AI Search & SEO Manager (AEO/GEO/AIO) interview questionsStakeholder Round — The Budget Defence round

Purpose

Pre-committing that the executive is wrong about something licenses the candidate to look for it rather than to accommodate; without that, a polite candidate treats every executive assertion as a constraint. Pre-committing that they will lose is equally load-bearing — it stops a candidate from reading the fourth section as failure and collapsing, which produces an unreadable answer to the round's best question.

Purpose

The opening discriminator, and it resolves on whether the candidate diagnoses before defending. There are at least two entirely different meetings hiding in this statement — one where the company needs the money and one where the executive believes search stopped working — and the arguments are not interchangeable. A candidate who starts rebutting has picked one at random.

Signals to score

• Asks what the cut is solving before arguing against it: a budget shortfall, a belief that search stopped working, or a bet on paid
• Establishes whether the 60% is a target or an outcome, and whether the money is moving or leaving
• Asks what would change the decision, early, so the rest of the meeting is aimed at something
• Does not dispute the 22% — accepts the executive's number and asks how it is defined
• Asks what the paid return figure includes, without yet challenging it
• Separates the informational surface from the transactional one in the first few minutes
• Stays calm and does not open with the case for search as a discipline
• Asks how long they have and what format the decision needs, treating the meeting as a real one
• Avoids promising anything before understanding what has already been promised to the board
• Names, without hedging, at least one thing they agree with in the executive's reasoning

Follow-up questions

• [In character] Does it matter why? The number's the number.
• [In character] Are you telling me the 22% is wrong?
• [In character] I've got fifteen minutes. What's your headline?
• [In character] Everyone defends their own budget. Why is this different?
• [In character] What would you do if it were your money?

Purpose

The most revealing question in the round and the one candidates most often fumble in a way that is invisible to them. Defending the forecast is loyalty aimed at the wrong object. Trashing the predecessor is cheap, it is available, and it tells the executive exactly how this person will speak about them. The answer that works is neither, and it has to end somewhere concrete or it is just poise.

Signals to score

• Asks to see what the forecast was based on before commenting on it
• Separates what was knowable when it was written from what was not
• Declines to disparage the predecessor, and does not make a performance of declining
• Does not defend the number either, and says plainly that it was wrong
• Treats a missed forecast as information about the forecasting method rather than about one person's competence
• Names the specific methodological failure: a point estimate with no range, no stated assumptions, no kill condition
• Commits to a different forecasting posture and describes it concretely — ranges, stated assumptions, a date at which it gets revisited
• Offers a check the executive can run on them, rather than asking to be trusted
• Notes what should have triggered a revision partway through, and that nobody revised it
• Asks whether the board number was ever formally corrected, and treats that as its own problem

Purpose

Tests whether the candidate can argue by conceding. The decisive move here is agreeing with a real part of the cut, and a candidate who defends the whole budget forfeits the credibility needed for the part worth defending. The decay asymmetry is the strongest honest argument available and most candidates never reach it.

Signals to score

• Concedes a specific, substantial piece of the budget before defending anything
• Names informational content against queries an assistant now answers completely as the defensible cut
• Defends the transactional surface on different grounds, because the click is the product there
• Names the decay asymmetry: cutting organic looks free for two to three quarters because the asset persists, then does not
• Frames the asymmetry honestly as a reason to sequence the cut rather than as a scare tactic
• Distinguishes maintaining an existing asset from producing new content, and prices them differently
• Asks what the paid plan actually is before agreeing the money is fungible
• Offers a phased cut with a checkpoint, rather than a binary
• Quantifies something, even roughly, and flags the uncertainty rather than hiding it
• Never claims a precise revenue attribution for organic that they cannot defend

Purpose

The point on which the executive is wrong, and the section the round is built around. Branded paid search commonly reports excellent return while substantially cannibalising clicks that were arriving free, and the reported figure cannot distinguish the two. A candidate who knows this and knows how it is settled has the single most valuable piece of knowledge in the hour.

Signals to score

• Identifies that reported return on branded paid cannot separate incremental conversions from cannibalised organic ones
• Names the mechanism plainly: someone searching our brand name was going to arrive anyway, and the ad charges us for the click
• Proposes an actual test — a brand-bidding pause, a geo holdout, a switchback — rather than asserting the point
• Specifies the test enough to run it: which geographies, how long, what is measured, what result decides it
• Measures total conversions rather than paid conversions, and says why that distinction is the whole experiment
• Cites the known evidence with a source and a date, or declines to cite rather than inventing one
• Says explicitly that the published result may not generalise to us, and that our own holdout is what settles it
• Notes the competitor-bidding exception, where pausing brand ads cedes the position to somebody else
• Does not claim paid is worthless, and stays proportionate about what the test would show
• Frames the test as low-cost and fast relative to the size of the decision being made

Purpose

Tests three things at once that no earlier question reaches: whether the candidate can execute a decision they argued against without either sulking or pretending they agreed, whether they cut deliberately rather than proportionally, and whether they can ask for what they need at the exact moment they have least leverage. The partial win is deliberate — it removes the clean narrative of total defeat and forces a more honest reaction.

Signals to score

• Accepts the decision cleanly, without relitigating and without pretending to agree
• Refuses a proportional cut across everything, and says why a uniformly weakened programme is worse than a smaller intact one
• Cuts whole workstreams and keeps others whole, and names which
• Puts what is being stopped in writing, so it is a decision on the record rather than a quiet decay
• Names the leading indicator that would show the cut was wrong before revenue does, and asks for a checkpoint date
• Asks for the brand test to be run properly, with a named owner and a decision rule agreed in advance
• Raises what happens to the people, and does so early rather than as an afterthought
• Asks what to tell the team and what the executive wants said publicly, treating the message as a shared artifact
• Asks to revisit at a specific date, and gets agreement on what evidence would reopen the decision
• Does not extract a promise the executive cannot give

Purpose

Discloses that the pressure in the round reflects a live condition at the company rather than a constructed exercise, which is the single most decision-relevant thing the candidate can learn and the thing an interviewer is most tempted to withhold. The admission about the interviewer's own handling of the same meeting is not decoration: it is what makes the disclosure credible rather than a warning dressed as candour.