AI visibility fails at the seams between teams, because the fixes live in four places at once: content, structured data, third-party review profiles, and product documentation. Unowned, it defaults to nobody.
The work is not hard. It is just distributed across four backlogs that do not share a standup, which is a different problem and a worse one.
Nobody refuses to do this work. It stalls because each of the four pieces belongs to a team that reasonably considers the other three somebody else’s, and the measurement that would show the gap sits in a fifth place.
Why does it fail at the seams specifically?
Because the causes and the fixes sit in different departments, and the symptom appears somewhere neither of them looks.
An engine describes your pricing wrongly. The cause is that the price exists only inside a styled table on a page product owns. The fix is a sentence. The person who noticed is in demand generation, the page belongs to web, the number belongs to finance, and the measurement lives in a tool nobody has budget for. Four handoffs for a one-line change.
That pattern repeats across every category of failure, and it is why the work looks harder than it is. Almost none of these fixes are technically difficult. Nearly all of them cross a boundary.
Where do the four fixes actually live?
Each one belongs to a different team for a documented reason rather than a political one.
| The work | Why it lives there | Usually owned by |
|---|---|---|
| Crawler access | OpenAI runs separate crawlers for search and for training, and its documentation notes each robots.txt setting “is independent of the others.” Blocking one does not block the other, and getting it wrong is invisible. | Web infrastructure or platform engineering |
| Retrievable content | Google states a page “must be indexed and eligible to be shown in Google Search with a snippet.” That is ordinary indexing work, not a new discipline. | SEO, or whoever already owns indexing |
| Structured data | Organization markup “can help Google better understand your organization’s administrative details and disambiguate your organization in search results.” | SEO with web development |
| Third-party profiles | Comparison pages and review platforms are written about you, and they answer the questions buyers actually ask an engine. | Product marketing, comms, or customer marketing |
| Product documentation | Integrations, limits and compliance claims are stated most precisely in docs, and docs are usually the most retrievable thing a company publishes. | Product or technical writing |
Read the second column and one thing stands out. Google’s documentation on AI features is explicit that there is nothing exotic to build: “There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary,” and “You don’t need to create new machine readable files, AI text files, or markup to appear in these features.”
There is no new technical discipline to hire for. There is an ownership question nobody has answered.
What can each team do alone, and where does each one stop?
- Can do alone
- Indexing, snippet eligibility, structured data, most of the retrievability work. This is the largest single share.
- Stops at
- Third-party profiles and product docs, neither of which sits in their backlog or their remit.
- Can do alone
- Category language, comparison pages, review platform profiles, the positioning an engine has to work from.
- Stops at
- Anything requiring a deploy, and anything requiring robots.txt.
- Can do alone
- The precise, checkable facts: integrations, limits, compliance, what the thing actually does.
- Stops at
- Knowing any of it is wrong in an engine, because nobody shows them.
- Can do alone
- Getting corrections made on sources you do not control, which is the slowest and highest-yield work here.
- Stops at
- Everything technical, and usually at knowing which sources matter.
So who should own it?
SEO, in most companies, and the reason is unglamorous: they already own the largest share of the work and the only measurement discipline that transfers.
Indexing, crawler access and structured data are already theirs. Google’s own framing supports this, since eligibility for AI features is the same eligibility as for Search. Handing it to a team that does not already own indexing means teaching them indexing first.
The two arguments against are real and worth stating. SEO teams report on rankings and sessions, and neither survives contact with an answer that produces no click, which is the reporting problem in what is zero click search. And third-party profile work is genuinely not an SEO skill.
So the honest arrangement is SEO owning the outcome with two standing dependencies, not SEO owning everything. Which is a different sentence from the one most companies say out loud.
What does the owner actually need to be given?
Three things, and the third is the one that gets forgotten.
On the first, insist that answers are stored rather than only scores. An owner who can show the sentence an engine produced about your product wins arguments that a percentage cannot, and the run count is what makes the percentage defensible at all, covered in why ChatGPT gives different answers to the same question.
What about hiring for it?
Premature in almost every company, and the reason is in the mechanism rather than in the budget. If Google states plainly that no new markup, files or optimisations are required, then the missing thing is not a skill your team lacks.
A dedicated hire makes sense at one point: when the measurement is running, the dependencies are agreed, and the backlog produced by that measurement is genuinely larger than the owner’s remaining capacity. Hiring before the measurement exists produces somebody whose first six months are spent building the case for their own role.
The more common gap is smaller and more awkward to name. Somebody needs about a day a week, permanently, and most companies would rather discuss a headcount than find a day.
How do you tell whether it is unowned right now?
Four questions, and if any of them stalls you have your answer.
Who would notice if an engine started quoting a price you retired last year? If the honest answer is a customer, nobody is watching.
Who can change robots.txt this week? Not who has permission in theory, who has done it recently.
Who owns the sentence describing what you do, on the review platform your buyers read? This one usually produces the longest pause.
And who reports the number? If the answer is that it appears in a slide once a quarter and nobody is accountable for it moving, the work is unowned regardless of what the org chart says.
None of this requires reorganising anything. It requires one named person, a measurement they can point at, and two teams agreeing to take tickets. The failure mode covered in third party sources ai citations audit is what happens when that person does not exist, and the arithmetic that tells you whether the measurement is worth trusting is in how to calculate AI share of voice.