The pulse of AI visibility.
Outerank measures its own visibility publicly, every two weeks — plus anonymized benchmarks contributed by consenting projects.
Outerank’s visibility, in public.
54 fixed prompts × 3 engines · snapshot 11/15/2026
Visibility trend
Last 5 comparable benchmark runs
Last 5 benchmark runs
| Date | Mention SoV | Recommendation | Citation | Engine coverage |
|---|---|---|---|---|
| 2026-11-15 | 6.5% +1.1pt | 19.0% +2.0pt | 100.0% | |
| 2026-11-01 | 5.4% -0.4pt | 17.0% +2.0pt | 100.0% | |
| 2026-10-15 | 5.8% -0.5pt | 15.0% +2.0pt | 100.0% +33.3pt | |
| 2026-10-01 | 6.3% +2.4pt | 13.0% +2.0pt | 66.7% -33.3pt | |
| 2026-09-15 | 3.9% | 11.0% | 100.0% |
Methodology before momentum.
- Repeatable set: the same prompt library runs against ChatGPT, Perplexity and Gemini on the 1st and 15th.
- Distinct metrics: mention, recommendation and citation are reported separately; one is never presented as another.
- Consent: aggregate research includes only projects that opt into the Research Data Program.
- Cohort gate: no segment is published below 30 independent contributors.
- Revocation: withdrawing consent removes every observation traceable to that participant.
The questions a skeptic should ask.
An index you cannot interrogate is just a chart. Here are the awkward questions, answered.
What is the difference between mention, recommendation and citation?
Mention means the brand name appeared anywhere in the answer. Recommendation means the answer actively advised using it. Citation means the answer linked to or attributed a source URL on that domain. We report three separate numbers because they are three different outcomes — collapsing them into one “visibility score” is how most tools in this category overstate results.
Which engines are measured, and why only three?
ChatGPT, Perplexity and Gemini. Each live run costs real money, so we would rather measure three engines consistently than five occasionally. A benchmark with gaps in its history cannot show a trend — and the trend is the only part of this that matters.
Why bi-weekly instead of daily?
Two reasons. Cost — every run queries live engines, a real line item for a bootstrapped company. Signal — AI answers fluctuate run to run, so daily sampling mostly measures noise. The 1st and the 15th against an identical prompt library gives two comparable points a month: enough to see direction, without paying for noise.
Isn't it self-serving to publish your own score?
It would be if the score were flattering. It is not especially flattering, and we publish it anyway — including the runs where a metric went down. A company selling AI-visibility measurement that will not show its own is asking for trust it has not earned. Publishing the number means we have to live with it.
What counts as a good score?
There is no industry baseline yet — which is part of why this index exists. For a young domain in a competitive software category, a citation rate in the low double digits is reasonable and above roughly 30% is strong. The honest answer: the absolute number matters far less than whether yours is rising across comparable runs.
Can my brand be added to the index?
Aggregate cohorts include only projects that opt into the Research Data Program, and no segment publishes until enough independent contributors exist to keep participants unidentifiable. Withdrawing consent removes every observation traceable to you. For your own numbers rather than cohort numbers, run the benchmark on your own project inside Outerank.
Measure your place in the answer layer.
Start with the free public scan, then track the prompts, engines and sources that matter to your market.