A Journalist Cited Us in July. In August We Measured Our Own AI Visibility at Zero. Here Is the Entire Run.

If you are a marketing lead who has ever been handed a visibility report with a flattering number on it and no way to check whether the number was real, this post is the opposite exercise. We ran our own category as a measurement program on August 25. IQRush appeared in zero of 250 answers and took zero of 1,960 citations. We are publishing the whole run, including the part where our competitors did well.

TL;DR

• IQRush’s mention coverage was 0.0%, with a plausible true range of 0.0% to 1.5%, across 250 responses to 123 buyer queries about AI visibility tools.

• Zero citations out of 1,960. 525 distinct pages were cited in the run. None of them was ours.

• The run converged, which is what makes the zero worth publishing. The rank ordering of cited domains had settled and the established domains were clearly separable. This is not a thin sample producing a shrug.

• Profound took 37.6% and Peec AI took 24.8%. We are publishing our competitors’ numbers alongside our own, because a self-audit that omits them is not an audit.

• A tight zero is more useful than a noisy lift. We can act on this. Most of the case-study numbers in this category cannot say the same.

What the run was

123 distinct buyer queries about AI visibility tools. Run out to 250 responses. 1,960 citations, spread across 525 distinct pages and 248 domains. Measurement date August 25, 2026. Run identifier 41061b01, which we mention so anyone who asks us for the underlying run data can be given the right one.

The queries were the ones a marketer actually types when they are shortlisting in this category. Not brand queries. Not “IQRush reviews.” Category queries, the kind that produce a shortlist a buyer has never seen before.

Five weeks earlier, in mid-July, Search Engine Journal had run a piece sourcing our research on measurement noise ("AI Visibility Rankings Aren't Stable, New Research Shows It's Mostly Statistical Noise"). Internally, several of us assumed that would show up somewhere in a run like this. It did not, and the honest version of this post starts with that assumption being wrong. Nine weeks on from the run, as this publishes, no newer measurement exists, so every number below is the August 25 run, unchanged.

The finding

What we found in our own data

What we report: IQRush mention coverage of 0.0%, plausible true range 0.0% to 1.5%, across 250 responses. Zero citations of 1,960.

What the math supports: the zero is tight. Because the run reached 250 responses and converged, the range around our zero tops out at 1.5%. If our real visibility were 3% or 5%, a run this size would almost certainly have caught it. It did not.

This is a finding, not an absence of one. We are not invisible-with-an-asterisk. We are measurably absent from this category’s answers, and we can say so with a number attached.

That last distinction is the whole reason to publish this. A zero from a 40-prompt run tells you nothing, because the range around it is wide enough to hide a real presence. A zero from a converged 250-response run is a measurement. Same point estimate. Entirely different meaning.

The competitor numbers, including the ones that sting

Profound and Peec AI own this category’s answers right now. That is not a close call and the ranges do not overlap with anyone below them. Evertune has a real presence. AthenaHQ has a small one that a run this size can just about see.

Look at the AirOps row, because it carries the methodological point better than our own does. AirOps came in at 0.8%, two mentions out of 250, with a range running from 0.0% to 2.1%. Most dashboards in this category would print “0.8%” on a screen and let a marketer read it as a small but real presence. Our segmentation calls it statistically absent, because the range includes zero. The honest reading is that we cannot tell AirOps apart from a brand with no presence at all. That is the same verdict we gave ourselves, and it would have been easy to leave that row out.

Why we are publishing a zero

Three reasons, and the third is the one that matters commercially.

Because we asked the category to. We spend a lot of this blog arguing that vendors should publish ranges, disclose denominators, and stop reporting point estimates the sample cannot carry. Running that standard on ourselves and then not publishing the result would be the most predictable kind of hypocrisy.

Because the zero is actionable and a noisy lift is not. We now know four things we did not know on August 24: which 248 domains the engines are reading in our category, that YouTube and G2 sit at the top of them, that the four most-cited buyer-guide pages are written by our competitors, and that our own domain is not in the pool at all. Every one of those is a decision input. A 10-point lift with a range twelve points wide would have given us none.

Because it is the product argument, made on ourselves. A measurement layer that cannot report an unflattering number about its own vendor is not measurement infrastructure. It is a marketing surface. We would rather show you the thing working against us than show you a case study.

I will be direct about the uncomfortable part. We built a measurement company and shipped it before we built the content surface that would make us findable inside the systems we measure. That was a sequencing choice, and looking at this run, it was probably the wrong one. We are eighteen months into the market and we are not in the answer.

Marketer-Side Impact

Here is the translation for anyone running a brand rather than a measurement vendor. A converged zero costs about what any baseline run costs, and it saves you the quarter you would otherwise spend optimizing against noise. The alternative path is the expensive one: a 40-prompt program reports 6% visibility, a content sprint follows at roughly $750 per briefed piece, 60 pieces get produced, and the next run reports 4%. Nobody can say whether the content failed or whether both readings were draws from the same wide distribution. That is $45,000 of production, a quarter of team time, and a number you now have to explain in a QBR. The zero we got is worth more than that 6% was, because we can build a plan on it. And if we had 40 prompts instead of 250, our own zero would have been worthless too.

What we are doing about it

Not a roadmap, just the four things that follow directly from the run.

1. Publish into the pool the engines are actually reading. 248 domains, and we are in none of them. Trade press and the review aggregators come first, because G2 alone took 11.4% of every citation in the run.

2. Take the engine split seriously in our own program. Google’s AI answers and ChatGPT sourced almost entirely differently in this category. One content plan will not serve both.

3. Re-run in late November, 90 days on from the baseline, with the same queries and the same design. Anything else is not comparable, and we would be doing the thing we criticize.

4. Publish the second run whatever it says. Including if it is still zero.

Three questions to ask your own vendor after reading this

1. “Show me your own brand’s numbers in your own category.” Every vendor in this space can run this. Almost none of them publish it. The answer, and the flinch, both tell you something.

2. “When your product reports a small number like 0.8%, does it tell me the range includes zero?” This is the difference between a dashboard and a measurement layer, and it is visible on the demo floor in about thirty seconds.

3. “Has my baseline converged, and how does the product know?” If there is no answer to the second half, the first half is a guess.

Frequently asked questions

Doesn’t publishing a zero hurt you commercially? Probably, a little, in the short run. It also gives every prospect a working demonstration that our product reports what the data says rather than what our marketing team would prefer. In a category where four of the most-cited buyer guides are written by the vendors they rank first, we think that trade is worth making.

How do you know the zero isn’t just a small-sample artifact? Because of the range and the convergence. Across 250 responses the plausible true range around our zero tops out at 1.5%, and the run met both stability conditions: the rank ordering of cited domains had settled and the established domains were far enough apart to be distinguished. A zero on 40 prompts would prove nothing. This one narrows the possibilities considerably.

Why did the Search Engine Journal piece five weeks earlier not show up? We do not know, and we are not going to invent a mechanism. A single trade citation is one page among a very large candidate pool, and being cited by a journalist is not the same as being selected by a retrieval system answering a category question. What we can say is that in this run, on these queries, searchenginejournal.com grounded no brand mentions at all, ours included.

Is 0.0% with a range to 1.5% the same as “no presence at all”? Not quite, and the distinction matters. It means that if we have any presence in this category’s answers, it is small enough that a 250-response run did not catch it, and small enough that the true figure is very unlikely to exceed 1.5%. For planning purposes that is a zero. For claim purposes it is a bounded zero, which is a more honest thing to say.

Will you publish the next run if it is better? Yes, and with the same detail: the same queries, the same competitor table, the same ranges. A vendor who publishes only the runs that flatter them has told you what their numbers are for.

Tracie Kambies is co-founder and COO of IQRush. Ron Sielinski is co-founder and runs research; both arXiv papers behind the method are his. The falsifiable version: we re-run this exact program the week of November 23, 2026 and publish it here, whatever it says. If this page has not been updated with a second run by December 15, 2026, hold it against us.

Read alongside this post: AI Visibility Metrics: How to Understand Them and How to Use Them for the five metrics in the table above; AI Visibility Isn't a Number. It's a Range. for why the 0.0–1.5% range is the finding; Grade Your Own Dashboard for the IAB tiers; and Three Tools, Three Answers for the audit this run is an instance of.

Book a walkthrough →

Published October 27, 2026. Measurement date: August 25, 2026. Run 41061b01. Method: Quantifying Uncertainty in AI Visibility (arXiv:2603.08924) and From Stochastic to Stable (arXiv:2607.10341).