What can I export, and from where?
| Data | Where | Format |
|---|---|---|
| Sampled answers | AI Visibility, Export N sampled answers in the report header | CSV |
| Citations | AI Visibility, Citations tab, Export N citations in the header | CSV |
| Tracked prompts and their results | Prompts tab, export menu | CSV or JSON, download or copy |
| Engine searches | Prompts tab, Engine searches, export menu | CSV or JSON |
| Your pages | Websites, the pages list | CSV |
| AI crawler visits | AI Traffic Analytics | CSV |
How do I export every answer for a period?
- Open AI Visibility for the brand.
- Pick the date range in the toolbar.
- Click Export N sampled answers, where N is the number of answers in the range.
- The CSV downloads with one row per answer, in the columns the server defines, so a new column is never dropped.
One file holds up to 20,000 rows. If a range holds more, the app says it exported the first 20,000, so a file never quietly contains less than the button promised. Empty cells mean no observation exists, never zero.
Can I export citation data for reporting?
Yes. The Citations tab exports every cited page in the range as CSV, ready for a spreadsheet or a BI tool, and on a Custom plan our team sets up a feed into Looker Studio or the BI tool you already use. For a page a client or a boss can open without an account, use Share Report instead, described in Integrations and reporting.
Is there a REST API?
Yes. The Reports API at app.proofsource.co/v2 returns your data as JSON: /v2/reports/visibility, /v2/reports/citations, /v2/reports/sentiment and /v2/reports/fanouts, authenticated with an API key from Settings. Its OpenAPI spec is published at app.proofsource.co/v2/openapi.json, so you can generate a client in any language.
Can my own AI agent read my data?
Yes, over the Model Context Protocol (MCP). An AI agent such as Claude Code, Claude Desktop, Cursor or one built on the OpenAI API reads the same numbers you see in the app, from the server at app.proofsource.co/v1/mcp.
- In the app, open Settings, then API Keys, and create a key. Copy it straight away: it is shown only once.
- Choose read for an agent that only looks. Choose write only if it should also add questions or mark fixes shipped. A key acts with your role in the workspace.
- Open Settings, then AI agents, copy the server URL and the setup snippet for your client, and paste your key where it says.
| Tool | What it returns or does |
|---|---|
| list_brands | Your tracked brands and their ids. |
| list_topics | Topics and the buyer questions tracked under each. |
| get_visibility | How often AI answers mention you, by engine or by question. |
| get_competitors | Share of voice against the competitors AI answers name. |
| get_citations | The sources AI answers cite, and where your domains rank. |
| get_changes | What moved: visibility trend, answer churn, rival site changes, new sources. |
| list_fix_opportunities | What keeps engines from citing you, and the fix. |
| get_fix_verification | Each shipped fix's before and after, and whether it worked. |
| create_prompt | Track a new buyer question. Needs a write key. |
| mark_fix_shipped | Date a fix so it can be verified. Needs a write key. |
Each key can make 60 tool calls a minute and 30 writes an hour. ChatGPT's and claude.ai's connector screens expect OAuth sign-in, so connect through one of the clients above with your API key.
