The AI audit ambush

A client ambushed me with a 30-page AI audit.

I joined the Zoom meeting thinking it was to meet my client's new marketing manager. I didn't expect to spend the next hour justifying my position.

The new marketing manager had invited another agency to audit the account using their AI tool.

He scrolled slowly through the report.

It was damning.

30+ pages of dire warnings, critical fixes and missed opportunities. Every few pages there was another chart accusing me of wasted spend or missed revenue. The report was polished and confident.

I panicked. I heard questions nobody asked:

• "If AI found all these mistakes in just a few minutes, why didn't you?"
• "Why are we paying you when the agency is offering to do better at half the price?"

My first thought was to defend myself. To say something like:

I've made you millions over the last 10 years. I've kept the account profitable through a pandemic, two wars, a switch to HubSpot, a switch away from HubSpot and a parade of marketing managers each with a new idea. Don't show me this clanker slop!

I didn't say it. I did something much more stupid.

I started rebutting the report point by point. Trying to explain complex technical nuances in plain terms so the client could make sense of them. Trying to remember why we had done something years ago that the AI was now flagging as an error. Trying to justify deviations from what the AI said was best practice.

It was exhausting.

The new marketing manager pushed back on everything because he trusted the AI, not me. It took us 90 minutes to get through the first three points.

Then the client gave up. He said he'd send the report and asked me to prepare a written response.

A few days later I read the report. Some of the points were valid. Many weren't.

For example, it said our PMAX campaigns were inefficient. The cost per conversion was double that of our search campaigns. But it missed that the search campaigns were optimising for raw leads and PMAX was optimising for booked appointments. The PMAX campaigns were actually our cheapest source of booked appointment conversions.

Some of the issues were technically correct, but commercially irrelevant. For instance, it warned that brand keywords were leaking into non-brand campaigns. That's Google 101 and I'd already excluded brand keywords. The AI found one click on a brand typo I'd missed. (If the business was called Smith Inc, the offending search term was Smiht Inc.)

And other recommendations ignored the reality of the business. It criticised us for running ads only during working hours, but we'd already tested this. After-hours leads didn't convert into customers. It flagged us for pinning ad copy without knowing that we have to comply with the industry licensing requirements.

But in spite of this, it looked authoritative. The language was confident. The presentation was polished. And it had heft. It's easy to see why it might make a client question their current provider's abilities.

What I did in the meeting was wrong

I wrote up a response, sent it to the client and luckily that was the end of it.

But as I thought about it, I realised that I had handled the meeting completely wrong.

The real issue wasn't that the audit had mistakes. Some points were fair, and AI audits are going to get better.

The issue is that AI can generate criticism faster than humans can evaluate it or respond to it. No human can win a debate against a machine that can generate endless plausible-sounding opinions at the push of a button.

And that makes a point-by-point rebuttal a trap.

Next time this happens I'm not going to debate every recommendation. Instead I'm going to try to help my client understand four things:

1. A free audit is a sales tool.
2. The audit will raise some valid points.
3. A useful audit looks at the system, not just the ad account.
4. The audit evaluates against one way of managing ads. There are many ways.

Let's look at each in more detail.

A free audit is a sales tool

I'm not saying that a free audit is worthless, but it's unlikely to be neutral.

It's not hard to produce an audit that is true but misleading:

• Generate pages of recommendations.
• Give every recommendation the same weight.
• Ignore anything that's working.

Format it nicely and even the most loyal client might think their current ad manager is a cretin. Which is exactly the point of the sales audit.

The audit will raise some valid points

No account is perfect. If you dig deep enough, the way a computer can, you will find unfinished work:

• More negative keywords.
• An automated recommendation that's worth testing.
• Keywords that aren't getting many impressions in their current campaign.

This isn't negligence. It's the reality of managing Google Ads under time, budget and operational constraints.

An ad account doesn't have to tick every box to make money.

A useful audit looks at the system, not just the ad account

The next time, I'll start by gently asking the client:

• Did the AI connect to the CRM so it could see which leads turned into sales?
• Did the AI look at a snapshot of the account as it is today, or did it ask about what we've already tried?
• Did the AI ask any questions about what services you sell, which ones are easiest to sell, which ones are the most profitable over the long term?
• Did the AI ask about operational trade-offs we've had to make?

You see, Google Ads doesn't operate in a vacuum. It's only profitable when it's part of a functioning system.

A useful audit looks at that system:

• The landing pages.
• What makes a lead qualified.
• Which leads become profitable customers.
• What operational constraints the business is working under.
• What's already been tested.

Without context, an audit is mostly measuring how closely the account matches a theoretical model of perfection.

The audit evaluates against one way of managing ads. There are many ways

Every audit system is built around a set of assumptions about how Google Ads should be managed.

• Automation vs manual control.
• Granular structures vs campaign consolidation.
• Performance Max vs exact match search only.
• Maximise conversions vs target CPA.

That's unavoidable. The problem is when these assumptions are presented as the one true way.

Experienced PPC managers adjust how they manage ads to suit the business. That means some highly profitable accounts can look "wrong" if you don't understand the trade-offs behind the current structure.

AI audits are going to become normal

They're going to get faster, cheaper and more convincing. Clients are going to see more of them.

The danger isn't that AI can find problems in an ad account. It's that AI can generate endless criticism without understanding the business behind the account.

It doesn't know which leads became customers. It doesn't know what you've already tested. It doesn't know the operational constraints the business is working under. It doesn't know which trade-offs were intentional.

It just knows how to produce plausible-sounding recommendations at industrial scale.

That's why arguing point by point is usually a mistake.

The real value of an experienced PPC manager isn't the ability to generate more recommendations. It's the ability to decide:

• what matters
• what doesn't
• what can wait
• what creates operational burden
• what actually moves the business forward

That's the real work. And at least for now, that's still human.

Related articles

Why good Google Ads campaigns still lose clients

Most Google Ads Reports Focus on the Wrong Things

Helping people make better Google Ads decisions

When there are no good options in Google Ads

Why I've Stopped Giving Quick-Fix Google Ads Advice

50% more leads… and still fired

Topics you'll find on this site