TL;DR
- An AI assistant connected to your actual ad accounts answers a different class of question than a dashboard: "where did I waste money?" instead of "here are 40 columns, good luck."
- Below: 36 real questions, grouped by the job you're trying to do — every one backed by a specific capability in Adstudio's assistant, not a hypothetical.
- Read questions get answered immediately. Write questions (pause, budget, keywords) come back as a proposal with a Confirm button — nothing changes in your account without your click.
- It's not magic: no billing, no policy appeals, no creative uploads, keyword actions are Google-only. The honest limits are at the bottom.
Every dashboard answers the question it was built for. The problem is that your actual questions on a Tuesday morning — "is anything on fire?", "why did CPA jump?", "which of these 60 ad sets is dead weight?" — cut across a dozen screens. An AI assistant with real, authenticated access to your Google Ads and Meta accounts answers those questions directly, because it can run the analysis instead of showing you a table and wishing you luck.
We build one of these (the assistant inside Adstudio), so the usual disclosure applies. But that's also why this list is trustworthy in a way most "50 ChatGPT prompts for marketers" listicles aren't: every question below maps to a specific tool the assistant actually has. Nothing here is aspirational. Where something only works on one platform, we say so.
1. "How did we do this week compared to last week?" The assistant pulls your account metrics for both periods and renders a side-by-side comparison — spend, clicks, conversions, CPA — with the deltas computed for you. This is the single most-used question, and it's the one every native dashboard makes you build a custom date comparison for.
2. "Give me a quick performance summary across Google and Meta." One cross-platform summary instead of two tabs. Because the assistant sits on top of both integrations, "how are ads going?" doesn't require you to specify which ads.
3. "Show me daily numbers for the last two weeks." Same metrics, day-by-day table — useful for spotting the exact day something broke rather than a smoothed weekly average.
4. "Break this down month by month." A monthly breakdown at the account or campaign level, so seasonality and slow drifts become visible. On Meta, this works down to the ad set level too — "show me this ad set's monthly trend" is a real question, not a pivot-table project.
5. "Which campaigns are running right now, and what's each spending?" A live campaign list with status, budget, and metrics. Sounds basic; it's the foundation everything else builds on, and it means "the three worst campaigns" in a later question refers to real, current entities.
6. "Tell me everything about the Summer Sale campaign." A single-campaign deep dive: settings, budget, delivery, and its metrics over whatever period you name. The follow-up questions ("and last month?") reuse the context — you don't re-specify the campaign every message.
7. "Where did I waste money last month?" This runs an actual waste analysis across your campaigns: spend with zero conversions, negative-ROI segments, budget going to things that demonstrably don't pay back. The output is a ranked report with amounts, not a vague "consider reviewing your campaigns." We wrote a whole post on finding wasted spend in Google Ads — this question is that post, compressed to one sentence.
8. "Is my conversion tracking actually set up correctly?" A conversion-setup audit: which conversion actions exist, which campaigns are optimizing toward them, and where the wiring is missing or contradictory. Broken tracking is the most expensive silent failure in paid media, and it's exactly the thing nobody checks until the numbers stop making sense.
9. "Are my bidding strategies working against my goals?" A bidding audit that flags misalignments — like a Maximize Conversions strategy on a campaign whose conversion tracking is thin, or targets that contradict the campaign's actual history.
10. "What's the overall health of my account?" A health score with category breakdowns, so "fine, mostly, except tracking" is an actual answer instead of a feeling.
11. "What should I fix first?" Prioritized, personalized recommendations across the account. The value isn't that the AI has opinions — it's that the recommendations are ranked against your data, so you start with the highest-impact fix.
Keyword intelligence is Google-only — Meta doesn't have keywords, so no tool honestly can offer this cross-platform.
12. "Which keywords have low Quality Scores?" Keyword performance with the Quality Score breakdown — expected CTR, ad relevance, landing page experience — so you see why a score is low, not just that it is.
13. "What actual searches triggered my ads this week?" The search terms report, conversationally. This is where the money leaks: the queries you're matching that you never intended to.
14. "Which search terms should I add as negatives?" A keyword-opportunity analysis that flags spending, non-converting search terms as negative candidates and promising terms as keyword candidates — the analysis a good freelancer runs monthly, on demand.
15. "Add 'project management software' as a phrase-match keyword to the Software ad group." This is an action, so it follows the confirmation flow: the assistant prepares the change, shows you exactly what will be added where, and waits for your Confirm click. Nothing lands in the account on its own.
16. "Add 'free' as a negative keyword to that campaign." Same proposal-then-confirm pattern. Ask in plain language, review the card, click Confirm — the negative is live.
17. "Pause the keyword that's spent $400 with no conversions." The assistant identifies it from the data you were just looking at and proposes the pause. Confirm, done.
18. "Which ad sets performed best this month?" Ad groups (Google) or ad sets (Meta) with their metrics, sortable by whatever you care about. "Best" can mean CPA, ROAS, or volume — say which and it ranks accordingly.
19. "Show me my ads ranked by CTR." Ad-level performance including creative details, so you're comparing actual headlines and images, not ad IDs.
20. "Why is this ad underperforming?" A detailed single-ad view — for Google responsive search ads, that includes asset-level performance, so you can see which specific headlines Google rates poorly.
21. "Write me five new headline variations for this ad." Generated ad copy suggestions, informed by the ad's actual context. You still choose; it drafts. (This mirrors how most PPC pros already use AI — copy drafting was the top LLM use case in the State of PPC 2026 survey.)
22. "Update this ad's second headline to the one we just discussed." An ad-copy edit as a change proposal: old text, new text, Confirm button.
23. "Duplicate this ad into the other ad set." Meta ad duplication, proposed and confirmed. Useful when a winner should be tested against a different audience.
24. "Clone this Google ad into the new ad group." The Google equivalent — same duplicate-with-confirmation flow, separate machinery under the hood.
25. "Pause the three worst-performing campaigns." A bulk status change: the assistant identifies the three from your data, lists them explicitly in one proposal, and pauses all of them on a single Confirm. One card, not three.
26. "Re-enable the retargeting campaign." Single status change, same pattern. Works on campaigns, ad sets, ad groups, and individual ads, on both Google and Meta.
27. "Raise the Summer Sale budget to $150/day." A budget update proposal showing current budget, new budget, and the implied change before you approve it.
28. "Archive those old test ad sets." Delete/archive — singly or up to 50 entities in one bulk proposal. This one is Meta-only: Google Ads removal semantics are different, and the assistant doesn't pretend otherwise.
29. "How much have we spent across all platforms this month?" A cross-platform spend summary — Google + Meta in one number with the per-platform split. The question behind every Friday spend spreadsheet, answered in one line.
30. "What's my total ad spend this billing period?" Your tracked ad spend, as the platform measures it — useful for pacing against a monthly budget without exporting anything.
31. "Build me a report with spend, conversions, and CPA by campaign for last month." A custom report generated from your description — you name the metrics, dimensions, and period; it assembles the table. This is the fastest path from "the client asked for X" to X.
32. "What plan am I on, and what does it include?" Subscription details on demand. Minor, but it means the assistant can answer questions about itself honestly.
33. "Which ad platforms are connected right now?" Integration status for every platform — and when something you ask for needs a platform that isn't connected, the assistant tells you that instead of hallucinating data. That failure mode (confidently inventing numbers) is the main reason generic chatbots can't do this job.
34. "Which demographics actually convert for us?" Audience insights broken down by demographics — age, gender, and segment performance — so "who is this working on?" has a data answer.
35. "Design a Google search campaign for our new product at $50/day." A full campaign proposal — structure, settings, keywords, budget — built as a reviewable draft. You adjust and confirm before anything is created. It's a starting scaffold, not a replacement for campaign strategy (more on that below).
36. "Set up a Meta awareness campaign for our brand." The Meta equivalent, with an interactive builder: you pick the page and creative image as part of the flow, then confirm. Awareness is currently the supported objective for Meta creation — conversions campaigns still deserve the native editor's full option set.
An honest capability list needs the other half:
Do I need to learn prompt engineering to use this? No. Every question above is written exactly as you'd type it. The assistant resolves dates ("last week", "this month"), looks up entities by name, and picks the right analysis itself. If a question is ambiguous, it asks — with clickable options, not a lecture.
Can the AI change my campaigns without me knowing? No. Read questions execute immediately; anything that writes to your account — pausing, budgets, keywords, copy, deletion, creation — is returned as a proposal card with Confirm and Reject buttons. Until you click Confirm, your account is untouched.
How is this different from asking ChatGPT about my ads? ChatGPT doesn't have your data — you'd paste in exports and get generic analysis on stale numbers. An assistant with authenticated account access runs these questions against live data and can carry approved changes back into the account. We compared the categories properly in Best AI tools for PPC managers.
Which platforms does this cover? Analysis and actions run on Google Ads and Meta today, with cross-platform spend reporting across connected integrations. If a question needs a platform you haven't connected, the assistant says so and points you to the integration — it never fabricates.
What does it cost to try? Adstudio has a free plan; paid plans start at $19/month, flat — not tiered on your ad spend. Connecting an account and asking question #7 at your own account is the fastest way to find out if this list is real.