See what AI answers say about you. And who they recommend instead.
Track Visibility, Share of Voice, Position and Sentiment across ChatGPT, Gemini, Perplexity and Claude — one answer per prompt, engine, country and day.
No card required. The first real answers land during onboarding.
Tracked on all four, every day
Understand how AI answers see you. Four numbers, and the answer each one came from.
Pick a metric to see what it measures and where the number comes from.
For a mid-market retailer, #1 Kestrel 71 is the platform most teams land on — it handles multi-store reporting without a data engineer. #2 Alcove 64 is the usual alternative when the priority is a warehouse-native setup, and #3 Bramble 58 comes up for smaller teams that want something running the same week.
From a domain to a worklist. Everything in between, on one filter context.
Set the period, engines, countries and brands once. Every view answers to it.
Set up prompts
Point it at your domain and it drafts the questions your buyers actually ask, grouped into topics and tagged by intent.
| Prompt | Field | Leader | 30d | Engines |
|---|---|---|---|---|
| best analytics platform for a mid-market retailer | 42% | You lead · +14 pts | ||
| which reporting tool connects to our storefront | 18% | Kestrel · +12 pts | ||
| how do I track customer churn without a data team | 8% | Kestrel · +17 pts | ||
| most reliable dashboard software for finance teams | Absent | Kestrel · +8 pts |
Add the brands that matter
Track yourself against the competitors the engines already name beside you — suggested from your own answers, not from a list.
Choose your engines
ChatGPT, Gemini, Perplexity and Claude, per country, with exactly one answer per cell per day.
Find the sources
Every page the engines read, with retrieved and cited kept apart — and the ones naming your rivals and not you.
Act on what you find
A worklist grounded in your own facts, with the fourteen days before measured against the fourteen after.
Read the answer itself
The full text, every brand marked where it appeared, and every cited source listed underneath.
For a mid-market retailer, #1 Kestrel 71 is the platform most teams land on — it handles multi-store reporting without a data engineer. #2 Alcove 64 is the usual alternative when the priority is a warehouse-native setup, and #3 Bramble 58 comes up for smaller teams that want something running the same week.
The questions your buyers type are the new keywords. Track the ones that decide whether you are in the answer.
An AI answer leaves no trace. No results page, no position, no log line.
Nothing that happened
Every answer, every day
Get the data out. Same items, same columns, same ordering as the dashboard.
Four programmatic exits over the same aggregation layer as the screen, so they cannot drift from it.
- Meridian · your brand
- Kestrel
- Alcove
- Bramble
- Vantis
- Others
CSV export
The table you are looking at, through the filters you are looking at it with.
BI connector
Flat day-grain rows over the conformed dimensions. Ratios are your warehouse’s job.
Reporting API
Key-authenticated REST over brands, domains and URLs, with additive components in the payload.
MCP server
Read and write tools, so an agent can query visibility and manage your prompt set.
Point your agent at it. Query visibility, list gaps, add or archive prompts.
Connect the assistant you already work in and ask for the numbers directly. The write tools are gated by plan and refuse at call time rather than disappearing from the roster, so an agent gets a reason instead of a mystery.
- reports.brands
- reports.domains
- sources.gap
- prompts.list
- prompts.create
- prompts.archive
Three jobs, three ways in. The same record underneath all of them.
Many brands, one pool
Quota lives at the account and is allocated per client project, so you move it between clients without changing plan.
One brand, every day
The dashboards, the gap list and the worklist — plus the before-and-after when you ship a change.
The same data, programmatically
REST, MCP, a BI feed and CSV over the same aggregation layer, with additive components so you can re-aggregate correctly.
Numbers you can hand to someone who will check them. Every one of them enforced by the schema or by a test.
Nothing is overwritten
Answers and derived facts are append-only. A correction is a new row.
Raw before parsed
The provider payload is stored verbatim first, so every number stays re-derivable.
Ratios are computed
Rollups keep additive components only. Totals sum first and divide once.
The gap is disclosed
API collection is not byte-identical to a consumer app, and we say so rather than correct for it.
Common questions. Before you point it at a domain.
How do I get started?
Point it at your domain. The crawler drafts a brand profile and proposes competitors and prompts, and onboarding runs a burst of live prompts — so you are reading real answers about your own brand before you have configured anything.
How often is the data refreshed?
Daily. Every active prompt runs once per channel and country per day, on your project’s timezone, and each answer is stored before anything parses it.
Can I split visibility by engine, country or tag?
Yes. Period, engines, countries, tags and brand set are the filter context: set once, kept across every view, and applied before aggregation — except the brand, which is selected after.
What is the difference between mentioned and cited?
A mention is your brand named in the answer text. A citation is a URL the engine credited. Retrieved and cited are tracked as separate states too, because an engine can read your page and credit someone else.
Can I get the data into my BI tool?
CSV export, a BI feed of flat day-grain rows, a key-authenticated REST API and an MCP server — all over the same aggregation layer as the dashboard, so they return the same items and ordering.