20 Meta Ads prompts that return something useful

Most Meta Ads prompts circulating online fail for the same reason: they ask for analysis of an account the assistant cannot see. The answer comes back well written, cites the right metrics and helps you decide nothing, because it contains none of your numbers.

The twenty below are organized by routine, and each block says what it needs to work.

Before the prompts: what access changes

There are two ways to work with AI in paid media.

Without access. You paste numbers into the chat. It works for a one-off analysis and collapses as a routine, because every week you export, slice and re-explain the context.

With access. The assistant queries the account live, through the API, and answers with your numbers. That is what a connector does. The mechanism is in what is MCP.

The prompts below work better in the second mode. In the first, use them as a script for what to paste.

Monday routine: what changed

1. "Compare spend, results and cost per result for the last 7 days against the previous 7, by campaign. List only what moved more than 20%."
2. "Which campaigns spent the most last week and brought no results at all?"
3. "Show daily spend for the last 14 days and point out which day the curve shifted."
4. "Did any campaign go down or stop spending in the last 3 days?"
5. "List campaigns by cost per result, worst to best, with each one's spend next to it."

What they share: they all ask for a comparison, not a snapshot. A loose number informs nothing; a number against the previous period informs.

Diagnosis when cost rises

6. "Which campaign's cost per result went up? Show frequency, CPM and CTR for that campaign in the same period."
7. "Did any campaign's frequency pass 4 in the last 14 days?"
8. "Compare CTR across the ads in campaign X and list the worst three."
9. "Did CPM rise or did CTR fall? Show both side by side over the last 4 weeks."
10. "Which ads have been running more than 30 days without a creative swap?"

Number 9 is the most useful on the list and almost nobody asks it. Rising cost per result has two possible causes with opposite fixes: more expensive media, or a less interesting ad. The full reading is in high CPL in Meta Ads and low CTR in Meta Ads.

Creative: what is wearing out

11. "List active ads by spend and show CTR and frequency for each."
12. "Which creative had the best cost per result in the last 30 days?"
13. "Did any ad lose more than 30% of its CTR comparing the last two weeks?"
14. "Which ads take more than 10% of campaign spend while performing below the campaign average?"

Creative fatigue shows up in this order: CTR falls, frequency rises, cost per result rises. Anyone watching only the last column finds out two weeks late.

Budget and scale

15. "If I doubled campaign X's budget, what in the account tells me it can take it? Show frequency, reach and audience size."
16. "Which campaign has the best cost per result with spend above US$ 1,000 this month?"
17. "Break down this month's spend by campaign in percentages and tell me where it is concentrated."
18. "Compare this month against the same month last year, by spend and results."

Number 16 has a deliberate filter. An excellent cost per result on tiny spend proves nothing: it is a short sample. The scaling criteria are in when to scale a campaign.

Client reporting

19. "Build a monthly client summary: spend, results, cost per result and the change against last month, in plain language with no jargon."
20. "List the three things that changed most in the account this month and explain each in one sentence."

Prompt 20 saves the most agency time. It does not ask for the report, it asks for the reading of the report, which is the part no spreadsheet automates. The full format is in Meta Ads reporting.

How to rewrite a weak prompt

The difference between a prompt that returns text and one that returns a decision is almost always three things: period, breakdown and comparison.

Notice that no right-hand column asks for an opinion. It asks for a slice of your data, and the opinion comes afterwards, once you are looking at the right numbers. That inversion changes the quality of the entire conversation.

One detail that saves rework: ask for the unit alongside. "In dollars" and "as a percentage" prevent the answer that mixes both and forces you to ask again.

What to ask before a client call

A separate short list, because the goal is different: you are not optimizing, you are preparing to be questioned.

• "What is the single biggest change in this account this month, in one sentence?"
• "Which campaign will the client most likely ask about, and what is the honest answer?"
• "Where did we spend money without a result, and what did we do about it?"
• "What did we change in the last 30 days, and what happened after each change?"

The last one is the one that builds trust. Clients rarely object to a bad month; they object to a bad month nobody can explain. Having the list of changes with their consequences turns the call from a defense into a report. The structure for the call itself is in presenting results to a client.

Three mistakes that ruin any prompt

Asking for a recommendation before asking for a number. "What should I improve in this account?" returns the same generic list every time.

Not stating the period. Without a date range, the assistant picks one, and it changes between answers.

Accepting the first answer. The conversation is the format's advantage. Ask where the number came from, request the breakdown by campaign, challenge the conclusion.

For the full weekly routine, see Claude for paid ads and the document itself on the paid media reporting page.