Analyze Facebook Ads With AI: A Ready-Made Prompt for ChatGPT, Claude or Gemini

AI tools like ChatGPT, Claude and Gemini can act as a second pair of eyes on your campaign numbers and point out what you missed. But most people who try it get generic advice like "improve your targeting and test new creative". The problem isn't the tool — it's what you give it. Here's how to analyze Facebook ads with AI and get answers you can act on.

Why AI gives generic answers

Because it doesn't know your campaign. Type "my campaign is expensive, what should I do?" and all it can offer is generic advice. Give it the full numbers, the goal, the targeting, the period and what success means for your business, and you'll get analysis based on your actual situation.

Analyze Facebook ads with AI: the data to include

• Objective and result type: sales, messages, leads?
• Core numbers: spend, results, cost per result, revenue and ROAS if relevant.
• Delivery and engagement: reach, impressions, CPM, CTR, CPC.
• Targeting: countries, ages, gender, interests — or broad.
• The period and what changed during it.
• What success means: your maximum acceptable cost per result or break-even ROAS (How to Calculate ROAS and Break-Even ROAS: When Is a Facebook Campaign Really Profitable?).

A copy-paste prompt

Preparing the data from Ads Manager by hand

1. Pick the period: at least 7 days, ideally 14 or 30, to smooth out daily swings.
2. Pick your columns: under Columns, add spend, results, cost per result, reach, impressions, frequency, CPM, link CTR, CPC, and purchase value and ROAS for sales campaigns.
3. Go down to ad level: campaign totals hide the difference between a winning ad and one that's burning budget. Give the AI per-ad numbers where you can.
4. Add the comparison: the same numbers for the previous period, so the AI sees the trend, not just a snapshot.
5. Export or copy: export the table and paste its contents into the chat, or type the numbers in order as in the prompt above.

Follow-up questions that get deeper answers

The first answer is usually fairly general. The real value comes from follow-ups:

• "Which metric do you think moved first, and what does that suggest?"
• "If I could test only one thing this week, what would you recommend and why?"
• "Write five different video hooks for this product that would suit this audience."
• "What assumptions did your analysis rely on, and which might be wrong?"
• "Summarize this as a short, plain-language message I can send the client."

Reviewing the AI's answer before acting on it

Treat the answer like a suggestion from a smart new colleague who has never seen your account. Before acting, ask yourself:

• Is the conclusion based on a number that actually exists? AI sometimes cites figures or percentages you never gave it.
• Does it know the market context? It may not know it's peak season, that the client changed prices, or that the product sold out.
• Can you actually implement it? Some suggestions assume a bigger budget or audience than you have.
• Does it contradict what the account has taught you? Your knowledge of the client and market should outweigh any generic recommendation.

The best use of AI here isn't letting it decide for you. It's saving you drafting time and pointing out an angle you hadn't considered.

Mistakes that make the analysis useless

• Missing numbers: without CTR and CPM it can't tell an ad problem from a market problem.
• No success benchmark: $10 per order might be excellent or disastrous depending on your margin.
• Trusting everything it says: AI can be wrong or make wrong assumptions about your account. Treat it as a second opinion to check, not a decision.
• Sharing sensitive data: it doesn't need client names or personal data. Numbers and targeting are enough.

A ready prompt from your data in one click

Assembling all of this by hand for every campaign takes time. In Raqeeb Ads' Campaign Analyzer, a button packages the campaign's data into a ready-made analysis prompt: objective, spend, results, cost per result, revenue, ROAS, targeting (countries, ages, gender, interests), reach, impressions, CPM, CTR and CPC — plus instructions asking for a verdict, a scale/kill/optimize recommendation, a diagnosis and a three-step plan. Copy it into any tool: ChatGPT, Claude or Gemini.

Ahmed Abdelhamed

A media buyer running Meta campaigns for clients across Egypt and the Gulf, who built Raqeeb Ads to fix the account-monitoring problems he kept running into himself.