Claude Meta Ads Audit Prompt: Find Tracking Issues & Wasted Spend

Claude Meta Ads Audit Prompt: Find Tracking Issues & Wasted Spend

November 20, 2026•20 min read

Claude Meta Ads Audit Prompt: Find Tracking Issues & Wasted Spend

I'm feeling super tired today, so I just asked Claude AI to audit a Meta Ads account for me.

And the results were pretty interesting.

Instead of manually going through all my campaigns, ads, performance metrics, and reporting breakdowns, I connected my Meta Ads account to Claude using Meta Ads MCP.

Then I gave it one prompt and my ad account ID.

Claude pulled the available data and gave me a structured account audit.

It flagged a potential problem with my purchase value tracking.

It identified one winning ad and a couple of ads that might be wasting budget.

It also analyzed where people were dropping off in my funnel and suggested three tests I could run next.

That's the kind of AI workflow I actually find useful for media buying.

Not asking ChatGPT or Claude for generic advertising tips.

I want AI to analyze my real account data and help me figure out what to do next.

In this article, I'll show you how I approach a Claude AI Meta Ads audit, what to include in the prompt, and why accurate tracking is so important before trusting the recommendations.

You can also watch the full walkthrough here:

Watch: How to Audit Meta Ads With Claude AI

What Is a Claude AI Meta Ads Audit?

A Meta Ads audit is basically a review of your advertising account to understand what's working, what's not working, and what you should improve.

Normally, I would open Meta Ads Manager and start reviewing the important numbers.

Things like:

  • Ad spend and campaign results

  • Purchases or leads

  • Cost per acquisition

  • ROAS

  • CTR and CPC

  • CPM

  • Creative performance

  • Landing page performance

  • Audience and placement breakdowns

Then I'd compare the campaigns, identify the strongest ads, and look for problems.

If you're working across several accounts, this can take quite a bit of time.

And if you're an agency, you might also need to prepare a report for the client.

Now, with Claude and Meta Ads MCP, you can automate part of this analysis.

The AI can help retrieve the account data, organize the results, and highlight potential issues.

For me, the biggest benefit isn't just saving time.

It's getting a clearer overview of what needs attention in the account.

But there's an important condition.

Your data needs to be reliable.

If the tracking is wrong, AI might analyze the wrong numbers and recommend the wrong changes.

I'll explain that in more detail later.

How Claude Connects to Meta Ads With MCP

The connection I use in the video is Meta Ads MCP.

MCP stands for Model Context Protocol.

Basically, it allows supported AI tools to connect to external systems and work with their data.

In this case, Claude can use a Meta Ads MCP connection to access information from an advertising account, depending on the connection's permissions and available tools.

That means you don't necessarily have to export reports and manually upload CSV files every time you want to analyze an account.

You can ask the AI to retrieve the information directly.

In my example, I connected Claude to Meta Ads, provided the account ID, and requested an audit.

The Meta MCP server is available without a separate server fee in the setup I demonstrate. However, Claude or another AI provider may have its own subscription or usage costs.

I mention a roughly $20 Claude subscription in the video, but you should check the current pricing and plan requirements rather than assuming that price will always apply.

If you haven't connected your account yet, I already have a step-by-step guide:

How to Connect Meta Ads to Claude Using MCP

Get the connection working first.

Then you can start analyzing your campaigns.

What Claude Found in My Meta Ads Account

Let me show you the main findings from the account I audited.

This was an account I was using to advertise my Skool community, with different offers and plans.

I have multiple ads and different naming conventions, so I wanted Claude to organize the data and highlight the important parts.

The audit identified four main areas.

1. A Potential Purchase Value Tracking Problem

This was one of the first issues Claude flagged.

It suggested that my purchase value tracking wasn't correct.

That's important because purchase value directly affects your reported revenue and ROAS.

For example, let's say your advertising account reports 10x ROAS.

That looks amazing.

But what if the purchase values are being recorded incorrectly?

Maybe the account is reporting the wrong subscription value.

Maybe certain events are duplicated.

Or maybe the tracking setup doesn't match the actual customer payment.

Those are possible tracking problems you would need to investigate.

Claude flagged an issue in my account, but the warning alone doesn't prove exactly what caused it.

I would still check Events Manager, the Pixel, and the actual purchase data before deciding what needs to be fixed.

The main point is that I want to verify tracking before making scaling decisions.

2. One Winning Ad

The next part of the audit was creative performance.

Claude identified an ad that stood out from the others.

That's useful because I don't want to treat all creatives equally.

If one ad is generating stronger results, I want to understand why.

Maybe the image is working.

Maybe the hook is stronger.

Maybe the offer is clearer.

Or maybe it's reaching an audience that responds particularly well to that message.

Instead of just looking at the winning ad and moving on, I want to use it as a starting point for the next creative test.

In the video, Claude suggested developing another concept similar to one of my image creatives.

That's a practical recommendation.

I'm not creating something completely random.

I'm using the information from an existing ad to decide what to test next.

3. Ads That Might Be Wasting Spend

The audit also identified a couple of ads that appeared to be wasting budget.

This is something I always want to check when auditing Meta Ads.

Which campaigns and creatives are spending money without producing enough results?

For example, an ad might have:

  • Consistent spend but very few conversions

  • A CPA above the account target

  • A weak CTR compared with other creatives

  • Traffic that doesn't convert after reaching the landing page

But I wouldn't automatically pause every ad Claude identifies.

First, I want to know how much the ad has spent.

Has it collected enough data?

Is it a new creative test?

Is it part of a prospecting campaign that's intentionally testing new audiences?

Maybe the ad looks weak over the last few days but has historically been a consistent performer.

You need context.

AI can highlight the ads worth reviewing, but I still want to make the final decision.

4. Funnel Drop-Offs

This was another useful part of the report.

Claude looked at where people were dropping off in the advertising funnel.

For example, you might have a campaign generating a lot of link clicks.

But what happens after the click?

Are people actually loading the landing page?

Are they visiting the offer page?

Are they completing the signup or purchase?

A good CTR doesn't automatically mean your funnel is working.

You need to understand what happens after someone interacts with the ad.

And that's exactly what I want the AI audit to help me investigate.

The Claude AI Meta Ads Audit Prompt

In the video, I show the prompt I used to generate the audit.

The main idea is to make Claude analyze the account like a senior media buyer.

I don't want it to simply tell me whether the account is doing well.

I want specific findings from the available data.

The prompt should answer four questions:

  1. What's working?

  2. What's wasting spend?

  3. Where is the funnel breaking?

  4. What three tests should I run next?

Here's a reusable version based on the workflow shown in the video.

Claude AI Meta Ads Audit Prompt

Analyze my Meta Ads account for the last 30 days like a senior media buyer.

Use the connected Meta Ads account data and identify:

  1. The best-performing campaigns and ads, including their key performance metrics and likely creative strengths.

  2. Ads or campaigns that may be wasting spend, with the data supporting your assessment.

  3. Potential funnel problems, especially the relationship between impressions, link clicks, landing page views, and conversions.

  4. Tracking or purchase value inconsistencies that should be investigated.

  5. Three practical tests I should consider next, based on the current results.

Review country, placement, age, and gender breakdowns where available.

Include spend, CTR, CPC, CPA, purchases, and ROAS when the data supports those metrics.

Separate verified findings from possible explanations.

Be direct. Don't give generic advice, and don't invent missing data.

This is an expanded, copyable version of the request described in the video, not a verbatim transcription of the prompt displayed on screen.

You can adjust it depending on your business.

For example, if you're running ecommerce ads, you might want to add your break-even ROAS and average order value.

If you're generating leads, you'd probably want Claude to focus more on cost per qualified lead and the conversion rate from lead to customer.

And if you're advertising a Skool community like I am, make sure the AI understands your different plans, trials, and subscription values.

Otherwise, it might misunderstand what a reported conversion is actually worth.

Step 1: Analyze What's Working

The first part of the audit should identify the strongest campaigns and ads.

This might sound obvious, but I want to go beyond simply finding the ad with the lowest CPA.

Maybe one ad has generated two purchases at a very low cost.

Another ad has generated 30 purchases at a slightly higher cost.

Depending on the budget, conversion volume, and business goals, the second ad might be much more useful.

That's why I want Claude to review spend and results together.

I also want it to explain what might be working in the creative strategy.

If the AI identifies a strong image ad, I might create another version of that concept.

If a certain message is performing well, I might test it with different visuals.

The goal is to understand what deserves another test.

And if you're already using a testing and scaling system, this information can become part of your weekly creative workflow.

I explain how I organize those campaigns in my Meta Ads campaign structure guide.

Step 2: Identify Ads That Are Wasting Budget

Next, I want Claude to find inefficient spending.

This is one of the most practical parts of an AI Meta Ads audit.

Instead of going through every ad manually, I can ask Claude to identify the ones that need attention.

But the report needs to explain why an ad looks inefficient.

For example, I want to know:

How much has it spent?

How many purchases or leads has it generated?

How does its CPA compare with the target?

Is the problem getting clicks, or is it converting those clicks?

And is there enough data to make a decision?

I don't want Claude to say, "Pause this ad," without explaining the numbers behind that recommendation.

Especially if I'm running creative testing campaigns.

Sometimes an ad needs more time or budget to produce meaningful results.

Other times, the numbers clearly justify reviewing whether to pause it.

The important thing is to understand the difference.

Step 3: Find Where Your Meta Ads Funnel Is Breaking

Now let's talk about funnel analysis.

This is something I mention in the video because it can completely change what you should optimize.

A Meta Ads funnel might look like this:

Impressions → Link Clicks → Landing Page Views → Purchases

If the campaign isn't generating enough purchases, you need to understand where people are dropping off.

Let's look at a few possibilities.

Problem 1: Low Click-Through Rate

If people see your ads but don't click, the issue might be the creative or message.

Maybe the hook isn't strong enough.

Maybe the offer isn't clear.

Or maybe the ad is reaching the wrong audience.

That's where I'd review the creatives and targeting.

Problem 2: Lots of Clicks, Few Landing Page Views

This is one of the most important metrics I like checking.

Let's say your campaign generates 1,000 link clicks.

But you're only getting 550 landing page views.

That's a significant difference.

Some of that gap might be related to tracking or differences in how the metrics are counted.

But it could also indicate problems with page loading, website speed, redirects, or traffic quality.

I want to investigate that before spending more money.

In the video, I mention that I use a custom metric to track this relationship.

A simple way to look at it is:

Landing Page View Rate = Landing Page Views ÷ Link Clicks × 100

This isn't a complete explanation of every tracking difference, but it's a useful starting point.

Problem 3: Landing Page Views but Few Conversions

What if people are reaching the page but not buying?

Then I'd start looking at the next part of the funnel.

Maybe the offer isn't convincing.

Maybe the checkout process is too complicated.

Maybe the page doesn't match what the advertisement promised.

Or maybe conversion events aren't being recorded correctly.

These are different problems.

And they need different solutions.

That's why I want Claude to diagnose the funnel instead of just recommending that I change my ads whenever performance drops.

For more context on how I use AI to interpret campaign data, see my AI Meta Ads Analysis guide.

Step 4: Check Your Pixel and Purchase Value Tracking

Before trusting the audit, I'd check the tracking.

This is probably the most important warning in the video.

Let's say you're running an ecommerce campaign.

Meta Ads Manager reports 10x ROAS.

Claude analyzes the account and recommends increasing the budget because the return looks amazing.

But then you discover the purchase values were wrong.

Maybe the reported revenue doesn't match your actual ecommerce orders.

Now the scaling recommendation is based on incorrect information.

That's why I don't want AI to blindly trust every reported metric.

I want it to flag possible tracking problems.

Then I want to verify them.

Depending on the account, that could include reviewing Meta Events Manager, the Pixel, Conversions API events, purchase values, event deduplication, and the relationship between platform-reported results and actual sales.

The AI may be able to access some of that information through the connected tools.

But if the necessary data isn't available, it should say so.

An MCP connection doesn't automatically guarantee a complete tracking audit.

And a clean-looking AI report doesn't mean your data is correct.

If the tracking is wrong, the recommendations can be wrong too.

That's why I would fix serious measurement problems before making major scaling decisions based on the AI audit.

If you're working with Skool community purchases, my Meta Pixel tracking guide for Skool groups covers another relevant part of this process.

Step 5: Ask Claude for Three Tests to Run Next

This is where the audit becomes more useful than a normal performance report.

After reviewing the data, I want Claude to recommend what I should test next.

In the video, it suggested three directions.

Test 1: Audience Testing

The first recommendation was to test another audience.

That's something I'd consider after reviewing where the current conversions are coming from.

But I wouldn't just create random audiences.

I'd use the existing campaign data to identify a reasonable hypothesis.

Maybe there's an audience segment that's showing promising results.

Maybe a different market deserves its own test.

The idea is to use the findings to plan a controlled experiment.

Test 2: A New Creative Based on an Existing Winner

The second recommendation was to create a new concept based on one of my existing image creatives.

I like this approach.

If an ad is already performing well, I want to understand which parts of the concept might be worth repeating.

Maybe it's the angle.

Maybe it's the copy.

Maybe it's the visual format.

Then I can create another version and test it.

I'm not assuming the next variation will automatically win.

But at least I'm starting from something that has already shown some promise.

Test 3: Retargeting Creatives

The third recommendation was around retargeting.

In this example, Claude suggested using proof-based messaging connected to my experience managing over $5 million in Meta Ads spend.

For a warm audience, that's a reasonable creative angle to investigate.

These people may already know the brand.

Maybe they've visited the website.

Maybe they've seen my content.

So instead of repeating the same cold-audience introduction, I could test a creative that provides more context or proof.

For example, a founder-led video or a direct explanation of what someone gets inside the community.

If you want to explore the setup side of this, I also have a Meta Ads retargeting campaign tutorial.

Step 6: Review Country, Placement, Age, and Gender Breakdowns

There's one more thing I always want to check.

Breakdowns.

Your main campaign metrics don't tell you everything.

Let's say Claude finds an ad with a strong CTR.

That's useful.

But where are those clicks coming from?

Facebook?

Instagram?

Audience Network?

Which countries are getting the results?

And which age groups are converting?

This information matters because campaign averages can hide important differences.

A high CTR doesn't necessarily mean the traffic is qualified.

A low CPA from one country doesn't automatically mean that country will produce valuable customers.

And one promising audience segment doesn't mean you should exclude everyone else immediately.

So I want Claude to review the available breakdowns and explain which patterns are worth investigating.

I cover this in more detail in my Meta Ads Breakdowns guide.

For this audit, I would specifically look at country, placement, age, and gender results, with enough conversion data to support any conclusions.

Using Claude Charts for Client Audits

Another thing I mention in the video is visualization.

Claude can help organize the account results into charts and other visual reports, depending on the tools and features available.

I find this particularly useful when preparing an audit for a client or potential customer.

Instead of going into a call with a large spreadsheet or dozens of Ads Manager screenshots, I can present a clearer summary.

For example, I might want a chart showing which campaigns are generating most of the conversions.

Or a funnel visualization showing the difference between link clicks and landing page views.

Or a table comparing the strongest and weakest creatives.

For the workflow demonstrated in the video, I liked the way Claude presented the results.

But the visualization itself isn't what makes the audit useful.

The numbers still need to be correct.

And the recommendations still need to make sense.

A beautiful chart built on inaccurate purchase value tracking is still a bad report.

Can Claude Replace a Meta Ads Specialist?

I wouldn't look at it that way.

Claude is useful because it can reduce the repetitive work involved in account analysis.

Instead of manually going through every metric, I can ask it to identify patterns and potential problems.

But it doesn't automatically understand the full business.

It may not know the profit margin.

It may not understand the customer's lifetime value.

It may not know that a campaign is intentionally collecting data.

And it may not recognize that the conversion tracking is wrong unless it has enough information to spot the inconsistency.

Those things matter.

So I still want to review the recommendations myself.

The best workflow for me is simple.

Let AI process the account data.

Use it to highlight possible issues.

Then make decisions based on the actual business context.

That's how I use AI to support my media buying, not replace my judgment.

My Checklist Before Trusting an AI Meta Ads Audit

Before making changes based on the report, here's what I would check:

  • Is Claude connected to the correct Meta Ads account?

  • Am I looking at the correct reporting period?

  • Is the Pixel and purchase value tracking reliable?

  • Does Claude understand the campaign objective and offer?

  • Are the winning and underperforming ads supported by enough data?

  • Has the audit reviewed funnel performance?

  • Has it checked relevant country, placement, age, and gender breakdowns?

  • Are the three proposed tests specific and measurable?

  • Have I verified the recommendations before changing budgets?

If the answer to some of these questions is no, I'd investigate that first.

You don't need an AI report with 50 different recommendations.

You need a smaller number of useful findings that help you make better decisions.

Frequently Asked Questions

Can Claude AI audit a Meta Ads account?

Yes, with access to the relevant advertising data.

A supported MCP connection can let Claude retrieve campaign performance information and analyze the results.

The exact capabilities depend on the connection, permissions, and available tools.

Is Meta Ads MCP free?

The Meta MCP server used in the video does not have a separate server fee, but the AI tool you connect may have subscription or usage costs.

Check the current requirements before setting it up.

Can Claude identify wasted Meta Ads spend?

Claude can help flag ads or campaigns that appear inefficient based on reported spend and results.

But you should verify the findings before pausing ads, especially when the account has limited conversion data.

Can Claude check Meta Pixel tracking?

It can help analyze available tracking information and flag inconsistencies.

However, access to a Meta Ads account doesn't automatically mean every Events Manager diagnostic or Pixel configuration is available through the connection.

You may still need to verify issues manually.

What should I include in a Meta Ads audit prompt?

I would ask for the strongest campaigns, ads wasting spend, funnel drop-offs, tracking warnings, and three practical tests to consider next.

I'd also tell the AI to use the actual account data and avoid inventing missing information.

Is Claude better than ChatGPT for Meta Ads audits?

Both can be useful.

In the workflow shown in the video, I liked Claude's reporting and visualization capabilities, especially for presenting results on client calls.

But I wouldn't choose a tool purely based on how the report looks.

Data access, reliability, and the quality of the analysis are more important.

Final Thoughts: Use Claude to Find What Actually Needs Fixing

There are a lot of interesting AI tools for advertising right now.

But for me, one of the most useful workflows is still a simple account audit.

Connect Claude to Meta Ads using MCP.

Give it the account ID and a clear prompt.

Ask it what's working, what's wasting spend, where the funnel is breaking, and what you should test next.

Then review the findings.

In my example, Claude flagged a potential purchase value tracking issue, identified a winning creative, highlighted possible wasted spend, and suggested three tests.

That saved me from having to start the analysis completely manually.

But I still need to verify the data.

Especially the tracking.

Because if your purchase values or ROAS are incorrect, a fast audit can still lead to bad decisions.

So use AI to make the reporting process easier.

Just don't blindly follow every recommendation.

If you want to see the actual Claude audit, the prompt, and the results from my account, watch the video here:

Watch the Full Claude Meta Ads Audit Tutorial

And if you want a broader guide to Meta Ads MCP audits and reporting, you can also read my Claude Meta Ads Audit guide.

Want More Meta Ads and AI Workflows?

Inside Meta Ads Systems by Wupscale, I share practical tutorials on Meta Ads, AI workflows, campaign optimization, creative testing, reporting, and tracking.

The goal is to help you understand the data and make better decisions inside your advertising accounts.

If you want to learn these workflows and get more practical Meta Ads training, join my Skool community:

https://www.skool.com/advertising

And start by watching the full video above to see the Claude AI audit in action.

I'll see you inside!

Máté

Máté Hunyor
Máté Hunyor is the founder of Wupscale, a Meta Ads-focused business built around AI systems.
Back to Blog