Ask any vendor these. Us included.
None of this is specific to us. If you are evaluating anyone in this category, these are the questions that separate measurement from theatre, and you should get straight answers to all of them before money changes hands.
- Show me one run, in full. The exact prompt and market, the product and model, the date and time, whether the session was signed in or anonymous, whether the engine actually searched the web, the raw answer, the sources it cited, and the screenshot.
- Define your words. What exactly counts as named, recommended, cited, and absent? A vendor who cannot draw those lines is not measuring; they are describing.
- How many times did you ask? One run of one question is a coin flip. Ask what the repeat count is and what happens when the answers disagree with each other.
- Show me a number that made you look bad. Any honest panel produces some. A deck with no zeros in it has been curated.
- What do you control, what do you influence, and what can you not control at all? If those three are blurred together, the guarantee is doing work the method cannot support.
- What happens to my pages, data and access if I leave?
What we control, what we influence, what we cannot control
This distinction came from that reviewer and it is a better articulation than we had, so we have adopted it.
| Category | Examples | What we'll commit to |
|---|---|---|
| We control | On-site facts and service pages, published prices and process detail, structured data, title and heading structure, profile completeness, name and address consistency across directories | Shipped on a schedule, approved by you before publication, logged and reversible |
| We influence | Review acquisition process, inclusion in third-party directories and roundups, local press and citation of your published facts by others | We run the process and report what happened. We do not control whether a third party publishes. |
| We cannot control | Whether a specific model names you at a specific moment, ranking position, how an engine weights reputation, proximity, or its own product policy | Nothing. We measure it and report it, and we will not sell you a promise about it. |
The presence guarantee sits deliberately at the boundary of the first two columns: presence across a question set over 90 days, which the work can move, rather than a placement on a given day, which it cannot.
The guarantee, measured and disputed
Named by at least one AI engine on at least 5 of your 25 client-approved questions by day 90, measured from a clean anonymous vantage, with the run log delivered — or month four is free.
Three things make that auditable rather than rhetorical. The question set is approved by you in writing before any work starts, so it cannot be swapped for easier questions later. The day-90 measurement runs from a clean anonymous vantage — no account signed in — so nothing is flattered by an account that already knows you. And the count comes from a run log you receive, so you can audit it line by line rather than take a dashboard's word for it.
If we disagree about whether it was met: name the runs you dispute, and we re-run those exact questions with you watching, from a clean vantage. If the re-run does not clear the bar, month four is free. The log is the arbiter, and you hold a copy of it.
Samples you can see now
The monthly scorecard is the actual retainer deliverable, populated from a real measurement panel with the firms anonymized. It shows the five-rung ladder per engine, the per-question detail, and the zeros. The durability report shows the underlying method applied at scale. Our own scorecard is the same instrument pointed at us, published with our own failing numbers, and corrected in public when we got one wrong.
Full client reports are confidential. If you want to see one, ask and we will walk you through a redacted one on a call rather than post a client's diagnosis on the open web.
What we cannot show you yet
Two items on that reviewer's list we cannot satisfy today, and it is more useful to say so than to fudge them.
A published before-and-after from a fix we shipped. We can show that positions churn, that published facts get quoted, and that businesses without them are absent — measured across many markets. What we cannot yet show publicly is a controlled sequence where we made a change and the answers moved because of it. That evidence takes a client engagement running long enough to produce it, and we would rather say that plainly than dress up a correlation. In the meantime we are running the experiment on our own property, where we control every variable, and the result will be published whichever way it goes — including if the answer is that our own fix did nothing.
Named client references. This is a young practice with a small number of clients, several in regulated categories where the engagement itself is confidential. If a reference matters to your decision — and it reasonably might — say so and we will ask a client directly rather than list anyone without their consent.
If either of those is a dealbreaker, that is a rational position and the free report is still free.
Why this page exists
The argument for hiring us is that we measure honestly and hand you the log. A page that only listed our strengths would undercut that argument. So this one lists the gaps too, and it will be updated when they close.