What to ask
AdPilot MCP works best when you name the app, a time range in complete days, and the decision you want. Copy a prompt below and adjust names.
- Orientation
- Performance and profitability
- Diagnostics
- Bounded optimizations
- Limits and budget planning
- Memory and follow-through
- Tips that improve answers
- Worked example: turn a vague question into a decision
- How AdPilot uses your prompt
- Next steps
- Official sources
Orientation
- List my apps and which ones have Apple Ads campaigns mapped.
- Orient me to this Apple Ads account: structure, Search Match, and anything unusual.
- Which campaigns are mapped to [App Name], and which are still unmapped?
Performance and profitability
- Review all campaigns for [App Name] for the last 30 complete days.
- Where did I waste money this month relative to installs and RevenueCat proceeds?
- Compare this week to the previous week for [App Name] using equal complete-day periods.
- Calculate TTR, conversion rate, CPT, CPI, and observed app-level ROAS for [App Name]. Show the formulas and data-quality caveats.
- Compare mapped Apple spend with blended RevenueCat proceeds for [App Name]. Do not assign proceeds to individual campaigns.
Diagnostics
- Why is [Campaign Name] barely spending?
- Find search-term waste for [App Name], exclude protected intent, and summarize candidates.
- Analyze under-delivery for campaigns that are limited by budget or bid.
- Show keywords with high spend and weak conversion over the last 14 complete days.
Bounded optimizations
State the outcome you want and any business constraints. Mutation tools execute directly within configured limits.
- Find search-term waste and prepare exact-negative operations for my review.
- Harvest converting search terms into keyword operations for [Ad Group].
- Reduce bids for keywords over my target CPA, capped by my workspace limits.
- Prepare pause operations for clearly wasted keywords that are not protected.
Then open Changes to inspect action IDs, before/after values, provider results, and evaluation windows.
Limits and budget planning
- Explain my workspace maximum projection campaign by campaign and compare it with every applicable cap.
- Which enabled daily budgets make my projected month exceed the workspace ceiling? Do not change anything.
- How much would restoring [Campaign Name] add from today through month-end?
- Would increasing [Campaign Name] from $5/day to $7/day fit both the app cap and workspace ceiling?
- Recommend campaign changes that would reduce exposure toward my confirmed limits. Do not execute them.
The assistant can explain the calculation, but the Limits page remains the authoritative current projection. Read Spend limits and projections before raising a ceiling.
Memory and follow-through
- What changes did we make last week, and did they help?
- Evaluate whether lowering the bid on [Keyword] improved performance using equal complete-day periods.
- Recall recent decisions and strategy notes for [App Name].
Tips that improve answers
- Prefer complete days and named apps over vague “lately / everything.”
- Ask for caveats when profitability numbers look too good or too bad.
- Keep protected terms updated on the app so negatives and pauses remain protected.
- Choose the matching skill when you want a repeatable workflow rather than a one-off question.
Worked example: turn a vague question into a decision
Instead of How are my ads doing?, ask: Review Calm Budget for the last 30 complete days. Compare the previous 30 days, calculate TTR, conversion rate, CPT, CPI, and observed app-level ROAS, identify the three largest changes, explain data-quality caveats, and recommend next actions without applying them.
This supplies scope, comparison, formulas, priorities, and an explicit read-only boundary.
How AdPilot uses your prompt
AdPilot passes the authorized workspace, app context, connected data, relevant available skill, and server-enforced limits to the requested tools. Precise names, dates, and decision criteria reduce ambiguity; they do not bypass permissions or validation checks.