FAQ and troubleshooting

Whether AI engines name your brand, how early, how they describe you and which sources they cite. Each run adds a score and up to ten ranked recommendations. Every metric, defined.

• ChatGPT
• Gemini
• Perplexity
• Claude

All four run on paid plans. Free runs ChatGPT and Gemini. Enterprise adds premium models plus access to 200+ AI models across regions. Questions go in as a buyer types them, with web search on.

No. It reads your public site, so any stack works. The Fix it in Contentstack skill and the Agent OS blueprint are optional extras.

A rank tracker shows where your page ranks. Contentstack Canoe shows whether the AI answer names you, and who it names instead. Use both.

One chat is one answer, from one engine, on one day. Contentstack Canoe runs the same balanced question set every time and scores it in code: a trend, not a screenshot.

Answers vary. Scoring doesn't: same questions and engines, mentions counted in code. Engine scores are combined so one outlier can't swing your number. Sentiment is a model's read, and we label it that way.

No. It tells you what to fix. Contentstack Agent OS or any MCP-ready agent can draft the fixes for your team to review.

Pick a region and language per monitor, so questions run in your buyers' language. Major languages are supported.

Whether AI crawlers can reach and read your site: robots.txt rules, CDN or firewall blocks, noindex or canonical problems on your homepage and whether your sitemap can be found. It also checks a sample of your pages for broken pages, titles, headings, thin content and more, and says how to fix each issue. It runs with your monitors on every plan, Free included.

Your website address, to draft your brand profile from your public site. No tag, no analytics or CMS access. The site audit also reads a sample of your public pages. Your first report takes about two minutes.

The monitor is still created and flagged. The engines find you through web search, and your profile is drafted from it for you to review. The site audit shows which AI crawlers look blocked and how to fix it. More in Getting started.

In AWS North America, on isolated infrastructure with separate workspaces. Data residency comes with Enterprise (coming soon). See privacy and data processing and the Contentstack trust center (opens in a new tab).

Passwords are stored hashed, never in plain text, and sessions are revocable. API keys are shown once, stored as a hash and stop at once when revoked.

Not yet. SSO and SCIM come with Enterprise (coming soon). You can sign in with Google today.

One question asked of one engine. A 20-question monitor on four engines uses 80 prompts a run. Free includes 20 a month, Growth 2,000.

No. No card, no expiry: one brand and a fresh report every month on ChatGPT and Gemini.

A run that would go over your allowance won't start. Allowances reset monthly. Growth can add prompt packs. Plans and limits

On Growth: $119 one-time (launch price) for 400 prompts, up to four packs. They're used after your monthly allowance, roll over and expire 12 months after purchase.

Reads don't. Calls count toward your monthly API and MCP requests (10,000 on Growth). Starting a run with trigger_geo_test_run or the webhook spends prompts.

It's coming soon and open to every team, Contentstack customer or not. Contact us with your brand count and needs.

The client didn't send the Authorization header, often because it tried OAuth, which isn't supported yet. Add Authorization: Bearer <key>. Config for each client

MCP needs a Read & write key, even to read. Scope is fixed, so create a new key and revoke the old one. Read only MCP keys are coming soon. Until then, block trigger_geo_test_run.

The workspace is on Free. Upgrade to Growth and the same key works.

VS Code uses servers, while Cursor and Claude Code use mcpServers. Check that CONTENTSTACK_CANOE_API_KEY is set where the client starts, then restart it.

Check run_status in get_geo_latest_report. Mid-run, the report shows the previous completed run, and _guidance says so.

already_running: a run is in flight, so poll until it ends. Out of prompts: the run is accepted but won't start, and run_status turns to error. Check Usage in Settings.

Not yet. It needs OAuth sign-in, which is coming soon. Until then, use it without a connector: a ChatGPT Project with our playbook and your report file. Every client marked Available under Connect your client works today.

Send a GET, not a POST, with the header name and passkey from the Web hook dialog. Only On Demand monitors show the Web hook option. A 200 with skipped: true means the URL is wrong or the monitor was deleted. Webhook trigger