How to Handle a Client Who Demands AEO Guarantees You Cannot Make

"I need you to guarantee we'll show up in ChatGPT within ninety days, or I can't sign." It is a reasonable-sounding demand from someone about to spend real money, and it puts you in a bind, because the honest answer is that you cannot guarantee it and no one who is telling the truth can. The temptation is to soften, to say something that sounds like a yes without technically being one, and hope the ambiguity survives the engagement. It will not. The guarantee you fudge in the sales call is the reason you lose the client in month four, when the citation they think you promised has not appeared.

There is a way to handle this that keeps the deal and your integrity at once. It is not a clever workaround for making the promise. It is a redirect: move the client's demand off the thing no one controls and onto the things you genuinely can stand behind, then write those into the contract so the guarantee is real and specific rather than impossible and vague. This is how to run that conversation, what to guarantee instead, and how to tell the difference between a client you can win and one you should let walk.

Why the specific guarantee is impossible, said plainly

Do not dodge the reason. Name it, because a client who understands why the guarantee cannot exist stops reading your refusal as weakness. You do not control the output of a large language model. The engine decides which sources to surface based on its training, its retrieval index, its ranking rules, and the exact phrasing of a user's prompt, and all four sit outside any vendor's reach. The models also change constantly, and their answers are non-deterministic, so the same question asked twice returns different brand lists. Anyone guaranteeing a fixed citation is either not understanding the mechanism or hoping you do not.

This is the direct parallel to the oldest rule in search: reputable practitioners never guaranteed a number-one Google ranking, for the same reason, that the platform, not the vendor, decides placement, and it updates its ranking systems constantly, as Google's own guidance on core updates documents. The industry consensus that guaranteed rankings are a red flag rather than a selling point carries straight into AEO, and stating it out loud positions you as the practitioner who knows the terrain rather than the one overselling it. The variance itself has a mechanism worth explaining to a technical buyer, laid out in why AI answers change every time you ask the same question, and walking a client through it converts the refusal from a shrug into a demonstration of expertise.

Redirect the demand to what you can actually guarantee

A refusal that ends at "we can't promise that" loses the room. The move is to immediately hand the client a different set of guarantees, ones that are specific, enforceable, and genuinely valuable, so the conversation shifts from what you won't promise to what you will. Four categories hold.

Deliverables. You control the work, so guarantee the work in exact units. Not "we'll do AEO" but "eight answer-structured pages per month, a monthly citation-share report across four engines, one technical crawlability audit per quarter, and a standing prompt panel of forty tracked queries." A client can hold you to that, and it is the opposite of vague.

A measured baseline and transparent reporting. Guarantee that you will establish exactly where they stand today, run a fixed prompt set across ChatGPT, Perplexity, Gemini, and Claude, and report the raw numbers every month whether they moved or not. Guaranteeing transparency, including in a down month, is a promise most vendors will not make and a serious buyer respects.

Movement on leading indicators, framed as effort not outcome. You can commit to improving the signals that drive citations, crawler access, entity clarity, schema coverage, and answer-structured content on the priority topics, without promising the citation itself. This is the same distinction reputable agencies draw when they guarantee strategy quality and execution consistency rather than the ranking.

Proof on their own brand. The strongest guarantee-substitute is a test. Offer a paid pilot with a defined scope and a clear success metric measured on their data, so the client buys evidence before they buy the retainer.

The two columns to keep in front of you

Under pressure, the failure mode is drifting from the defensible promise to the flattering one. Keep this split visible through the whole negotiation, because the fastest way to lose a careful buyer later is to guarantee something today that the engine controls.

Every promise on the left is enforceable and honest. Every promise on the right requires you to control an output you do not. An engagement that lives on the left column survives the quarter where a citation is slow to appear, because you never staked the relationship on the one thing you could not deliver.

Write it into the contract so the guarantee is real

A redirect that stays verbal evaporates the moment expectations drift, so move the guaranteeable things into the agreement itself. Specify the deliverables as counts and cadences, define what a baseline and a monthly report contain, and state the reporting commitment as an obligation. Then, in the same document, name the uncontrollable factors explicitly: that engine outputs are non-deterministic and vendor-independent, that no citation, position, or date is promised, and that success is measured by movement on the agreed leading indicators and pilot metric. Putting the disclaimer next to the deliverables is not defensive lawyering; it is the artifact that keeps month four honest, because both sides signed the same understanding of what was and was not promised.

Set the timeline in the contract too, because an unstated clock is where a fair engagement gets misread as a failed one. Different engines discover, index, and first cite new content on different schedules, and a client expecting a citation in week three who sees nothing by week seven concludes the method is broken when the result was simply on schedule. The realistic engine-by-engine timeline is in how long does AEO take to work, and writing that arc into the expectations section converts a future disappointment into a met milestone. The kickoff is where you cement all of this, and the agenda for doing it is in how to run an AEO client kickoff workshop that sets expectations.

Sell the pilot as the guarantee substitute

For the client who genuinely cannot proceed without proof, the pilot is the answer, and it is a better deal for both of you than a fake guarantee would have been. Structure a fixed-scope, fixed-price engagement of thirty days that establishes the baseline, optimizes one defined cluster of pages, and measures the citation movement on those pages against a comparable cluster left unchanged. That holdout design controls for the platform-wide noise that makes naive before-and-after charts untrustworthy, so the result the client sees is causation they can trust rather than correlation they will later doubt. The full method for a pilot that converts skeptics into retainers is in how to run a 30-day AEO pilot that wins the client retainer.

Show the client the machine behind the pilot so the thirty days reads as a process with a mechanism, not a hopeful experiment. The loop is legible: measure where engines name competitors and not them, produce content built to be the answer to the exact prompts their buyers ask, and track citation share as it moves, which is what how OnlyAEO works describes, with the always-on production handled by the AI Feed Engine. If they want to see the arc on a named brand before they commit, the FastTrackr AI case study shows a company moving from unnamed to cited by the same method a pilot would run. And a free llms.txt generator gives them a real, immediate action that widens the surfaces engines can read, proof the work starts on day one rather than after a long ramp.

Know when to let the client walk

The last skill is the hardest: recognizing the client you should not sign. A buyer who hears the honest explanation, sees the four real guarantees and the contract that backs them, and still insists on a promised citation by a fixed date is not a client you can satisfy. They are a churn event with a deposit. Signing them means either lying to close, which detonates in a quarter, or spending the engagement managing a disappointment you designed into the deal. Walking is cheaper than either.

Say it cleanly: "I won't promise a specific citation because no one honestly can, and I'd rather lose your signature than earn it on a promise I'd have to break. Here is what I will guarantee, and here is a pilot that proves it on your data. If that is not enough, we are not the right fit, and I would be wary of anyone who tells you otherwise." A client who respects that answer is one you can keep for years. A client who rejects it was going to fire you the moment the guaranteed number failed to appear, and you are better off finding that out before the contract than after.