How to Audit Meta Ads With Claude + Meta Ads MCP in 2026
How to Audit Meta Ads With Claude + Meta Ads MCP in 2026
I gave Claude access to my Meta Ads account.
Then while I had my morning coffee, it pulled a full account audit in minutes.
No manually exporting spreadsheets.
No copying campaign numbers one by one.
No generic prompt asking AI what a good CTR should be.
Claude was reading the real data from the Meta Ads account using Meta's official MCP connection.
The interesting part was not just how quickly it pulled the numbers.
It was what Claude found.
It identified my winning creative, campaigns wasting money, major differences between Facebook and Instagram, problems with Audience Network, an issue with my retargeting setup, and the point in my funnel where I was losing money.
Then I used follow-up prompts to turn the data into charts, country breakdowns, conversion reports, and next actions.
In this article, I will show you the workflow I used and how you can apply the same Meta Ads audit process to your own account.
Watch the full video here:
https://www.youtube.com/watch?v=aKpR42Nfogs
What Is Meta Ads MCP?
MCP stands for Model Context Protocol.
You do not need to care too much about the technical name.
The useful part is that it allows AI tools such as Claude to connect to your Meta Ads account and work with the actual account data.
That is a big difference from the way most people use AI for advertising.
If you manually tell Claude:
"My CTR is 1.8%, my CPM is $20, and my CPA is $50. What should I do?"
Claude only knows those three numbers.
It cannot see what is happening across your campaigns, ads, countries, placements, audiences, and funnel.
With the Meta Ads MCP connected, the AI can work with much more of the actual context.
If you have not connected Claude to Meta Ads yet, I already have a complete step-by-step guide here:
https://www.wupscale.com/post/how-to-connect-meta-ads-to-claude-2026
Get the connection working first.
Then you can start doing the interesting part.
The Meta Ads Audit Prompt I Used
I wanted Claude to act more like a senior media buyer than a generic AI assistant.
So instead of asking:
"How are my Meta Ads doing?"
I made the job very specific.
My audit focuses on four main questions:
What is working?
What is wasting spend?
Where is the funnel breaking?
What should I test next?
That sounds simple, but it gives the audit a clear structure.
1. What Is Working?
First, I want Claude to identify the best campaigns and ads.
That includes looking at:
Spend
Results
Cost per result
CPM
CTR
CPC
Creative performance
Conversion volume
I also want it to identify patterns.
If one creative is outperforming four other ads, what is different about it?
Is it the design?
The hook?
The format?
The placement?
The audience?
The goal is not just to get a list of winners.
I want to understand why they are winning.
Finding the Winning Creative
In the account I audited, I was running a creative test.
The ads were using the same copy because I wanted to isolate one variable.
The variable was the creative.
That makes the test easier to understand.
If five ads use the same targeting, same copy, and same setup but one image clearly outperforms the others, I have much more confidence that the creative itself caused the difference.
Claude spotted the strongest creative and compared it against the other ads based on cost per booked demo, CPM, and overall performance.
That gives me something useful to do next.
I can take the winning concept and create more iterations of it.
This fits directly into the testing system I use across Meta Ads accounts.
You can see the full test-to-scale campaign structure here:
https://www.wupscale.com/post/meta-ads-campaign-structure-2026
The important part is always the same.
Test one variable.
Find the winner.
Iterate.
Then scale what already has evidence behind it.
2. What Is Wasting Spend?
This is where an AI Meta Ads audit gets more useful.
I do not want Claude to just tell me that my account generated X number of conversions.
I want it to find where money is being wasted.
In this audit, one of the biggest differences came from platform performance.
Instagram was generating booked demos at roughly $300 each.
Facebook was around $117.
That is a huge difference.
If I only look at the overall campaign numbers, I might miss it.
The total CPA could look acceptable while one platform quietly takes a large percentage of the spend at a much worse cost.
Now I have a clear hypothesis for my next test.
Do I need Instagram?
Should I test Facebook-only placements?
Is there a creative problem specifically on Instagram?
Those are much better questions than simply asking whether the campaign is profitable.
Why Meta Ads Breakdowns Matter
Your overall account averages can hide a lot.
A campaign might have a solid CTR overall.
But then you break the results down and discover Audience Network has an extremely high CTR while generating low-quality traffic.
Or Facebook may be producing most of your conversions while Instagram is absorbing budget at a much higher CPA.
Or one country may appear extremely cheap because the conversion tracking is reporting low-quality leads as successful results.
This is why I regularly use breakdowns.
I have a separate guide on using Meta Ads breakdowns to find wasted spend here:
https://www.wupscale.com/post/meta-ads-breakdowns-explained
The dashboard gives you the averages.
The breakdowns tell you what is actually creating those averages.
Claude Found an Audience Network Problem
Another interesting issue Claude found was Audience Network.
At first glance, the click-through rate looked strong.
That sounds positive.
But once we looked deeper into the funnel, those clicks were not turning into landing page views at the same rate.
That means the high CTR was misleading.
Cheap clicks do not matter if the traffic disappears before it reaches the page or never converts.
This is exactly the type of thing AI needs to be instructed to look for.
If you only ask:
"Which placement has the best CTR?"
Claude could easily tell you Audience Network.
Technically, that may be correct.
Strategically, it could be completely wrong.
The real question is:
Which placement produces meaningful traffic and conversions at an acceptable cost?
3. Where Is the Funnel Breaking?
This was the next part of the audit.
Instead of stopping at ad metrics, I want to follow the customer through the funnel.
That means looking at:
Impressions → Clicks → Landing Page Views → Leads or Demos → Purchases
You want to know where the biggest drop happens.
In this account, impression to click performance looked fine.
The CPM and CTR were reasonable.
Then clicks to landing page views started showing problems, partly because of Audience Network.
Further down the funnel, the page-to-demo conversion was another area where money was being lost.
Now the optimization plan becomes clearer.
A low CTR is probably an ad problem.
A strong CTR but poor landing page view rate might be a placement or website problem.
Strong landing page views but weak conversion rates might be an offer, messaging, or funnel problem.
You need to diagnose the problem before trying to fix it.
That is also the framework I use on my AI Meta Ads Analysis page:
https://www.wupscale.com/ai-meta-ads-analysis
Do not jump straight from data to recommendations.
Diagnose first.
Then decide what action makes sense.
Claude Also Caught a Retargeting Problem
One of the more useful findings was around my retargeting campaign.
It was labelled as retargeting.
But it was not really functioning as pure retargeting.
The audience was too small, so Advantage+ expansion was helping Meta reach beyond the original audience.
Claude called this out.
That matters because otherwise I might look at the campaign performance and assume:
"Retargeting is performing like this."
But I would not actually be measuring true retargeting.
This is another example of why campaign names alone are not enough.
The AI needs to examine how the campaign is actually configured and interpret the numbers in context.
4. What Should I Test Next?
The final part of my audit asks Claude for the next three tests in priority order.
Not 25 recommendations.
Three.
This forces the audit to become actionable.
Based on this account, the next actions included things like:
Testing placements
Rebuilding the retargeting setup
Testing more image creatives
Now I have an actual plan.
That is what an audit should produce.
More dashboards do not automatically improve your advertising.
Better decisions do.
The Exact Structure of My Meta Ads Audit Prompt
The core prompt is basically:
Act as a senior media buyer.
Use the real data from the Meta Ads account, not generic best practices.
Analyze the last 30 days.
Tell me:
What is working, including the best campaigns, ads, and creative patterns.
What is wasting spend at the campaign, ad set, and ad level.
Where the funnel is breaking.
The next three tests I should run in priority order.
If something is broken, say so.
That last part matters.
I do not want Claude trying to make everything sound positive.
If the retargeting setup is bad, tell me.
If a placement is wasting money, tell me.
If an ad needs to be paused, tell me.
The goal is not to get a nice report.
The goal is to make the account better.
Use Follow-Up Prompts to Go Deeper
The first audit is only the start.
Once Claude understands the account, you can keep asking follow-up questions.
For example:
"Give me the performance breakdown by country."
"Compare cost per demo by country."
"Show spend versus conversions as a chart."
"Break performance down by age and gender."
"Compare the last seven days to the previous seven days."
"Turn this into a weekly client report."
Now the same Meta Ads data can become much more useful.
Turn Meta Ads Data Into Charts
This is one reason I like Claude for this workflow.
The output can be very clean when you want to visualize results.
In my example, I asked it to illustrate:
Spend versus demos by country
Cost per demo by country
Conversion rates
Country performance
I could switch between charts and tables depending on how I wanted to view or present the information.
This is especially useful if you manage client accounts.
Instead of opening Ads Manager during a client call and jumping between 10 different breakdown menus, you can create a cleaner report showing the numbers that actually matter.
You Can Use This for Weekly Client Reporting
This workflow is not limited to one-off audits.
You could ask Claude to compare:
This week versus last week.
This month versus last month.
Current campaigns versus the previous campaign structure.
Country performance over time.
Creative performance over time.
Then summarize:
What improved
What declined
What was changed
What should happen next
That gives you a much more useful client report than dumping screenshots from Ads Manager into a presentation.
Automating the Reporting Workflow
You can take this even further.
Instead of manually requesting the same report every Monday, AI workflows can help turn recurring account analysis into a repeatable reporting system.
For example, your weekly report could include:
Total spend
Conversions
CPA
ROAS
Best ads
Worst ads
Country breakdown
Placement breakdown
Funnel issues
Next actions
The goal is not automation for the sake of automation.
The goal is to stop spending your time repeatedly pulling the same numbers.
Use that time to make decisions.
AI Can Still Give You Bad Advice
There is an important warning here.
AI is only as good as the data it receives.
Imagine your Meta Ads tracking is double-counting purchases.
Your real ROAS is 2x.
Ads Manager reports 4x.
Claude sees the 4x ROAS and recommends scaling aggressively.
Claude did not necessarily analyze the data badly.
The underlying data was wrong.
The same problem can happen with:
Duplicate purchases
Missing purchases
Incorrect revenue
Bad lead quality
Wrong conversion events
Attribution problems
Before you rely heavily on an AI Meta Ads audit, make sure your tracking is actually working.
AI can help you read the account.
It cannot magically fix bad source data.
AI Should Help You Think, Not Replace You
I would not blindly follow every recommendation Claude gives me.
I still want to make the final decision.
AI is extremely useful for:
Finding patterns.
Comparing large amounts of data.
Spotting unusual performance.
Creating breakdowns.
Producing reports.
Prioritizing tests.
But I still know the context behind the business.
Maybe a country generates cheap leads but terrible customers.
Maybe a campaign is intentionally spending more because I am collecting data.
Maybe an ad looks weak over seven days but has historically been one of the most consistent performers.
The best workflow is not:
AI makes every decision.
It is:
AI processes the account faster, highlights what matters, and gives me better information to make the decision.
Build a Better Meta Ads System
This type of AI audit becomes much more powerful when the rest of your account is structured properly.
If your account has 30 random campaigns, inconsistent naming, broken tracking, and no clear testing process, Claude still has to make sense of chaos.
A clean testing and scaling system makes both human analysis and AI analysis easier.
You can learn more about the Meta Ads systems I use here:
https://www.wupscale.com/meta-ads-systems
And you can see real campaign examples and results here:
https://www.wupscale.com/results
Watch the Full Claude Meta Ads Audit
In the full video, I show the actual audit, the account findings, my exact prompt, funnel analysis, country breakdowns, charts, and follow-up prompts.
Watch it here:
https://www.youtube.com/watch?v=aKpR42Nfogs
Try the audit on your own account.
Do not just ask AI whether your ads are doing well.
Ask:
What is working?
What is wasting money?
Where does the funnel break?
What should I test next?
Those four questions will give you a much more useful starting point.
Want More Meta Ads Prompts and AI Workflows?
Inside Meta Ads Systems, I share the systems I use for campaign analysis, creative testing, scaling, tracking, Meta Ads MCP, AI workflows, and account optimization.
If you want to apply these workflows to your own campaigns, join here:
https://www.skool.com/advertising
Use AI to spend less time pulling reports and more time making better decisions.
