What we refuse to do

No proxy, no DNS change, no script that rewrites your pages, no agent with publish access, and no automated fix applied while you are not looking. Your site serves its own bytes and we never sit in that path. If you commission build work from us, that is a human engagement you approve and deploy yourself, and it is recorded in your change log under our name.

Several tools in this category apply fixes for you by routing your traffic through their infrastructure. That makes them a dependency of your homepage: their outage is your outage, and their bug is your bug. It also means the party editing your pages is the party reporting whether the edit worked. The distinction that matters is not whether a vendor ever touches a site. It is whether anything reaches your site without a person deciding, and whether the touching is visible afterwards.

What it costs us: We lose every buyer who wants it applied automatically while they are not looking. That is a real segment and we are not the product for it.

If we ever help you make a change, the measurement that judges it is the same one we would run on a stranger. The significance gate does not know who made the edit and cannot be told.

The dominant pattern in this category is a vendor whose agent writes your content and whose dashboard then reports the improvement. That is marking your own homework. Doing the work is not the problem; grading it with a private rubric is. This holds because of machinery rather than goodwill: the significance test is applied to the numbers and has no field for who caused them, and every change we make is written to your change log with us named as the actor, where you can read it.

What it costs us: We cannot show you a flattering number when a change did not work. Our own dogfood runs have produced results we would rather not have published.

No paying people or bots to ask assistants about you, to engage with answers that name you, or to plant brand mentions on forums so a model picks them up.

An engine's answer is worth measuring because it reflects something real. The moment a vendor manufactures the inputs, the number they sell you measures their own activity. You are then paying twice: once for the farming, and once for a metric that only moves because of it.

What it costs us: It is the fastest way to move an AI-visibility number, and we will not sell it. Vendors who do will show a steeper chart than ours.

We do not sell any service whose success our own instrument reports on. That includes content, links, PR placement and engagement.

Several vendors that built their business on manipulating a signal are now adding AI-visibility measurement surfaces. When one company both moves the needle and owns the needle, there is no version of the resulting report you can independently check.

What it costs us: Bundling would be a straightforward second revenue line, and refusing it means we stay a smaller company with one thing to sell.

Every number on this site is either something we measured, with its sample, or something we cite to a named third party with a link.

This category runs on the uncited two-decimal statistic. A figure like "63.41% of buyers now start with AI" is precise enough to sound measured and specific enough that nobody checks it. The decimal places are doing persuasion, not arithmetic. Precision without provenance is a rhetorical device.

What it costs us: Our marketing copy has fewer big numbers in it than anyone else's in this category, because most of the available ones cannot be traced.

Numbers about our own results carry a baseline, a window and a sample, exactly like the numbers we borrow from other people.

There is a tell worth learning: many vendors carefully cite every industry statistic they borrow, then state every number about their own results with no source at all. The citation discipline is real, and it is applied in precisely one direction. Watch what a company does with the numbers nobody can check.

What it costs us: It rules out most of the case-study formats that convert best, because they need a percentage with no denominator attached.

We publish an outcome only with the starting value, the time window, the sample size and the confidence range. If we cannot state all four, we do not publish it.

"200% increase in leads" from a base of two is four leads. Without the denominator you cannot tell the difference between a transformation and a rounding error, and the format exists because you cannot.

What it costs us: It is why our proof page is short.

Answers come from the engines' own APIs and documented surfaces. We do not scrape assistant interfaces, and we do not infer a number from a proxy signal and present it as measurement.

A scraped number breaks silently. When a layout changes, the scrape starts returning something subtly different and the chart keeps drawing a confident line. You cannot see the difference between a real move and a broken selector.

What it costs us: Some surfaces have no official interface, so we cannot measure them at all. We list those as not measured rather than estimating them.

No bot clicks, no engagement pods, no traffic products.

Vendors selling click manipulation cite leaked search-engine documentation as proof that clicks are a ranking signal. Read the same documents closely and the striking part is the sheer volume of infrastructure devoted to FILTERING clicks: classifying them, discounting them, and identifying the automated ones. Evidence that a system spends heavily on detecting fake clicks is not evidence that fake clicks work.

What it costs us: None. This one costs us nothing and we would refuse it anyway.

When we cite a rating, we name the platform, the review count, and who the reviewers actually are.

A high average can measure something other than the product. A rewards platform can hold an excellent rating from the people who SEND rewards and a poor one from the people who try to spend them. A tool can rate well with the buyer who pays and badly with the team forced to use it. The number is real; the question is which audience produced it.

What it costs us: It means we point out when our own review footprint is thin, and it is: we are young.