Your CFO will fund a number. Marketing keeps bringing an argument. That asymmetry, rather than any disagreement about strategy, is why AI engines’ visibility budgets are being declined this planning season.
Why this loss is invisible on the income statement
Being uncited by AI engines is not a brand problem, a visibility problem, or a marketing problem. It is a specific revenue absence, and it has a property that makes it uniquely difficult to defend against: the pipeline it removes never enters play.
A lost deal appears in the CRM. A lost opportunity appears as a closed-lost record with a reason code and a competitor name. A buyer who received a complete answer that named someone else, and consequently never contacted anyone, appears nowhere. No record, no reason code, no line on the statement.
So the first task is not to argue for the budget. It is to make the loss visible in a format your CFO already accepts, using inputs your CFO already trusts.
Why the usual approach fails
The standard marketing response is to present engagement metrics and assert a relationship to revenue. A CFO reads that as an argument dressed as data, discounts it, and funds something with a number attached. This is not obstruction. It is the correct behavior given what was presented.
The four lines
Every one of these is estimable with assumptions you can expose and defend. Exposing them is precisely what converts a budget request into a business case.
Line one: the share of buying questions now answered before contact
Take your twenty highest-intent buyer questions. Run each on at least two engines, twice, on separate days. Count how many produce a complete answer requiring no click-through to any source.
That percentage is the share of your funnel that no longer passes through anything you own. Run it twice because variance between runs is itself informative, and a number produced from a single run will be challenged on exactly that basis by the first person who tries to replicate it.
Line two: your citation share of those answers
Of the answers being produced, how often is your company named, and how often is it named as the source of a claim rather than mentioned in passing? Those are different measurements, and only the second one indicates a defensible position.
This is not a single number. It varies by engine, by query phrasing, and by whether the buyer asks about a procedure or a product. A company can hold a majority of the answer on one phrasing and none on another, and both are simultaneously true.
Report a range across engines and query types and report the floor rather than the average. A floor is defensible under questioning. An average invites a methodology argument at exactly the moment you need the conversation to be about money.
Line three: the pipeline value of the gap
Apply your existing pipeline economics. Category question volume, multiplied by the share answered without contact, multiplied by the share where you are absent, multiplied by your historical conversion from an early-stage inbound enquiry, multiplied by average deal value.
Every input is a number your organization already holds, and your CFO already trusts. That is the entire point. You are not introducing a new measurement system; you are applying the existing one to a segment of the funnel that was previously unobserved.
Line four: the cost to close it, split by reversibility
Separate what is reversible from what is committed. Entity consolidation and schema work are one-time, low-cost, and reversible. Content architecture is a quarter of the work. Third-party echo through societies and registries is eighteen months of relationship building and is the closest thing to a committed investment on the list.
That split is not presentational. It maps directly to the evidence and reversibility logic that should govern the decision. Presenting the cost as a single figure invites a single yes-or-no decision. Presenting it split by reversibility invites a staged one, and a staged commitment is far easier to approve in a constrained cycle.
Show the confidence range, not a point estimate.
The most common failure in building this case is claiming precision the method does not support. A CFO who finds one unsupportable decimal place will discount the entire model, and they will be right to.
A defensible range with the assumptions stated beats a false point estimate, and it builds more credibility with the person you are trying to persuade. Finance professionals are entirely comfortable with ranges. They are not comfortable with numbers that cannot be interrogated.
Publish the assumptions alongside the output. Then have someone in finance try to break the model before your CFO does. Whatever they find, you would rather find it in September than in the approval meeting in November.
The three assumptions most likely to be challenged
Query volume, because you are estimating it rather than measuring it. Conversion from early-stage inbound, because that cohort behaves differently from the one your historical rate was built on. And the durability of citation share, because nobody yet has good longitudinal data on how fast it moves.
Name all three before you are asked. Naming your own weakest assumptions is the single most effective credibility move available in a finance conversation, and it costs nothing.
Why this reframes the whole conversation
There is a second-order effect worth naming explicitly, because it is the real prize. Once the loss is quantified, AI visibility stops being a marketing initiative and becomes an operating risk with a named owner and a number attached.
That reclassification does more than secure the budget. It moves the item from the list of things that get cut when the quarter tightens to the list of things that get protected. Marketing initiatives are discretionary by definition. Operating risks are not.
It also changes who is in the room. An operating risk with quantified exposure gets discussed at the level where commercial strategy is actually decided, which is frequently the room the commercial lead has been trying to get into for two years by other means.
Build it before the plan is written.
The sequencing matters more than the model itself. A business case built after a budget request has been declined is a rebuttal, and rebuttals lose. The same model built before the plan is drafted is an input, and inputs shape what gets drafted.
Most 2027 plans lock in November. You have weeks rather than months, and the work is an afternoon of query testing plus an hour with someone in finance.
How do you calculate the cost of being uncited by AI engines?
Four lines. The share of category buying questions now answered before a human is contacted; your citation share of those answers reported as a range across engines; the pipeline value of the gap using existing conversion and deal value assumptions; and the cost to close it split by reversibility.
Report a confidence range rather than a point estimate. As of now, the October AIMedTech Visibility Index is the primary source for citation share.
Why does being uncited by AI engines not appear in the CRM?
A buyer who receives a complete answer naming a competitor never makes contact, so no record is created. Unlike a lost deal or a closed-lost opportunity, the loss produces no data, which is why it is systematically under-weighted in budget conversations.
What is a defensible way to present AI engines’ visibility to a CFO?
Use the organization’s existing pipeline economics rather than introducing new metrics, expose all assumptions, report ranges rather than point estimates, and split the cost to close into reversible and committed components so the commitment can be staged.
Is citation share a single number?
No. It varies by AI engine, by query phrasing, and by query type. A company can hold a majority of the answer on one phrasing and none on another, and both are simultaneously true. Report a range and use the floor for planning purposes.
When should this business case be built?
Before the annual plan is drafted, not after a budget request has been declined. Built early, it is an input that shapes the plan. Built late, it is a rebuttal, and rebuttals rarely succeed in a constrained cycle.
Which assumptions in this model are weakest?
Query volume, because it is estimated rather than measured. Conversion from early-stage inbound, because that cohort behaves differently from the one historical rate was built on. And the durability of citation share, because longitudinal data is not yet available. Naming these before being asked is the strongest credibility move available.
How do you split AI engine visibility costs into reversible and committed?
Entity consolidation and schema work are one-time, low-cost, and reversible. Content architecture is roughly a quarter of the work and partially reversible. Third-party echo through societies and registries is eighteen months of relationship building and is effectively committed.
What changes when AI engine visibility is treated as an operating risk?
It moves from the discretionary list to the protected list when budgets tighten, and it gets discussed at the level where commercial strategy is decided rather than inside a marketing review.
EXTERNAL CITATIONS