A fixed-scope AI visibility audit. You keep the report either way.

Why a fixed-scope audit exists at all

Most agencies in this category sell a retainer or nothing. That works badly for two kinds of buyer: the one who needs to know how bad the problem is before they can get budget approved, and the one who has an internal team and only needs the diagnosis.

A price band has formed for this internationally, roughly fifteen hundred to five thousand US dollars for a serious audit. We price after a scoping call rather than publishing one number, because a single-market English audit and a five-market bilingual one are not the same job and pretending otherwise means one of you is overpaying.

If your budget is zero, run the free grader first. It will tell you whether this is worth paying for.

What we measure

The prompt set

Between forty and eighty questions, agreed with you before we start, written the way your buyers actually ask rather than the way your marketing describes you. It covers six question types: the direct question about your name, the open recommendation where you are not mentioned, the comparison against your real alternatives, the trust question your sales team dreads, the question about your founder or key people, and the same set again in the other language your market uses.

The unnamed recommendation question is the one most audits skip and the one that most often reveals the expensive problem.

Six engines, separately

ChatGPT with search on and off, Perplexity, Google AI Overviews, Google AI Mode and Gemini, Claude, and Microsoft Copilot. Reported per engine. We do not produce a single blended visibility score, because six engines disagree constantly and an average hides the one that is failing.

The retrieval-versus-memory diagnosis

For every wrong answer, we determine whether the claim came from a live page you can fix, from training memory you cannot, or from a gap the engine filled by inventing something. Doing the wrong work for the wrong cause wastes a quarter, and this is the finding that prevents it.

Source tracing

Every problem answer traced back to what produced it. Perplexity numbers its citations, which does most of the work. The others need comparison runs. What comes back is a specific list of pages, listings and threads rather than a general sense that the internet is unkind.

Access and technical check

Whether AI crawlers can reach you at all, whether anything is blocked by accident, whether your critical content renders without JavaScript, and whether your structured data says what your pages say. Several audits stop before this and it is where a surprising number of root causes are.

What you receive

The answer archive. Every prompt, every engine, the full answer text, the date, and whether search was on. This is the only evidence base that exists, because nothing is retained on the engine side. It is yours as raw data, not a summary.

The findings document. What each engine says, sorted into stale, merged, invented and absent, with the traced source for each.

The fix list. Prioritized by expected effect against effort, written so your own team could execute it. Anything that cannot be fixed is marked as such with the reason, rather than left implying it can.

A walkthrough call. One hour, with whoever needs to be in the room.

What we will tell you

If the audit finds that your AI answers are basically fine, we will say so and we will not follow it with a retainer proposal. Some brands are in better shape than they fear.

If it finds that the negative thing being repeated is accurate, we will say that too. No amount of source work outruns a real operational problem, and you should hear it from the audit rather than after six months of invoices.

And if the honest answer is that your problem is not AI visibility but conversion, or ranking, or something else entirely, that goes in the report.