Skip to content
Syncfusion GTM Playbook

AI answer monitoring

Type
checklist
Status
draft
Owner
COE Team
Last reviewed
2026-11-07
Products
All products

Draft. Not yet reviewed. Check facts before you use it with customers.

Run monthly (first week) for every product to record whether AI assistants mention, cite and describe us accurately; done means the results table is filled for every prompt and engine and the gap list is handed to the SEO/GEO program. The Claude skill ai-answer-monitoring runs this checklist.

Before you start

  • Why this matters and what we can influence: Search and AI visibility.
  • Use the fixed prompt set below unchanged month to month. Add prompts only at a quarterly review.
  • Use a clean session (no memory or personalization) and record the country.
  • Engines: ChatGPT (with search), Gemini, Google AI Overviews (a normal Google search), Microsoft Copilot, Perplexity, Claude.
  • Asking engines through the API: when the openrouter MCP server is connected (see MCP servers and tool guardrails), use its ask_engine tool for the ChatGPT, Gemini, Perplexity and Claude rows. Each call is a clean session (one message, no memory), so it also covers Claude when you are running inside this repo. Call list_engines first and use the models it lists. API answers approximate the consumer apps but aren’t identical, so start each such row’s Note with API: {model} and keep the answer text and cited URLs in raw/. Google AI Overviews and Microsoft Copilot have no API: check them by hand or mark them “n/a”.
  • Product-specific prompts live in each product’s GTM plan, section “AI-monitoring prompts”. Competitor names come from _reference/competitors/.

Fixed prompt set (per product)

ID Type Prompt pattern
P1 Category “What are the best {category}?”
P2 Enterprise “What is the best {category} for an enterprise / midsize company?”
P3 Comparison “{Our product} vs {top competitor}: which should I choose?”
P4 Alternatives “What are the best alternatives to {top competitor}?”
P5 Brand facts “What is {our product}, how is it licensed and who uses it?”
P6 Use case One high-intent job from the GTM plan (for example a migration or integration question)
P7 Trust “Is {our product} secure and compliant for enterprise use?”
Product {category}
essential-studio UI component library for .NET/JavaScript; PDF/Word/Excel library for .NET (run P1–P2 for both)
bold-bi embedded analytics / embedded BI platform
bold-reports embedded reporting tool; SSRS alternative
boldsign e-signature software and e-signature API
bolddesk AI help desk / AI customer support software
boldcreate TODO: verify category from the public site
bolddocserver Office Online Server replacement for SharePoint Server

Checklist

Phase 1: run prompts (marketing analyst or Claude, first week of the month)

  • Run P1–P7 for each product on each engine; keep the answer text or a screenshot. Use ask_engine for the engines it lists (see “Before you start”); judge Mentioned, Position, Cited and Accuracy from the answer text and cited URLs it returns.
  • For each answer, record: Mentioned (Y/N), Position (1st, 2nd… in a list, or ”-”), Cited (Y/N, our URL linked as a source), Cited URL, Competitors mentioned, Accuracy (Correct / Outdated / Wrong / n/a) with a note.
  • Any wrong pricing, licensing, platform or compliance statement counts as Wrong.

Phase 2: compare and act (SEO lead, same week)

  • Compare with last month: mention rate and citation rate per product and per engine.
  • List gaps: prompts where we’re absent, where only competitors appear, or where facts are wrong.
  • For each gap, name the page that should answer it and the fix (refresh, new FAQ, structured data, third-party listing). Hand the list to the SEO/GEO program and the blog program.
  • Post the summary deltas for the weekly GTM review.

Results format

Claude prints the result in chat; save it to the team’s shared results log (location TBD). If an output directory is given (for example by pbmaintain), also write the report files defined in reports/schema/README.md there. Otherwise don’t create results files in the repo. Use exactly this format:

## AI answer monitoring: {product slug}, {YYYY-MM}
Run by: {name or "Claude"} | Location: {country} | Date: {YYYY-MM-DD}


| ID | Engine | Mentioned | Position | Cited | Cited URL | Competitors mentioned | Accuracy | Note |
|---|---|---|---|---|---|---|---|---|
| P1 | ChatGPT | Y | 2 | N | - | {names} | Correct | |


### Summary
- Mention rate: {x of y} ({change vs last month})
- Citation rate: {x of y} ({change vs last month})
- Wrong or outdated facts: {list, with engine and prompt ID}


### Gaps and fixes
| Prompt ID | Engine | Gap | Page to fix | Fix |
|---|---|---|---|---|

Report keys (for the report files): Item IDs are the prompt ID plus the engine (P3-chatgpt, P5-perplexity). Metric keys: ai.mention_rate, ai.citation_rate (dimension engine, or none for the total), ai.wrong_facts.

Done when

  • Every prompt has a row for every engine, for every product.
  • Month-over-month deltas are calculated and the gap list is handed off.
  • Wrong facts are logged with the page that should correct them.

If something goes wrong

  • Engine unavailable: mark the row “n/a” and rerun within the week.
  • An answer makes a harmful or seriously wrong claim about a product (security, compliance, pricing): tell the product marketing lead and the COE Team the same day.
  • Results vary: run twice and record both if they differ.