In short: Ask for the smallest piece of work that produces evidence, not for a programme. One category, one market, ten questions, one quarter, with a stated cost and a stated condition that would make you stop.
A vague ask is refused for a reason that has nothing to do with GEO: nobody can tell what they are agreeing to, so the safe answer is no.
Four people usually have to agree and only one of them controls budget. Finding the other three before the meeting is most of the work.
Why Big Asks Get Refused
"We need to invest in AI visibility" names no scope, no cost, no end and no test. Approving it means approving something undefined for an indefinite period, and a careful person declines that regardless of the subject.
It also fails a quieter test. A request with no stopping condition sounds like the start of a permanent line item, and permanent line items are what a budget holder is paid to be suspicious of.
The ask that works is boring: a fixed scope, a number of hours, a date, and a sentence saying what result would end it. Boring is the point. Everything interesting about the field belongs in the evidence you showed them, not in the request.
The Shape Of The Ask
One category and one market. Retrieval barely overlaps between languages and markets, so a pilot spread across four of them produces four thin results instead of one readable one. Pick where the revenue is.
Ten questions, fixed in writing. They become the measurement, and their value depends on nobody editing them later. Ten is small enough to run by hand and large enough that one odd answer does not dominate.
One quarter. Long enough for the fast steps to land, short enough that somebody will actually look at the result. Say in advance that the slow step will probably not have moved yet.
A named cost. Hours, whose hours, and any spend. Vagueness here reads as an attempt to avoid scrutiny, and it is the easiest thing in the whole request to be exact about.
A stopping condition. One sentence: what result would make you recommend not continuing. This is the part that converts the request from an open commitment into a bounded experiment, and it is the part most people leave out.
Who Has To Agree
Usually four, and only one of them is the budget.
Whoever owns the site's technical work. Nothing in the first pass happens without them, and most of it is small: a robots.txt review, a rendering check, a look at what the logs say arrived.
Whoever owns content. The extraction work is editing what exists rather than writing more, which is a smaller ask than it sounds and a different one than they will assume.
Whoever owns the brand's name and description. One canonical name used everywhere is an entity decision and a brand decision at the same time, and it is the one most likely to stall if you meet this person in the approval meeting rather than before it.
Whoever owns search reporting. Their numbers can fall while presence rises. Bring them in early enough that the change in reporting is something you designed together rather than something that happened to them.
The blocking decision people forget: if the site restricts AI crawlers, that is a commercial policy somebody set, and reversing it is not a technical ticket. Find out what the file says and who decided it before you promise anybody a result, because access removed acts quickly and access restored takes as long as the crawler takes to return.
What You Commit To In Return
Three things, all of which cost you nothing if the work is honest.
A measurement that does not change between quarters, so the second reading is comparable to the first. A report that names its limits, which is the subject of The Report Somebody Can Check. And no claim of a revenue effect you cannot trace.
Committing to those in advance is unusual enough to be persuasive on its own, and it costs you only the option of overclaiming later.
If The Answer Is No
Ask for the hour instead.
Running ten questions and recording what came back needs no budget and no permission from anybody. Do it anyway, keep the sheet, and bring it back in a quarter with a second reading beside the first. A refusal is often a refusal of the size of the request rather than of the subject, and a second data point turns the same conversation into a different one.
Key Takeaways
- Ask for a bounded experiment: one category, one market, ten fixed questions, one quarter, a named cost, and a stopping condition.
- A request with no end date reads as a permanent line item, which is what a budget holder is paid to refuse.
- Four people usually have to agree: technical, content, brand, and whoever owns search reporting. Only one holds the budget.
- Find out what your robots.txt says and who decided it before promising anyone a result.
- Commit in return to an unchanged method, a report that states its limits, and no revenue claim you cannot trace.
Check yourself
Before you move on
Not scored, not recorded, and not part of the certificate. Both answers are settled by a sentence in this lesson, and the reasoning appears whichever option you pick.
- 01
Which part of the ask most often gets left out, and turns an open commitment into something approvable?
- 02
Besides whoever holds the budget, who most needs to be found before the approval meeting rather than in it?