The hardest part of selling answer engine optimisation is not the pitch. It is deciding what you are actually selling, because the obvious framing, "we will get you into ChatGPT answers", is a promise nobody can make with a straight face.
This is a pricing structure that does not need that promise. It sells three things: a measurement, a fixed body of work, and ongoing monitoring. Each one has a defensible price because each one has a deliverable you can hand over.
Why the usual retainer framing fails here
A standard SEO retainer works because rank tracking gives everyone a shared scoreboard. Positions move, reports get sent, the relationship survives a slow month because the graph is visible.
AI answers do not give you that. They are not deterministic, attribution back to a phone call is mostly broken, and a business can genuinely improve while a re-run of the same eight questions shows a lower number by luck.
If you sell a monthly retainer against an unstable metric, month three is a hard conversation. If you sell a discrete audit, a fixed sprint of work, and then monitoring, month three is a status update.
The three products
1. The visibility audit
What it is: eight question shapes, at least two assistants, run logged out in fresh conversations. Every business name in every answer recorded. A score for the client and for the top competitors. The list of URLs the assistants cited. A one page summary of where the gap comes from. The 15-minute manual method sets out how to run it.
What it is not: a strategy document. Resist padding it. The value is that the client reads one table and sees a zero next to their name and a 75 next to a competitor.
What to charge: this is a fixed-price deliverable, and the price should reflect that, by our own estimate, it takes a couple of hours to do properly and produces something the client can act on with or without you. Price it where a business owner can approve it on their own, not where it needs a budget line and two meetings.
Why it sells: it is the only part of this whole field where the finding is unarguable. You are not predicting anything. You are showing them a screenshot of an answer that does not contain their name.
Sell it standalone, sell it as a paid discovery that credits against the sprint, or run it free at volume as prospecting. The third is the one to start with, because it is also how you find the people who would buy the other two.
2. The visibility sprint
What it is: a fixed scope, fixed price, four to six week engagement that does the boring work the audit identified.
A defensible scope is some version of:
- Google Business Profile completed and the primary category corrected
- Listings on the specific directories that appeared in the citation set
- Name, address and phone standardised everywhere
- One plainly written page per question shape worth winning, usually two or three
- Homepage and service page copy rewritten so the differentiators are facts rather than adjectives
- A review request process that asks for specifics rather than stars
- A re-audit at the end, same questions, same conditions
What to charge: a fixed project fee. This is a defined body of work with a defined end, priced like a website project rather than a retainer. The work itself is not exotic, so the price should be honest about that: what the client is buying is that someone competent does all of it in six weeks rather than none of it in a year.
The critical clause: the deliverable is the work plus the re-audit, not a guaranteed score. Write that down. "We will complete the following work and re-measure with the same method" is a promise you can keep. "We will get you named in ChatGPT" is not.
When the re-audit is done and the client agrees to be named, the 60-day case study template is a structure for writing it up, failures included.
3. Monitoring
What it is: the same eight questions re-run monthly, a short report on what moved, and a flag when a competitor appears or disappears.
What to charge: a small monthly fee. Deliberately small. This is not a disguised retainer, it is a tripwire, and the price should be low enough that nobody has a reason to review it.
Why it matters commercially: monitoring is the only reason to still be in contact three months after the sprint ends. When a client drops out of the answers because a competitor got onto a directory, you are the one who notices, and that is the conversation the next sprint would start from.
Where the margin actually is
Not in the audit. The audit is a customer acquisition cost you are getting paid for.
Not really in the sprint either, once you account for the copywriting.
The margin is in the fact that the audit is the same eight questions every time, and the sprint is the same checklist every time. This is a productisable service, which means the second one costs you a fraction of the first, and the fiftieth costs almost nothing but the copywriting.
Two practical consequences.
Systematise the audit before you sell the tenth one. Doing it by hand teaches you what matters. Doing it by hand fifty times is a bad use of a month. Whether you build the automation or buy it, the point is that the question set, the conditions and the recording format stop varying between people on your team, because inconsistent inputs make the month-three comparison meaningless.
Prospect with the audit at volume. The cold email that works in this category is not a pitch, it is a finding: here are eight questions, here is the answer that did not name you, here are the three that were named instead. The cold email templates for AEO outreach are built on that structure. That email needs the audit to already exist before the conversation starts, which means running it for businesses who have not asked you to.
Running it for one prospect is fifteen minutes. Running it for every plumber in a city is the part that decides whether this is a service line or a hobby. That is the specific job AIgnored does: it scans a city and a niche, scores every business, and marks the ones with real reviews and a real website that the assistants never name, so the prospect list and the audit arrive together. It runs on your own API keys, so the cost per scan is whatever the provider charges you and nothing on top. The pricing page has the plans, including a free one for a single city.
What to put in the contract
Three clauses save arguments later.
Method, stated. The exact questions, the assistants used, the conditions (logged out, fresh conversation). If you re-measure differently, the comparison is worthless, and writing the method down is what stops that happening when someone else on your team runs it. If you use a tool for this, link its method: ours is on the methodology page, thresholds and limits included, and a client is entitled to read it.
Variance, acknowledged. State plainly that these systems are not deterministic, that a re-run can move a score without anything changing, and that the trend across several runs is the signal rather than any single number. Said at the start, it is a property of the method. Said in month three, it sounds like an excuse.
Attribution, acknowledged. State that AI-driven enquiries frequently arrive with no referrer and that you will measure visibility, not attributed revenue, unless the client adds a "how did you hear about us" step. Without it, the client is waiting for a revenue number that nothing in the work can produce.
What not to sell
Guaranteed placement. Nobody controls the shortlist.
"AI SEO" as a mystery. If the client cannot describe what you did in a sentence, you are charging for a word.
Content volume. Twenty blog posts do not get a plumber into an answer about emergency callouts. One clear page and four directory listings might.
Anything that depends on scraping or on gaming a platform's terms. Beyond the ethics, it is a business you cannot sell later, and the audit is a business you can.
The honest sales conversation
Say this, roughly:
"We asked four assistants who the best {trade} in {city} is, eight different ways. You were named in none of the answers. These three companies were named in most of them. The reason is not your reviews, it is that you are not on the four pages those answers are built from and your profile category is wrong. We can fix that in six weeks for a fixed fee, then re-run the exact same test and show you what moved. We cannot promise a number, because these systems change their answers run to run, and anyone promising you a number is guessing."
That paragraph gives a finding, a scope, a timeframe and a stated limit, which is everything the owner needs in order to say yes or no. It also survives month three, because it promised work and a re-measurement rather than a number.