Written by Uptimal, the team that builds and operates Upbuild and runs marketing analytics for multinational clients across the globe. Everything below reflects how Upbuild actually behaves in production.
Prompts for connected assistants are a different species from prompts for chatty ones: the platform does the fetching, so the prompt’s job is to specify scope, comparison and format. These 25 are the survivors from our client work; adjust names and thresholds and they run as written in Claude, ChatGPT, Copilot Studio or Gemini Enterprise with Upbuild connected.
Daily pulse
- “Blended spend, conversions and ROAS by platform yesterday, versus the same weekday last week; flag moves over 15%.”
- “Any campaign that spent over 100 with zero conversions yesterday, any platform.”
- “Which platform’s CPA moved most, and show the single campaign most responsible.”
Weekly review
- “Write the Monday summary: totals, three movers each direction, one paragraph of interpretation, Slack-ready.”
- “Meta creative fatigue check: frequency over 4 or CTR down 25% week over week, by ad.”
- “Google search terms: over 50 clicks, zero conversions, 30 days; group by theme and propose negatives.”
- “Budget-limited campaigns with lost impression share, trended four weeks.”
Investigations
- “CPA jumped Tuesday: check delivery changes, then GA4 landing-page behaviour, then GTM version history, and tell me which moved first.”
- “Reconcile Meta purchases against GA4 key events for last week; state both attribution windows before comparing.”
- “Which audiences overlap enough to be bidding against ourselves?”
Exec and client-facing
- “Month-to-date versus target, with the two risks I should mention proactively.”
- “Turn this thread’s findings into five slides of bullet content, numbers cited.”
Blended and cross-platform
- “Blended ROAS by platform for the last 28 days versus the previous 28; show spend share next to conversion share and name the platform that is over-funded.”
- “Rank every campaign across Google Ads and Meta by CPA, last 30 days, spend over 500; put the ten worst in a table with their share of total spend.”
- “Weekly spend split between search, social and video for the quarter, and where the mix moved by more than five points.”
Creative
- “Top ten Meta ads by spend this month with CTR, frequency and cost per purchase; mark the ones whose CTR fell two weeks in a row.”
- “Which Google Ads responsive search ad assets have the lowest performance rating, and which would you rewrite first?”
Search terms and keywords
- “Search terms from the last 14 days that contain a competitor name: clicks, cost and conversions, and whether we already have a negative for any of them.”
- “Keywords with quality score under five and spend over 200 this month; list the landing page each one sends to.”
Budgets and pacing
- “Month-to-date spend against a monthly budget of 40,000 per platform; project the month end at the current daily pace and flag anything that will land more than 10% off.”
- “Which campaigns hit their daily budget before noon in the last seven days, and how much impression share did that cost?”
GA4 reconciliation
- “Compare Google Ads conversions with GA4 key events from Google Ads traffic for last week, by campaign; explain the three largest gaps in plain language.”
- “GA4 sessions and engaged sessions by default channel group, last 30 days versus the previous 30, and the landing pages behind the biggest drop.”
Client-facing
- “Draft the client’s monthly performance email: headline numbers, what we changed, what we will change next, under 200 words, no jargon.”
- “Build a one-page quarter review for the client: spend, conversions and CPA by month, top three wins, top three risks, as a table plus five bullets.”
More worked examples are on the connector page; the connection itself is the two-minute setup. Format tip that doubles usefulness: end prompts with the destination (“Slack-ready”, “as a table”, “one paragraph”), assistants follow it religiously.
Ask your ad platforms directly, in Claude, ChatGPT, Copilot Studio or Gemini Enterprise.
Upbuild is an AI platform for marketing data: AI dashboards, insights, exports and alerts, with a read-only MCP server as part of its toolset that brings your accounts into Claude, ChatGPT, Copilot Studio and Gemini Enterprise. Every tool, live data. Free for 14 days on the connector and 7 days on the platform.
Get a demo Start free →Frequently asked questions
Do these work in both Claude, ChatGPT, Copilot Studio and Gemini Enterprise?
Yes; they are written against Upbuild’s tools, not any assistant’s quirks. Claude tends to excel at the long investigations, ChatGPT at quick formatted summaries.
Why do prompts specify thresholds like “over 15%”?
Because connected assistants execute literally: explicit thresholds turn vibes into reproducible checks your team can share and rerun.