Loading
How often AI names or cites you at all
Average points when it does, from -0.5 to 1.0
Score falls, Presence holds
The assistants still name you, and recommend you less firmly.
Presence falls, Preference holds
You are being left out of answers.
A minor factual error
A major error, being confused with another company, or appearing only under an old name
A warning always stays negative
Each square is one answer that cited the company's website. The red ones also named it. Salienz, AI Security Shortlist Study, October 2026, 7,385 answers from seven AI assistants.
One security vendor's website, mostly its guides. AI Security Shortlist Study.
What a buyer actually saw. A citation without a name earns 0.05.
In 300 simulated markets, where a mention rate showed a gain every time.
How often ChatGPT and Google's AI repeated a brand list, in SparkToro and Gumshoe's 2,961 runs.
Markets showing a fake gain when the grader nudged the client up one step, one time in ten. Salienz's own tests.
Salienz's own tests, 5 October 2026, at 8,000 answers a month, for the brand under test.
Inside each assistant
Movement is measured against the category inside one assistant, so a model update that hits every brand is not your win or loss.
False alarms when one assistant turned cooler towards every brand.
A tenth held out
About 10 percent of questions are never shared with the content team. Gains that skip them are not real.
Median lift, in points, from content written only for the tracked set.
Blind grading
The grader never knows which brand is the client.
Every brand named
Every brand the assistants name is tracked, not a hand-picked list.
| Test | Salienz Score | Mention rate | Sentiment composite |
|---|---|---|---|
| Ranks brands by how often buyers chose them (1 is perfect) | 0.98 | 0.93 | 0.97 |
| The same, when answer order is random | 0.98 | 0.93 | 0.97 |
| Finds the true category leader | 92% | 72% | 82% |
| Mentions up, recommendations down: flags the drop | 90% | 0%, shows a gain | 80% |
| Catches old-name confusion | 100% | 2% | 4% |
| Noise on one month's reading, 95% range | ±0.8 points | ±1.0 points | ±0.9 points |
A good Salienz Score looks low next to a mention rate, because the scale counts recommendations and not mentions. In Salienz's simulated test markets, the middle half of category leaders scored between 35 and 45, and mid-pack brands between 11 and 16. A score of 100 would mean being the correct top pick in every relevant answer.
No. A mention rate counts how often a brand is named, which the Salienz Score reports separately as Presence. The score multiplies Presence by Preference, how strongly the brand is recommended. In testing, a plain mention rate showed a gain in cases where recommendations were falling, and the Salienz Score flagged the drop 90 percent of the time.
Yes. An answer that warns buyers away from a brand scores minus 0.5, and a warning always stays negative. If warnings outweigh recommendations across the question set, the Salienz Score goes below zero. The scale runs from 1.0 for a top pick down to minus 0.5 for a warning.
Every Salienz Score is reported with a 95 percent range, and the same question set is run twice to measure normal noise. In testing at 8,000 answers a month, a month-to-month move had to pass about 1.4 points to count as real. Roughly 10 percent of questions are held out to catch gains that exist only on the tracked set.
Not by itself. The Salienz Score is an early signal of how firmly AI assistants recommend a brand to buyers, and pipeline is still the outcome. Use it to steer the work month to month, and judge the programme on pipeline. It is not revenue, market share, a rank tracker or a guarantee.
Run the method on your category first.
The Foundation audit: 100 questions across five engines, $800, delivered in seven business days.