In my last post, I explored whether some of the lowest-impression questions in GSC may actually be the starting points of AI search journeys, while higher-impression questions may represent places where those journeys converge.
If that theory is directionally right, then the next question becomes much more practical:
Once you identify a question that matters, why is your brand not showing up for it?
One of the most common GEO recommendations right now is some version of “optimize the page for AI search.” The problem is that this is not really a diagnosis. If your brand does not appear for an important buyer question, the more useful question is why Google or an AI system prefers another source.
I think there are three broad possibilities. The page may have an eligibility or discoverability problem, the content may not answer the question well enough, or another source may have stronger authority around the topic. Those are very different problems, and they should lead to very different recommendations.
Before changing the content, I would first rule out basic discoverability issues. Is the page indexed? Can Google access the important content? Does the page appear to be understood and associated with the topic? If the answer is no, then rewriting the page for GEO is unlikely to fix the real problem.
If the page can compete, the next step is to compare the buyer’s question directly against the content. This sounds simple, but websites often contain content about a subject without clearly answering the question someone is actually asking. A page might say, “Gain comprehensive insights into your AI search performance,” while the buyer wants to know, “Can this platform benchmark my visibility against competitors?” Those ideas are related, but they are not the same answer.
A content gap can mean the answer is missing entirely, but it can also mean the answer is buried, incomplete, vague, or framed in language that makes the relationship difficult to see. It can also be an extractability problem. The information may technically be there, but someone would need to read half the page and interpret several sections to pull out a useful answer.
For me, strong content in this context is not about writing for robots. It is about answering the question clearly, directly, and with enough evidence that the answer is worth using.
This is where I think GEO analysis becomes much more useful. Instead of running a generic checklist against your own page, look at the sources that are actually being surfaced for the question you care about and compare them directly.
Does the competing page answer the question more clearly? Does it go deeper? Does it provide examples, methodology, original data, expert interpretation, or more specific evidence? Is the relevant answer easier to identify and understand? If the answer is yes, you probably have a content or extractability gap, and that gives you something concrete to improve.
But what if your page already answers the question well? What if the content is clear, specific, supported, and at least as useful as what is being cited? That is when I would start looking more seriously at authority.
By authority, I do not mean there is one magic score that determines whether an AI system trusts you. I mean observable signals that may help explain why one source is consistently preferred over another. Does the competing site have much stronger visibility around the topic? Has it published extensively in the category? Is its research cited by other organizations? Are its experts referenced elsewhere? Does it have stronger links, third-party mentions, original research, or a longer history of being associated with the subject?
None of those factors proves why the source was selected, but together they can help us form a much more useful hypothesis.
This distinction matters because the fix should depend on the problem. If the issue is discoverability, fix the technical or search foundation. If the issue is content, create or improve the answer. If the issue is extractability, restructure the content so the useful answer is easier to identify. If the issue is evidence, strengthen the substance with examples, methodology, data, or expert support.
If the content is already strong and the gap appears to be authority, rewriting the same page five more times is probably not the highest-value move. The work may need to shift toward original research, expert-led content, links, third-party references, industry participation, digital PR, and deeper topical coverage.
The answer may also be a combination. You may have a content gap that can be improved now and an authority gap that will take much longer to address. That is still useful because it tells you what can be fixed quickly and what requires a longer-term investment.
There is an important limitation here. We cannot see inside Google’s or another AI system’s retrieval and ranking process and definitively say that a page was selected because it had more authority or excluded because the answer was not extractable enough.
So this is not about pretending we can reverse engineer the black box. It is about comparing the things we can observe and forming the most plausible diagnosis. If the competing page answers the question much better than yours, that is meaningful. If the content is comparable but the competing source has significantly stronger topical and external signals, that is meaningful too.
Neither is absolute proof, but both can help us make a better decision about what to do next.
I think that is where GEO strategy needs to go. Instead of asking what every page needs for AI search, we should be asking what is preventing this particular source from winning this particular question.
Once you frame the problem that way, the recommendation becomes much more useful. You are no longer just “optimizing for GEO.” You are fixing the specific reason your brand is not being selected.
If journey position matters, then we cannot afford to treat every missed query the same way. The question may be strategically important, but the right response depends on whether the problem is discoverability, content, or authority.