Core concepts
Every report is built from a few pieces: a brand, a monitor, buyer questions and the engines that answer them.
How the pieces fit
Workspaces hold brands. Brands hold monitors. Every time a monitor asks its questions, that's a run: the report you read.
- Workspace Your team
-
Brand
Tarnwell Cycles
API:
project -
Monitor
City e-bikes
API:
geo test -
Run
Oct 1 report
API:
report
Workspaces
Where your team, plan and data live, with unlimited usersFair use policy applies on every plan. The creator is the Owner and invitees join as Member. Only the owner buys add-ons or changes the plan. Switch workspaces from the menu in the top bar.
The monthly prompt allowance and API keys belong to the workspace, so every brand and monitor in it draws from one pool.
A fair use policy applies.
Brands
The company you track. Contentstack Canoe reads its public site and drafts a profile, and every question is written from that profile.
| Profile field | What it's for | Limit |
|---|---|---|
| Domain | Identifies the brand. Every monitor targets it. Can't be changed. | One |
| Brand name | The name searched for in answers. | One |
| Brand name variations | Acronyms, casings and misspellings that count as you. | Up to 5 |
| Business type | B2B, B2C or Both. Frames questions and personas. | One |
| Vertical | Your category, like "electric bikes for city riders". | One |
| Competitors | Brands you're compared against, each with a domain and name variants. Five are suggested. | Up to 10 |
| Products | What you sell. Each monitor focuses on one. | Up to 10 |
| Personas | Your buyers: job titles, industries, company sizes and locations. | Up to 10 |
Edit it any time in Brand settings (the gear icon). Changes shape monitors you create afterwards; existing monitors keep their questions. Separate sites for separate brands should be separate brands.
Monitors
A saved set of buyer questions asked on a schedule. It's the thing you measure, defined by:
- One product, so the questions stay specific.
- One to ten personas, the buyers the questions are written for.
- A region and a language.
- A question set, written once and reused on every run.
- A schedule: Daily, Weekly, Monthly or On Demand.
Reusing questions is the point: run twelve asks what run one asked, so a score change means the answers changed.
Tip: one monitor per product and market. Tarnwell Cycles might run "City e-bikes, urban commuters, United Kingdom" and "Cargo e-bikes, parents, European Union" as two monitors.
Schedules
Pause, resume or Run now from the monitor's row on the brand page.
| Schedule | Runs | Plans |
|---|---|---|
| Monthly | Once a month. The default on Free. | Free, Growth |
| Weekly | Once a week. The default on Growth. | Growth |
| Daily | Every day. | Growth |
| On Demand | When you run it, or when a webhook triggers it on Growth. | Free, Growth |
Questions
Drafted from your products, personas and competitors, using live web research on what buyers in your space ask AI. Every set is balanced on purpose across four kinds of question:
Four kinds of question
- Category questionsNever name you.
Best e-bike for a 10 km commute?
- Brand questionsAsk about you directly.
Is Tarnwell City reliable in winter?
- ComparisonsYou against a named competitor.
Tarnwell City vs Aldren Metro
- Alternatives questionsWho else to consider.
Alternatives to Tarnwell for city riding
Each question is tagged Awareness, Consideration or Decision. Questions that name you usually sit later in the buyer journey. The question limit is split across the monitor's personas.
Free runs up to 10 generated questions; Growth runs up to 100 you can edit. See Plans and limits.
The app calls each question a "prompt". For usage, a prompt is one question asked of one engine, so a 20-row Prompts tab on four engines is 80 prompts.
Engines
- ChatGPT
- Gemini
- Perplexity
- Claude
Free runs ChatGPT and Gemini; Growth runs all four. Engines outside your plan show as locked in the app.
Each question goes to each engine exactly as a buyer would type it, with web search on. Brand Ranking and share of voice come from a separate read of who the engines think leads your market.
Answers vary run to run, as they do for your buyers. The questions, scoring and citation analysis don't, so trust the trend.
Runs and reports
A run is one pass: every question on every engine in your plan. On Growth, any run downloads as a PDF. The first starts when you create the monitor and takes about two minutes:
The engines answer.
Mentions, position and sources are counted in code, with sources sorted into yours, competitors' and third-party.
A model reads the tone, and the report labels it that way.
Scores and changes are worked out.
Gaps become recommendations: up to ten ranked fixes.
Execution History logs every run with its scores and a status:
- SuccessEvery engine answered.
- PartialAt least one engine was unavailable. Scores cover the rest.
- FailedNo engine responded. It costs nothing.
Prompts: the unit of usage
One prompt is one question asked of one engine. A run costs its questions times its engines.
Only runs spend prompts. Adding a brand, reading your site, refreshing the profile and writing questions are free. Some screens, like Usage in Settings, say credits.
Region and language
A monitor's Location scopes its questions, competitor suggestions and the market leaderboard. Options: Global (no regional bias), United States, Canada, United Kingdom, European Union, Germany, France, India, Japan, Singapore, Australia, United Arab Emirates, Brazil and Mexico.
Questions are written in the language you pick, and major languages are supported. Reports and recommendations are in English.
Names in the API and MCP
The API and MCP server (Growth and Enterprise) and some older screens use internal names. They map one to one:
| In the app | In the API and MCP | Example |
|---|---|---|
| Brand | project | list_projects, /api/projects |
| Monitor (GEO monitor) | geo test, test | list_geo_tests, /api/test |
| Run | report, execution | get_geo_latest_report |
| Question | prompt | get_geo_prompt_details |
| Persona | icp | Personas on a brand profile |