How to Turn an AI Visibility Audit Into a Retainer

An audit is a wedge. It is not a business.

The wedge works because it is cheap, fast and undeniable: here is what engines say about your category, here is who was cited, here is where you stand. That earns a second meeting. What earns the second year is a queue of work, a cadence, and a report that keeps classifying the gap.

Most AI visibility audits die in the gap between those two things. The findings land, the client agrees the problem is real, and then there is no agreed next step because the audit ended at insight. Insight is not a retainer.

The five-step motion

Step one: audit. Run the prompt set that maps to their revenue, across the engines that matter to their buyers. Capture the answers, the cited sources, and what was said about them. Keep the raw output. It is the evidence for everything that follows.

Step two: findings. Present it in their language. Not "your AI visibility score is 34". Instead: "when a buyer asks for the best supplier in your category, these three firms are named, and this is the page each name rests on". Then the classification: absent, present but not cited, or cited but wrong.

Step three: gap. Translate the classification into commercial consequence. Absent from comparison prompts means excluded from shortlists. Cited but wrong means an objection you have to overcome before the conversation starts. Present but not cited means the page exists and the passage fails.

Step four: plan. The ranked queue. What changes, in what order, who owns each item, and what each phase costs. This is the document that becomes the retainer, so it has to be specific enough to be scheduled.

Step five: retainer. Agree the cadence, the queue depth and the reporting rhythm. Then renew on evidence rather than goodwill, because the same prompt set rerun is the whole argument.

What makes the findings land

Three things, and none of them are charts.

Their words. The prompt set is written the way their buyers write. If the client reads the questions and does not recognise their own customers, nothing after that page is trusted.

Their competitors. Not a generic market map. The domains actually cited in their category, named. Clients frequently discover their real competitor set is not the one they put in pitch decks.

Their claim, quoted back. The sentence an engine currently uses to describe them, shown verbatim, ideally with the source that produced it. This is the single most persuasive artefact in the whole engagement.

Turning the plan into a retainer

The plan becomes a retainer when three things are fixed.

Scope by work type, not by vague effort. Architecture, passages, corroboration. Each has a definition of done. A retainer that promises "ongoing AI optimisation" renews on trust; one that promises a named number of corrected passages and a monthly citation report renews on evidence.

A cadence the client can plan around. Monthly is the usual shape. The audit reruns, the classification is refreshed, the queue is re-ranked, and the deliverables from last month are checked against the prompts they were meant to move.

A price anchored to the alternative. The comparison is not another tool subscription. It is the cost of a strategist's month, or a writer's month, or the deals lost to a shortlist they were not on.

"Deploy a cluster per client per month, charge £300-500 each, keep £250+ margin per cluster. That's fractional CMO economics."

John Russell, Director, Digital Womble

Why one-off audits fail to renew

Four reasons, all structural.

No queue. The audit produced insight and no scheduled work, so there is nothing to renew.

No cadence. A single report is a snapshot. Clients renew rhythms, not documents.

No baseline discipline. If the prompt set or method changes between reports, movement is unprovable and the retainer becomes a matter of faith.

No owner on the client side. AI visibility work touches content, site, PR and sometimes product. Without a named counterpart, the plan stalls quietly and the retainer gets blamed.

The margin question

The reason this works as a service line rather than bespoke consulting is productisation. The prompt set, the citation capture, the classification and the content production are the same process every time, so the marginal cost of the second client is far below the first.

That is also where most agencies get the maths wrong: they price the retainer as if every month were a bespoke project, then cannot afford to deliver the cadence the client was promised.

"You're probably spending 20 hours planning content that takes 1 hour to execute. Topical Authority Content Strategy reverses that ratio. 1 hour to plan and generate, 20 hours to refine and execute."

John Russell, Director, Digital Womble

The renewal argument

At renewal, the conversation should not be about whether the client believes in GEO. It should be a before-and-after on the same prompt set: which prompts moved, which did not, which sources changed, what was published, and what is queued next.

That is a report any client can defend internally, which is the actual job. Budget holders do not renew vendors. They renew things they can explain to their own board.

Delivering it without building it

Running audits is straightforward. Running the work behind them, at a cadence, across several clients, is where agencies lose margin.

Our platform produces the audit evidence and the content that follows it, and agencies deliver both under their own brand. Start with the white label GEO service, or read what an AI visibility report should actually show to see the evidence standard we hold ourselves to. Pricing covers the commercial shape, and the 15-minute GEO pitch covers how to open the relationship.