crawlability and rendering, corrections to wrong answers, the monthly operating pass, and the quarterly review. Two sit awkwardly between roles and are worth assigning first: the platform change log, because watching vendors is nobody's day job, and corrections, because they span content, communications and support. Titles matter less than the separation: whoever ships the pages shouldn't be the only person computing whether the pages worked.
If your AI search programme started as one person's side project, that's fine until the same person writes the pages, runs the samples and presents the result. Nobody's being dishonest. The structure just guarantees that every number passes through the one person with a stake in it.
A RACI for a small in-house team
Accountable means one person who answers for it; responsible means whoever does the work. The roles are illustrative; map them to your own.
| Job | Accountable | Responsible | Consulted | Informed |
|---|---|---|---|---|
| Prompt set and versions | Measurement owner | Measurement owner | Sales, support | Content lead |
| Sampling runs | Measurement owner | Analyst or tool | Engineering | Content lead |
| Surface register and change log | Measurement owner | Analyst | Engineering | Marketing lead |
| First-party AI data archives | Analytics lead | Analyst | Measurement owner | Marketing lead |
| Prompt-to-page map | Content lead | Content team | Measurement owner | Marketing lead |
| Content changes | Content lead | Writers | Measurement owner | Marketing lead |
| Technical fixes | Engineering lead | Engineers | Content lead | Measurement owner |
| Corrections to wrong answers | Marketing lead | Content or comms | Support, legal | Sales |
| Monthly operating pass | Measurement owner | Analyst | Content lead | Marketing lead |
| Quarterly review | Marketing lead | Measurement owner | Content, engineering | Leadership |
The one separation that matters
Content changes are accountable to the content lead; the prompt set and the sampling are accountable to the measurement owner. Keep those two apart, even in a team of three. The measurement owner is the person who can say 'the control set moved just as much' without it being a judgement on their own work.
When one person has to do both
In a team of one, add a second reader rather than a second owner. Once a month, someone else reads the report and asks the control-set question and the 'what did we exclude' question. It's a short read, and it removes most of the risk the separation exists to remove.
The owner card
Job: ______
Accountable: ______ (one name)
Responsible: ______
Cadence: weekly / monthly / quarterly
Inputs: ______
Outputs: ______ (where they're stored)
If this person is away: ______
Hand-over notes: ______One card per job, kept with the programme's other records. The last two lines are the ones that save a quarter when someone leaves.
FAQ
Can a tool be 'responsible'? — For running samples, yes. A tool can't be accountable: someone has to answer for the numbers it produces, including the day it's wrong.
Who owns the budget? — Whoever is accountable for the quarterly review, because that's where spend gets justified or cut.
What about agencies? — The same table, with a column for who sits on which side of the contract. Approvals and corrections to public claims usually stay with the client; measurement can sit with either, as long as it isn't with the team being measured.
Put this to work on your own website.
Fenn finds what your customers ask, drafts the articles and site fixes, and measures what ChatGPT, Claude, Gemini, Perplexity and Grok say about you — with every change waiting for your approval.