AI Search Consultancy vs. AI Visibility Tool

An AI visibility platform automates something you could, in principle, do by hand: it sends a list of prompts to ChatGPT, Google AI Overviews, Perplexity, Gemini, and Copilot on a schedule, records the answers, and parses them for brand mentions, cited URLs, and competitor appearances. It then turns that into a time series.

That machinery is genuinely valuable, and it is not trivial to build:

• Scheduled sampling. AI answers are non-deterministic. Running the same prompt repeatedly and aggregating is the only honest way to measure, and doing that across several engines every month is real infrastructure.
• Citation extraction. Pulling the source URLs out of an answer and attributing them to domains gives you the source ecosystem an engine leans on for your category.
• Competitor tracking. Seeing which rival brands appear in which prompts, and how often, is a straightforward and useful output.
• Alerting and reporting. Change detection and an exportable dashboard make the channel legible to people who are not in it every day.

If you want the sourced specifics — which engines each product covers, whether it separates citation from recommendation, and what the vendor actually publishes as a price — that lives on our AI visibility tools comparison, where every claim carries a link to the vendor's own page.

The gap is not a flaw in the software. It is a category boundary. Measurement products report state; they do not decide what should be true instead.

• It does not choose your prompts. Every platform makes you supply, or accept, a prompt list. A prompt universe assembled from keyword exports measures search behaviour that predates AI assistants. A prompt universe built from win-loss interviews, sales-call transcripts, and the competitive set measures the questions that precede a purchase. Same dashboard, different truth.
• It does not explain cause. A dashboard can tell you a competitor is recommended in 60% of decision-stage prompts and you are recommended in 12%. It cannot tell you whether that is a third-party review-site gap, an entity disambiguation problem, a page that is structurally unquotable, or a rendering issue that keeps crawlers from seeing your content at all. Those four causes have four different remedies.
• It does not distinguish being cited from being recommended. These are separate outcomes with separate remedies — one is about your own content being usable, the other about what third-party sources say. Where a product does track both, we say so on the tools page with a source. Our own treatment of the distinction is on the methodology page.
• It does not do the work. No platform writes the comparison page, ships the schema, briefs the engineer, or earns the mention on the industry publication that the engine keeps citing.

Several situations make buying software and running the work in-house the correct call. We would rather say so than pretend otherwise.

• You already have a strong SEO or content lead with capacity. If someone on staff understands retrieval, entities, and structured data, and has ten to fifteen hours a week free, they will do this well with a dashboard and a roadmap they write themselves. Hiring an outside consultancy on top of that buys speed and pattern recognition, not competence you lack.
• You are still deciding whether the channel matters. If nobody internally can yet say what AI-sourced pipeline would be worth, a licence is the cheapest way to find out whether buyers in your category are asking assistants about you at all. Measure for a quarter, then decide.
• Your category is small or highly technical. Where the buying universe is a few dozen prompts and a handful of named competitors, the measurement problem is small enough to run yourself.
• You mainly need reporting for someone else. If the actual requirement is a board slide showing the trend, a platform is the direct answer to the direct question.
• Your fundamentals are not in place yet. If your site does not render for crawlers, has no comparison or pricing content, and has never earned third-party coverage, spend the first budget there. AI visibility work compounds on top of those things; it does not substitute for them.

The inverse is the honest case for a consultancy: you lack the hours, the channel is already producing pipeline you cannot explain, the diagnosis keeps coming back ambiguous, or nobody internally owns the execution once the dashboard turns red.

Licence price is the smallest term in this equation. The variable that dominates is who spends the hours. The table below compares the three realistic configurations on the four costs that actually appear in a budget.

One deliberate omission: we do not publish a headcount salary estimate here. Loaded cost varies enough by market and seniority that any single number would be a guess dressed as data.

In practice the mature configuration is both, and it is worth being plain about that rather than arguing software is unnecessary. Teams that run this channel seriously tend to end up with a monitoring platform for continuous coverage and alerting, and a consultancy — internal or external — that owns prompt design, diagnosis, and the work that changes the answer.

The division of labour that works: the platform answers what changed, the consultancy answers why and what now. Where those two functions sit inside one team, you do not need us. Where they do not, buying only the first one leaves the second undone, and the dashboard becomes a monthly reminder of a problem nobody is assigned to fix.

Our engagements include the measurement rather than requiring you to buy it separately, and the prompt universe stays locked so month-on- month comparisons hold. If you already own a platform you like, we will work from your data instead — that is a reasonable request and we do not charge extra to honour it. See what engagements cost.