Track the buyers who ask ChatGPT first.
ChatGPT, Claude, Gemini, and Perplexity are recommending vendors in your market right now — with or without you. AI Visibility measures them the way researchers study customers, not the way SEO tools chase rankings: read what buyers are told, then follow the citations to where you need to be.
usage-based · from $50/mo
An assistant does not rank you. It describes you — differently every time.
There is no position to hold. The same question returns a different answer depending on how it’s worded, which assistant is asked, who seems to be asking, whether the model searched the web or answered from memory, and what day it is. Your site can sit perfectly still while the answers about you move.
A single tracked number can’t tell you why it moved. A measured set of answers can — which is why this tool works like market research, not like a rank tracker.
You write the questions. Personas fan them out.
Write the questions your buyers ask. For each one, the tool generates a version per persona you sell to — because a founder asks about price, an ops lead asks about integrations, and the assistants answer each of them differently. Every generated prompt is on the table before anything runs: read each one, edit each one, and see what the whole run will cost.
startup founder“Cheapest way for a five-person team to stop approving invoices by hand?”
operations lead“Best invoice approval tool that plugs into our existing ERP?”
enterprise buyer“Invoice approval automation with audit trails for a compliance-heavy org?”It shows you which sites the assistants trust, so you know what to publish next.
Every cited source is collected and sorted by what kind of page it is. Comparison questions pull comparison pages; “how do I” questions pull documentation. The pattern tells you which format to publish — and where the assistants are settling for something weak because nothing better exists.
A four-year-old forum thread getting cited is not trivia. It’s a question with no owner, and owning it is cheap.
You decide how much to measure, and you pay for exactly that.
No tiers and no seats. You choose how many questions, how many personas, how many assistants, and how often the panel runs — that’s your usage, and that’s your bill. Scale up for a full market study; scale down to a small standing panel; change it whenever you want.
The cost of a run is shown before you start it. Never discovered on an invoice.
Usage-based pricing
Includes $50 of usage credits every month, good on any Foundry tool. Every run shows its exact price before you start it.
Match runs to how often you publish.
Your publishing cadence sets your measurement cadence.
An answer only changes when the things behind it change. So measure as often as you give it something new to read — and when a competitor launches, measure through it.
Measure slower than you publish and you cannot tell which change moved the number. Measure faster and you are watching noise.
- One press piece a monthMeasure monthly
- Five videos a weekMeasure weekly
- Rivals shipping every few daysMeasure daily
Monthly is the floor even at a standstill: the sources behind your answers turn over that fast whether you publish or not.
Every run produces the same report.
Same sections, same arithmetic, every time — which is what makes one run comparable to the last.
| Visibility score | One number with its denominator attached: named in 41% of 320 answers, not a naked 41%. |
|---|---|
| Results by assistant | Where you stand on each platform — strong on one, invisible on another is the common finding. |
| Results by persona | The founder and the enterprise buyer get different answers. See both. |
| Every answer, in full | The complete stored text of every response, brand mentions marked. |
| Cited sources, categorized | Which pages the assistants relied on, sorted by what kind of page they are. |
| Movement since last run | What changed, per question — never a delta across two different question sets. |
| PDF export | The whole report as a dated document you can put in front of anyone. |