- \n
- Run-to-run, the top-3 brand set matched only ~46% of the time<\/strong>\u2014the same prompt, same minute, different answer.<\/li>\n
- Genuine day-over-day change happens far less often: roughly 1 in 10 days<\/strong> for the top recommendation, once you filter out the noise.<\/li>\n
- Week-over-week, about 22% of a top-5 list turns over<\/strong>\u2014the level where real signal reliably clears the noise floor.<\/li>\n
- Live-retrieval engines (Perplexity, Google AI Mode) are noisiest run-to-run; training-bound engines (Claude, Gemini) are calmer daily but jump on model updates.<\/strong><\/li>\n
- For most B2B brands, weekly aggregated tracking is the right cadence.<\/strong> Daily checks mostly measure noise.<\/li>\n<\/ul>\n
\nWhat is AI answer volatility?<\/h2>\n
AI answer volatility is the frequency and size of change in how an AI engine surfaces a brand\u2014which names it lists, in what order, and whether it cites a source.<\/strong> It is the AI-search analog of ranking fluctuation, but far larger and partly random by design.<\/p>\n
Volatility shows up on three axes:<\/p>\n
- \n
- Inclusion<\/strong> \u2014 does your brand appear in the answer at all?<\/li>\n
- Position<\/strong> \u2014 are you named first, buried mid-list, or last?<\/li>\n
- Citation<\/strong> \u2014 are you linked as a source, mentioned without a link, or omitted?<\/li>\n<\/ul>\n
A brand can hold steady on one axis while swinging wildly on another. You might appear consistently yet rank first only a third of the time\u2014the exact pattern SparkToro found<\/a> when one cancer hospital, City of Hope, appeared in 97% of ChatGPT responses but led the list in only ~35% of them. Tracking a single axis hides this. Useful llm brand tracking<\/strong> watches all three at once.<\/p>\n
Two kinds of change: noise versus real drift<\/h2>\n
Here is the distinction almost every volatility article skips: most AI answer volatility is sampling noise, not movement.<\/strong> You cannot detect a real shift smaller than your noise floor, and separating the two is the whole game.<\/p>\n
We classify every change into three types.<\/p>\n
Run-to-run noise (the floor)<\/h3>\n
Ask the same engine the same question twice in the same minute and the answer still changes. This is stochastic decoding<\/strong>\u2014the model samples from a probability distribution, so the order flips, a name drops, a synonym swaps in. Nothing in the real world moved. In our data this "noise floor" alone accounts for the majority of apparent volatility.<\/p>\n
Day-over-day drift (the signal)<\/h3>\n
Drift is genuine movement in the underlying answer distribution: a competitor's new comparison page gets indexed, a Reddit thread gains traction, a fresh review reshapes sentiment. It is what you actually want to catch. The problem is that on any single day, drift is usually smaller than the noise<\/strong>, so one check can't tell them apart.<\/p>\n
Structural breaks (model updates)<\/h3>\n
A structural break is a step change\u2014a model version upgrade or index refresh that resets the baseline overnight. These are rare but large: a single update can move your mention rate more than a month of organic drift. They are the one event that justifies an unscheduled re-check.<\/p>\n
How we measured it: 90 days, 8 engines, 1,000 prompts<\/h2>\n
To separate noise from drift, you have to measure both at once\u2014so we did. Over 90 consecutive days we ran a fixed set of 1,000 commercial-intent prompts<\/strong> ("best X tool," "X alternatives," "is X good for Y") across eight engines<\/strong>: ChatGPT, Gemini, Perplexity, Claude, Copilot, Grok, Google AI Mode, and AI Overviews.<\/p>\n
The design has two layers:<\/p>\n
- \n
- Noise layer<\/strong> \u2014 each prompt ran 10 times within the same hour<\/strong> to measure run-to-run variance with zero real-world change.<\/li>\n
- Drift layer<\/strong> \u2014 the same prompts ran once daily<\/strong> to capture genuine day-over-day movement.<\/li>\n<\/ul>\n
By subtracting the noise-layer variance from the daily variance, we isolate real<\/em> change from sampling jitter. We logged brand inclusion, list position, and ai citations<\/strong> (linked sources) separately. A few caveats shape the numbers: results come from the consumer web apps, not the APIs\u2014and those can diverge, which is why whether you monitor the API or the web app<\/a> matters before you trust any volatility figure. Prompt design matters too; we used a representative, intent-balanced prompt set<\/a> rather than a handful of vanity queries. The full platform-by-platform 90-day breakdown<\/a> carries the per-engine detail.<\/p>\n
![\"Methodology](\"https:\/\/maxaeo.ai\/blog\/wp-content\/uploads\/2026\/07\/1782474437826-21-37847-1.jpg\")
<\/figure>\nHow often do AI answers actually change?<\/h2>\n
Short answer: far less than it looks.<\/strong> Run-to-run, the top-3 brand set was identical in only 46% of same-hour repeat runs<\/strong>\u2014meaning more than half the apparent "changes" happened with nothing changing in the world. But genuine, noise-adjusted change is much rarer: the top recommendation truly shifts about once every 10\u201311 days, and a top-5 list turns over roughly 22% week-over-week.<\/strong><\/p>\n
- Drift layer<\/strong> \u2014 the same prompts ran once daily<\/strong> to capture genuine day-over-day movement.<\/li>\n<\/ul>\n
- Position<\/strong> \u2014 are you named first, buried mid-list, or last?<\/li>\n
- Genuine day-over-day change happens far less often: roughly 1 in 10 days<\/strong> for the top recommendation, once you filter out the noise.<\/li>\n
![\"Line](\"https:\/\/maxaeo.ai\/blog\/wp-content\/uploads\/2026\/07\/1782474437826-21-37847-2.jpg\")
<\/figure>\n