You already use ChatGPT. This is what changes when it's pointed at the campaigns, budgets, and reports actually on your desk — not a generic tour of prompts.
Five minutes of setup so the lessons themselves are hands-on rather than theoretical. If you already have accounts and access, skip straight to Lesson 1.
Check your AI tool's data settings. If you're on a personal ChatGPT or Claude account, look for the toggle that controls whether your chats are used for model training — turn it off before pasting any real campaign data.
Bookmark your three reporting pages. Google Ads campaign report, Meta Ads Manager breakdown view, and GA4 Explore — one click away, not buried in menus.
Do one low-stakes test run first. Export a single week of one campaign and try the Lesson 1 prompt below before you use this on a live client report.
On data handling: aggregate performance numbers are fine to paste, anything with a named individual's contact details isn't. Full detail is in the "Don't forget" note at the end of this module, but keep it in mind from the first export.
Your numbers live in three places — Google Ads, Meta Ads Manager, and GA4 — and manually cross-referencing them to find what actually changed eats the first hour of your day.
Paste the raw export, not a description of it. Ask AI to find what changed and why, ranked by impact — not to summarize every row, which just restates numbers you can already see.
Here's this week's Google Ads campaign report, pasted below. Identify the three biggest changes vs. last week, ranked by revenue impact. Flag anything that looks like a genuine anomaly rather than normal day-to-day variance. Don't summarize every row — only the ones that matter.
Describing performance in words instead of pasting the actual export table. Without real numbers in front of it, AI will confidently invent a plausible-sounding trend that isn't real.
Meta ad fatigue means constant new variants, and Google's Responsive Search Ads need multiple headline and description combinations. Generic AI copy doesn't pass brand review.
Build a constraints-first brief: audience, tone, banned phrases, and three real examples of your own past top-performing ad copy. Feed the same brief into both platforms, adjusting only for format — Meta primary text versus Google RSA character limits.
Using the three ads below as tone reference, write 5 Meta primary text variants (max 125 characters) and 5 Google RSA headlines (max 30 characters) for [product]. Avoid: [banned phrases]. Match the tone of the examples exactly — don't default to generic marketing language.
Skipping your own past winning ads as reference material. Without them, "on brand" means nothing to the model — it has no idea what your brand actually sounds like.
Stitching Google Ads, Meta, and GA4 data into one narrative for a client or leadership update is repetitive, and the tone needs to shift depending on who's reading it.
One reusable template that turns a metrics dump into what happened / why / what's next, with a second version of the same prompt tuned for client-facing tone versus internal tone.
Combine the Google Ads, Meta, and GA4 data below into a 3-bullet executive summary in a [client-facing / internal] tone: what happened, why it happened, and what we're doing next. Keep it under 100 words.
Accepting the first draft without checking it against the raw numbers. AI tends to smooth over bad news unless you explicitly tell it to stay blunt.
Meta's Campaign Budget Optimization can silently over-allocate spend to one ad set, and Google Ads daily budget pacing can drift for days without triggering an obvious alert.
Run a repeatable pacing health check against a spend-by-day table, deliberately tuned to flag genuine anomalies rather than normal day-of-week variation — like the usual weekend dip in Meta CPMs.
Here's 14 days of spend and results by campaign, across Google Ads and Meta. Find every case where spend and results diverged unexpectedly — ignore normal weekday/weekend variance. List each anomaly with the likely cause.
Asking a leading question like "does this look okay?" AI, like people, tends to confirm what you seem to want to hear. Ask it to find every anomaly, however small, instead.
The weekly or monthly stakeholder deck means turning Google Ads, Meta, and GA4 data into slides — not just a written summary, but a structured narrative with the right level of detail per slide.
Take the report you built in Lesson 3 and reshape it into a slide outline before you touch PowerPoint or Google Slides: one slide per idea, a title that states the takeaway (not just "Overview"), and no more than 3 bullets per slide. Ask AI to draft the outline first, then build the deck from that structure instead of improvising slide-by-slide.
Turn this month's Google Ads, Meta, and GA4 summary (pasted below) into a 5-slide outline for a [leadership / client] audience: Overview, Channel Breakdown, Top Wins, Risks & Anomalies, Next Month's Focus. For each slide, give a title that states the takeaway, not just the topic, plus up to 3 supporting bullets.
Letting every metric earn a slide. The same discipline from Lesson 1 applies here — three things that matter beat twelve things that don't. If a slide's takeaway could be "nothing changed," cut it.
It's easy to paste client campaign data, competitive intelligence, or customer records into a personal AI account without thinking about where that data goes next. Aggregate performance numbers — spend, clicks, conversions — are generally fine. Anything with a named individual's contact details isn't, and a paid or enterprise AI subscription doesn't automatically make everything safe to share. Check your own company's data policy, not the AI vendor's terms — they're not the same thing.
Run through these five on your actual accounts before moving on — Practitioner builds directly on top of them.
