Most enterprise AEO and GEO projects fail before anyone writes a line of schema. They lose momentum in the first cross-functional meeting, because nobody owns the question of how company knowledge should be structured for machines to read.

Teams open by talking about tools and content. What schema do we need, how should the FAQ be written, does the site need rebuilding. The earlier questions go unasked. Who defines this content? Which version of a fact counts as the company's official one, and who decides? Who maintains it afterward? When two departments disagree, who has the authority to settle it? Skip those, and an AEO project quietly turns from content engineering into a fight over territory nobody wants to claim.

Nobody designed who owns this

AEO changes more than the words on a page. It changes who writes them, who approves them, who updates them, and who answers for consistency a year from now. Where those lines were never drawn, every proposed change triggers the same three responses: that isn't my job, we've always done it this way, that belongs to another team. Teams read this as an execution failure and try to fix it with better project management. The failure sits further upstream, in a design decision nobody ever made.

AEO asks a company to take knowledge scattered across departments, pages, and systems, and give it a structure an outside system can interpret reliably. Without an owner, that work does not get maintained. Without maintenance, it does not stay consistent. Without consistency, a company can publish endlessly and still fail to register with AI systems as one coherent entity.

What a first AEO meeting reveals

Nobody in the room objects to AEO. Each person understands it as something slightly different, and the gap shows up in three recognizable ways.

The decision-maker says little. That does not mean they hold little power. The person controlling budget and headcount often lacks the background to judge a technical structuring decision. That is where the governance risk sits: authority and comprehension have come apart. The people who understand the problem cannot approve the work, and the person who can approve it cannot evaluate what they are approving.

Owners defend their boundaries. The moment the discussion touches an existing workflow, people stop examining the problem and start demonstrating that the problem lies elsewhere. This has little to do with individual attitude. When accountability was never mapped, any request to change how work is done reads as an accusation that the work was being done wrong.

Vendors cannot supply what is missing. An outside consultancy can provide technical recommendations, diagnostics, and a schema roadmap. It cannot define who inside the company is accountable for what. Conflate the two, and you pay for a solution and receive a report — one more document telling an executive that the company's digital presence needs improvement. Who decides, who executes, who maintains it over time: no vendor answers those for you.

One schema field, and the system behind it

A company publishes a press release. In the article schema, does the author field carry a person's name or the company's?

The decision looks too small to argue about, and it reaches further than it looks. That field shapes how search engines and AI systems understand where the content came from. Put a person there, and the system may read the piece as an individual's opinion rather than an official company statement. Put the company there, and the company's entity definition in the knowledge graph has to be clear enough to carry it.

Which answer is correct matters less than whether the company has ever set the rule. Does this content type represent an individual view or an organizational one? Who has the standing to decide that? Should every official publication follow the same convention? Most companies discover, the first time they hit this, that their web governance model has no line item called knowledge ownership.

That is what misassigned accountability costs. A wrong page can be fixed in an afternoon. Having no basis on which to make even the smallest structural decision is the expensive part.

AI search has raised the price

Where AI systems generate the answers, unstructured content costs more than it used to. These systems favor information they can identify, cite, and attach to a known entity. If your content has no clear ownership, the systems will still answer questions about you. They will answer using competitors, third-party summaries, or stale data. Once a wrong answer has been repeated often enough, the cost of correcting it climbs.

Define yourself, or external systems will define you with someone else's material. There is now a name for what happens when machine-generated errors get cited and regenerated: AI slop.

Minimum Viable Governance for AEO (MVG-AEO)

If a company wants to start, what is the smallest governance structure it needs? My answer is Minimum Viable Governance for AEO. It handles one problem: keeping an AEO project from stalling in its first meeting.

1. What is the company optimizing for? Different AEO goals carry very different governance loads. Undefined, every department believes it is working on the same thing while nobody is accountable for the same outcome. Minimum: choose one primary goal, write it down, and put it in front of everyone involved.

2. Which departments does this touch? The reach of AEO routinely exceeds the initiating department's authority. Undefined, you discover halfway through that you need IT, legal, or a business unit that was never in the room. Minimum: map the impact before you start — whose workflow changes, and who holds a veto.

3. Who has final say? Most AEO projects have enough people to do the work. What they lack is someone who can settle a cross-functional dispute. Undefined, every contested decision waits. Minimum: name an arbiter with a written scope, before the project starts.

4. Which knowledge needs long-term maintenance? AEO structures are not build-once. FAQs, product data, and schema rules all decay. Undefined, the first wave of work is obsolete in six months with nobody watching. Minimum: list the content types that need upkeep, and give each one an owner and a review interval.

5. Are you buying technology, or buying governance? Companies assume that hiring a consultancy gives them governance capability. Often it gives them a diagnosis. Undefined, the recommendation list has no one to receive it. Minimum: separate, in the contract, what is advice, what is execution, and what requires an internal decision.

When a company starts writing for both people and machines, the pages are the smaller half of the rewrite. The division of responsibility around them is the larger one.

MVG-AEO is the smallest set of answers that lets the work begin: five questions, five clear answers, and a project that survives its first meeting.