Short answer, as of August 2026: do not walk in with a methodology — walk in with a measurement. Run their real buyer questions through the assistants before the meeting, bring the verbatim answers, and let the gap do the selling. Then scope a small, fixed first phase — an audit and a baseline, typically two to four weeks — rather than an open-ended retainer for a discipline neither of you has calibrated yet.
The most common way this conversation goes wrong is not ignorance. It is over-promising in the first fifteen minutes, and then owning that promise for six months.
Before the meeting: forty minutes of homework
Everything that makes this meeting go well happens beforehand, and it is quick.
Write down eight to twelve questions a genuine prospect of theirs would ask an assistant before choosing a supplier. Not keywords — questions, in the words a buyer would actually use. Then run each one through two or three assistants and record, verbatim, the answer and the sources cited.
You will end up with one of three findings, and each leads to a different meeting:
- They are absent and competitors are named. The strongest position. The gap is demonstrable and specific, and you can show it rather than assert it.
- Nobody in the category is named. An open field. Lower urgency, higher upside, and an easier case for moving first.
- They are already named reasonably often. The honest answer is that this is a monitoring and defence job, not a rescue. Saying so buys you enormous credibility, and it is usually true more often than consultants expect.
Bring screenshots. A client reading an assistant recommending their competitor, by name, in their category, does not need to be persuaded of anything else.
The five questions to ask them
1. What do your customers ask you before they buy? This is the prompt set, and they already know it — it is in their sales calls and support inbox. Their answer is more valuable than any keyword research you could do in the first week. The structured version is in choosing which prompts to track.
2. Has anything changed in your traffic recently? Look for flat impressions with falling clicks, which usually means the query moved into an AI answer rather than that they lost rankings.
3. Who do you actually compete with? Named competitors make share-of-voice measurable. Vague answers ("other agencies") produce vague reporting.
4. Who controls the website, and how fast can changes ship? This determines whether the engagement is achievable at all. A client whose site changes require a quarterly release from an outsourced developer cannot buy a twelve-week programme in good faith.
5. What would make you consider this successful? Ask it early, get it in writing, and push back if the answer is a revenue figure with no baseline attached. Nailing this down is what stops month four becoming an argument.
What you can promise
Three things, comfortably.
A measured baseline. You can tell them, this month, in what fraction of their buyers' real questions they are currently named, per engine, and who is named instead. This is genuinely valuable, deliverable in days, and almost nobody has it.
Access and eligibility. You can verify — and fix — whether assistant crawlers can reach them, whether pages are indexed, and whether content renders without JavaScript. These are the gating failures, they are common, and they are fixable inside a fortnight. The audit is described in the AI crawler access audit.
A defensible trend. With scheduled sampling, you can show whether citation rate is moving by about week twelve. That is the honest first checkpoint, and the phase-by-phase expectation is set out in how long GEO takes.
What to refuse, out loud
Ranking guarantees. There is no rank to guarantee. There is a citation rate, it is probabilistic, and the mechanism is not something either of you controls. Say this plainly and early; the client who wanted the guarantee will find someone who offers one and will be back in five months.
A date for a specific citation. "You will be named in ChatGPT for this question by October" is not a commitment anyone can keep.
Refusing both, in the first meeting, is the single highest-leverage credibility move available — and it is also what most distinguishes a competent supplier from a confident one. The client-side version of this filter is how to choose a GEO specialist, and it is worth reading from the buyer's side before you sit down.
Scoping the first engagement
| Phase | Duration | Deliverable | What it proves |
|---|---|---|---|
| 0. Baseline | Week 1 | Citation rate per engine across 30–60 real questions, plus who is cited instead | Whether there is a gap worth funding |
| 1. Access audit | Weeks 1–2 | Crawler access, indexing and rendering verified and fixed | Whether they were even eligible |
| 2. Entity consistency | Weeks 2–4 | Name, address, phone and category agreement across sources | Whether engines can identify them confidently |
| 3. Retrofit | Weeks 3–8 | Highest-authority existing pages upgraded to answer directly | Fastest available route to a first citation |
| 4. Net-new content | Week 6 onward | Pages for questions nothing on the site answers | Sustained expansion |
| 5. Report and decide | Week 12 | Trend, per engine, against the week-1 baseline | Continue, adjust, or stop |
Sell phases 0 and 1 as a fixed-fee first engagement. They are fast, cheap, produce a concrete artefact, and frequently uncover a blocking problem that reframes everything else — which is a far better second meeting than a status update. The execution detail for the rest is in the first 30 days agency playbook.
Note the ordering: retrofit before net-new. Upgrading pages that already rank is materially faster than building authority from nothing, for the reasons set out in retrofitting existing pages for citation.
What to charge, and what it costs you
Price the first phase as a fixed audit fee and the ongoing work as a retainer, the way you would any other discipline. The input costs are modest and worth knowing before you quote: tracking software in this category starts around $99 a month, agency-grade multi-client platforms sit in the low hundreds, and the dominant cost is your team's hours. If you are building the internal case, the GEO ROI business case is written to be handed to a finance team, and GEO versus traditional SEO budget split covers what to move rather than add.
Do not bill this as a bolt-on to an existing SEO retainer without changing the reporting. If the deliverable is new, the report must be too, or you will be judged against the old numbers.
DigiRank's AI Visibility Tracker is built for this shape of work: multi-tenant from the Agency plan, so each client gets their own campaigns and prompt sets, with white-label share pages and a custom domain so the baseline you present carries your brand rather than ours. Agency is $249/mo for up to 15 client tenants, 10 campaigns and 2,500 prompt-checks, on a 14-day trial — enough to run the week-1 baseline for a prospect before you have signed them. The reporting side is covered in white-label SEO reporting for agencies.
Frequently asked questions
My client asked me how to rank in AI search and I don't know where to start. What do I do? Run their real buyer questions through two or three assistants before the meeting and record the verbatim answers and sources. Bring that to the conversation as a measured gap rather than a methodology. Then scope a fixed two-to-four-week baseline and access audit instead of committing to an open-ended programme.
What can I honestly promise a client about AI search? A measured baseline of where they are and are not cited, verified crawler access and indexing, and a defensible trend by about week twelve. Do not promise a ranking, a guaranteed citation, or a date by which a specific engine will name them.
How is this different from the SEO work I already do? Much of the underlying work overlaps — clear structure, specific claims, consistent entity data, technical access. What differs is the measurement and the selection logic: there is no ordered list, so the metric is a citation rate across a prompt set rather than an average position.
How long before a client sees results? Access and indexing fixes can show effects within days of re-crawling. First citations on retrofitted pages typically appear around week six, and a trend worth reporting at about week twelve. Anything faster is usually the removal of a blocker rather than newly earned visibility.
Should I sell this as a separate retainer or fold it into SEO? Either works commercially, but the reporting must change if the deliverable changes. Folding new work into an old retainer without new reporting means being judged on metrics that do not reflect what you did.
What if the client is already being cited? Say so. That converts the engagement from rescue to monitoring and defence, which is a smaller but far more defensible scope — and the honesty typically wins you the work over someone claiming a crisis.
How many prompts should the first baseline cover? Thirty to sixty real buyer questions is a workable first set for most businesses, sampled repeatedly rather than once. Fewer than about thirty is too noisy to trend; far more is hard to keep meaningful early on.
What does the tooling cost for an agency running several clients? Multi-client platforms in this category generally sit in the low hundreds per month. DigiRank's Agency plan is $249 a month for up to 15 client tenants with white-label reporting, on a 14-day trial. The larger cost in any case is staff hours, and quotes that omit them tend to be the ones that go wrong.