Null is not zero

RankX AI distinguishes "we did not measure this" from "we measured this and found nothing", everywhere, and it renders the two differently. A blank means unknown. A zero means measured and genuinely absent. The second is a finding worth acting on; the first is a gap in the sample and acting on it would be acting on nothing.

This is the single most useful convention in the product, and it is also the one that surprises people, because most tools in this category do the opposite.

Why this matters more here than elsewhere

An AI answer either names your brand or it does not, and a lot of the time the check cannot produce a verdict at all: an assistant times out, an answer comes back in a shape the analyser cannot read, a Google connection has never synced, a crawl never reached a page. Every one of those is an absence of measurement.

If a tool reports each of them as a zero, three things happen. A customer with no data concludes they are invisible. A real absence, which is the actionable finding, becomes indistinguishable from a failed job. And every rate built on top is quietly wrong, because the denominator counted checks that produced nothing.

So RankX AI counts them separately, reports them separately, and shows you the denominator so you can see how much of the picture you have.

The rule, stated once

A rate divides by checks with a known verdict. Checks without one are excluded from the numerator and from the denominator, and counted on their own so the gap is visible. If no check has a known verdict, the rate is null and renders as unknown.

That single rule generates every behaviour below.

Where you will meet it

AI visibility

The two zeros are the pair worth internalising. 0% share of voice is the finding the product exists to surface. A blank share of voice is a schedule.

Google positions

Three states, never merged:

• Ranked, with a position from 1 to 20.
• Not in the top 20: the check ran and your page was not in the first twenty results.
• Never checked: no check has run for this keyword.

RankX AI also reports the reason a scheduled check was skipped, when one was, so a gap in a series has an explanation rather than a shrug.

AI Overviews

Unknown is a real third state. An AI Overview was detected but not captured, so RankX AI cannot say whether you were named inside it. That is not the same as "you were not mentioned", and turning it into one would manufacture an absence nobody measured.

AI Shopping

Shopping visibility divides by answers that showed products at all. An answer with no product carousel has no opinion about your catalogue.

The ads view carries the same discipline in a sharper form. One shopping surface reports which results were sponsored; the other does not expose a sponsored flag at all. A zero on the second surface therefore means "not observable here", never "no advertisers here", and the product says so rather than letting a reader draw the wrong conclusion from an empty column.

Google Search Console and Analytics

Both refuse rather than reporting zero. A connection is in one of four states and only the last returns figures:

A separate, related honesty: Search Console publishes on a two to three day delay at source. The last day or two being absent is a healthy connection, and RankX AI marks those days provisional rather than plotting a decline. Why traffic data is missing covers all four states.

Website Audit

A check whose audit path did not run is not assessed, not passed. This one was a real defect before it was a rule: seven checks were rendering confident green assurances for checks that had never run, on sites where nothing had been measured.

Three checks are honestly reported as assessable on no path at all, because the crawler returns no field for them. Keeping them visible with "not assessed on any path" records the gap; deleting them would lose it.

AI Readiness

Two consequences of the same rule:

• A score never renders without its coverage fraction. "72" alone is a claim about a site; "72, from 14 of 19 checks" is a measurement.
• Below the coverage floor there is no score at all. Not 0, which reads as "your site is broken". Not 100, which reads as "nothing to do". The honest output is the gap and the action that closes it.

A check that could not run moves the score in neither direction.

How to read a screen with blanks on it

1. Find the denominator. RankX AI shows it. If you are quoting a figure elsewhere, carry it: "named in 12 of 40 checks" survives being quoted and "30%" does not.
2. Ask which kind of blank it is. Not measured yet, or measured and absent? The screens say which, and the answers change what you do next.
3. Do not fill a blank in yourself. Averaging over a missing assistant, or treating an unknown AI Overview verdict as a negative, reintroduces exactly the error this convention removes.
4. Treat a coverage gap as a task. A large not-assessed count usually has a single cause: an unfinished crawl, a disconnected integration, a Website with no tracked keywords. Fixing the cause fills the column.

If you are building on this

Every RankX AI surface follows the same rule, including the MCP tools: a metric field is nullable, null means unknown, and every tool that can return one says so in its own response. When you report a number to a person, carry the denominator and the unknown count with it.