Rethinking Meta Ads AI: Best practices for better results

Meta CEO Mark Zuckerberg made a bold claim about the future of advertising in a recent interview (emphasis mine):

• “You’re a business, you come to us, you tell us what your objective is, you connect to your bank account, you don’t need any creative, you don’t need any targeting demographic, you don’t need any measurement, except to be able to read the results that we spit out. I think that’s going to be huge, I think it is a redefinition of the category of advertising.”

Zuckerberg paints an interesting future. But it’s not quite reality.

Meta Ads’ AI isn’t replacing digital marketers anytime soon. 

Still, its best practices – from automation to targeting – can help social ads teams drive better results and streamline workflows.

This article breaks down some of the most common Meta Ads best practices and pitfalls. 

Use goal-based campaigns

This is not exactly a best practice, but Meta has been pushing this into its campaign-building workflow pretty hard since the beginning. 

Meta’s focus on helping less experienced advertisers used to be a plus, but now it’s become central to how the platform works. 

But even so, you need to approach campaign goals with nuance.

Why is that? 

While AI now runs a large part of the Meta Ads show, it is still based on very literal algorithms. 

They will achieve your goal but never challenge it.

For example, if your goal is to generate qualified traffic, you could think, “Let’s use a Traffic campaign goal”. 

However, any experienced Meta Ads advertiser will tell you it generates garbage web visitors. 

Instead, you need to use the Conversions goal and manually set up a qualified visit custom conversion. This subtle difference will have a massive impact on your funnel.

What to do instead

You need to find a way to align your business goals with ad delivery. 

And that’s what Meta encourages advertisers to do by selecting campaign objectives like conversions, traffic, or reach. 

But don’t forget AI is not smart. You are.

Implement server-side tracking

Since iOS 14, this has been a staple in the industry. 

Meta’s been pushing for it with its Conversion API solution, and shady advertisers have been using it to sell additional solutions. 

The goal: record more conversions, bypassing privacy-first browsers and ad blockers.

You don’t always need server-side tracking. Here’s why.

First, in legal terms, you cannot ignore user consent. 

This means server-side tracking does not solve for iOS 14+ or GDPR, for example. 

The whole argument of recovering lost tracking capabilities is diminished at best (ad blockers are not that widespread).

Also, while there are natively integrated solutions for server-side tracking (think Shopify, which makes it a no-brainer to implement), more complex CMSs often mean much tougher, longer, and resource-intensive projects to get this up and running.

What to do instead

If you generate high volumes, implementing server-side tracking is probably a good idea. 

However, for SMBs with low conversion volumes, it’s most certainly not worth the effort.

Trust Advantage+ targeting

This is what Zuckerberg was alluding to when he said, “You don’t need any targeting demographic.” 

Advantage+ is Meta Ads’ product to entirely delegate targeting to the ad network.

The upside? 

It is highly capable of finding users whose chances of performing your target goal are higher.

By entering you into broader auctions, it delivers cheaper CPMs. 

The downsides? 

The wrong campaign and ad set-level goal will:

• Crash your results. 
• Remove control. (For example, you cannot exclude retargeting audiences.)

What to do instead 

Use “manual” broad targeting. It offers the same capabilities, plus control over ad saturation.

---

Serve ads on all placements

Meta Ads wants you to show ads across all of their platforms and trust them that all of their ad inventory is worth it. 

Facebook, Instagram, Messenger, and the Audience Network all deliver performance equally, they say. 

Again, the upside is better quantitative results: You can enter cheaper auctions for the exact same users, simply in a different setting.

The downside: the Audience Network often results in spammy placements and low-quality traffic, especially in prospecting campaigns.

What to do instead

Use auto placements but exclude the Audience Network in prospecting campaigns to avoid junk traffic.

Enhance creatives with AI

Again, we come back to Zuckerberg’s quote, “You don’t need any creative.” 

But is that true? I disagree.

Meta Ads’ creative enhancement tool promises better performance by adding effects, cropping, or animating assets, and perhaps soon, by generating full creatives from scratch.

That’s a useful fallback if you lack creative resources, but it still comes with real trade-offs.

In practice, it often clashes with brand guidelines, introduces off-brand edits, or changes layout unpredictably. 

For advertisers with established branding, it’s still too risky. 

No doubt, it will improve in the near future. 

But when everyone leverages AI to produce creatives, the differentiator will be humans. 

Either because they will better drive the AI, have more creative ideas, or have a stronger overall brand.

For example, one of my go-to moves in a lead generation setting is listening to sales calls. 

They always have those performance nuggets (think pain points, sales pitches, etc.), which we re-use at the creative and landing page levels. 

Will Meta Ads’ AI know that? Not a chance – at least, not yet.

What to do instead

Keep this off for now unless you have zero creative resources and brand standards.

Don’t edit budgets more than 20% at a time

I still hear this one fairly often. 

They say that increasing or decreasing your budget by more than 20% will cause the algorithm to go crazy, and performance will tank. 

This is only partially true. 

The real “rule” should be to generate enough results for the algorithm to perform steadily. 

Meta Ads used to advise for 50 results per week per ad set. 

From my experience, you can lower that figure, especially if it makes sense in terms of the customer journey (e.g., splitting prospecting and retargeting).

Indeed, Meta Ads’ algorithms are based on machine learning. 

If they can learn quickly (by having enough conversions frequently), they will adapt seamlessly. 

Conversely, if you don’t feed them enough conversions, their experiments will last longer, making it look like the algorithms are going nuts.

What to do instead

Monitor conversion volumes and forecast budget edit impacts accordingly. And if moving those by 30% makes sense, go for it.

Get out of the learning phase, ASAP

Meta Ads performs best once it has enough data. But is it that big of a deal?

Many businesses won’t hit 50 conversions per week per ad set, especially in niche or high-ticket industries. That’s not a reason to panic or rebuild.

Exiting the Learning Phase is basically Meta Ads’ way of telling you to either increase the budget or give them more control by merging ad sets. But is it in your best interest? Unfortunately, it often isn’t.

What to do instead

Optimize toward getting out of the Learning Phase, but don’t chase it blindly.

Sometimes “learning limited” is just your reality, and that’s fine: You have more structural steps to take.

Be critical and balance best practices

It’s easy to fall into the trap of accepting Meta Ads recommendations as gospel: it’s reassuring and cost-efficient.

What’s not to like?

Performance often improves when you challenge the defaults and infuse business and ground-level insights into your Meta Ads campaigns. 

That applies to targeting, creatives, and measurement. (Sorry, Mark!)

Like with Google Ads, success comes from knowing when to let AI drive – and when to take the wheel yourself.