AI answer monitoring
Your buyers already asked.
We track what AI answered.
We run the questions people actually type into ChatGPT, Claude, Gemini, Perplexity and five more models — on a schedule, with your brand name left out of every organic prompt. What comes back is the answer your buyer got.
Share of Voice
Rankings
best protein-bar question set — brand names are never injected into organic prompts, so absence is a signal.
Organic share of voice
Ask the question your buyer asks. Not the one that flatters you.
Put brand names in the prompt and the model will say them back. That is a mirror, not a measurement.
Rank these brands: Alpine Bars, Summit Snacks, Ridgeline Nutrition.
All three get named. You learn nothing you did not already put in.
What's the best protein bar?
The model answers freely. Being left out is a result, and it counts.
No brand injection
Prompts are written from your market category, never from a list of brands. Every name in the answer got there on its own merit.
never sent with the prompt
Tiered mention parsing
Each answer is graded, and the tiers are not worth the same. Being the top pick counts for fifteen times what a passing mention does.
Statistical confidence
Share comes from repeated runs across models and days, so one lucky answer never moves your number.
runs × models × days
Prompt intelligence
Nine kinds of question. Nine models. One rotation.
Buyers do not ask one question. Coverage is graded across the whole intent spread, not just the prompt with your name in it.
Brand discovery
“Who makes climbing-friendly protein bars?”
Recommendations
“What's the best protein bar for climbing?”
Comparisons
“Alpine Bars vs Summit Snacks — which is better?”
Use case
“What should I eat on a multi-pitch day?”
Trust & authority
“Which protein bar brands are third-party tested?”
Pricing & buying
“Where can I buy protein bars in bulk?”
Industry landscape
“How is the protein bar market changing?”
Troubleshooting
“Why do protein bars upset my stomach?”
Listicle & roundup
“Best protein bars of the year”
Provider rotation
Prompts are executed across the models on your plan — two on Starter, four on Pro, all nine on Premium — so no single model’s mood becomes your score.
Freshness
Every stored answer keeps its timestamp and its source list, so you can tell a score built on this week’s answers from one built on last quarter’s.
Source influence
Retrieved is not the same as cited.
A model can open your page and still not name it. The distance between those two is where the work is.
- Retrieved
- The model fetched the page while it was working out what to say.
- Cited
- The model named the page in the answer your buyer actually read.
85%
of brand mentions sit on domains a brand does not own — reviews, forums, roundups, comparison pages. Your own site is one voice in the room.
Industry research on AI answer citations. Not an Omni Impact benchmark.
Source InfluencePro+
7 days30 days90 daysLeaderboard
8 of 63 domains
Your own domain is retrieved more often than it is cited — the page is being read, then set aside.
Search Lenses
Native search is one retrieval path. Run the same prompt through two more.
Every model already retrieves in its own way when it answers. A lens reruns the identical prompt through Tavily or Exa instead, so independent retrieval sits right next to what native search already showed you.
- Native
- Each model answers with whatever retrieval it already does on its own — nothing added, nothing swapped.
- Lens
- The same prompt, re-run through Tavily or Exa instead of the model’s own search — set per prompt, not per account.
Tavily
ExaLens runs never change your scores — they sit beside them.
Perception
Not just whether you are named. How you are framed.
Every answer is read for sentiment, prominence, certainty and the story the model is telling about you.
Brand perception· alpinebars.com
30 days9 modelsNarrative framing
Specialistdominant
Recent AI Mentions
“For alpine routes specifically, Alpine Bars is the specialist pick — precise macros, built for long days on the wall.”
ChatGPT · Recommendations · 2h ago
Sentiment
+0.41Positive
Up close
Four readings from the same set of answers.
Every run leaves a trail: how the score moved, how each kind of question frames you, which domains the models lean on, and where an answer got the facts wrong.
Watch the score move
Your visibility score is recomputed after every run and plotted across the window. Each point is a day of answers, so a fix you shipped on Tuesday shows up as movement rather than a hunch.
How every answer frames you
Answers are read for sentiment and graded by the kind of question that produced them — recommendations, comparisons, use case, trust, pricing. The shape shows which questions you win and which ones leave you thin.
Who speaks for you
Every citation is traced to the domain that earned it and ranked by how often the models lean on it. Your own site sits in that list next to the reviews and roundups that outrank it.
Alpine Bars discontinued its climbing line.
Seen in 3 of 41 answers this week
The climbing line is core — 6 SKUs, restocked monthly, cited from alpinebars.com.
What the models get wrong
Models repeat details that are out of date — a line you never discontinued, a spec that was never yours. Each claim is checked against what your pages actually say, and the correction is filed next to it.