A Founder's Guide to Understanding AI in Meta Ads

Over the last few years you'll have noticed more and more AI features creeping into Meta's Ads Manager. And as a founder, you may have questioned some of the developments…

"I know my business better than Meta does. I'm increasingly uncomfortable spending more money on a system I can't influence."

"Am I paying an agency to oversee a system which basically does what it wants, regardless of what we/they ask it to do?"

Both reactions are understandable, and we’ve heard them directly from brands.

You've built your business by understanding your customers, making decisions based on real information, and staying in control of where your money goes. Handing that over to an algorithm, albeit a sophisticated one, can feel like a loss of control. The problem is that Meta isn't really giving you a choice anymore, the platform has fundamentally shifted towards AI-driven optimisation.

But here's what many agencies aren’t explaining clearly, that shift doesn't mean your ads are now automated, and it doesn't mean you've lost control. It means the type of control you have has changed.

This article is for founders who want to understand what's actually happening when you run AI based ads in Meta. You still need good creative, you still need strategy, and running ads is absolutely still a specialist skill, it's just that the skills are evolving.

WHY META MOVED IN THIS DIRECTION

AI optimisations have been around for about a decade. At first, many of us fought automated features and we wanted to manually set everything and retain control. Then, slowly, Meta made manual control harder. They started hiding manual options and  rewarding automation with lower costs, and they've been gradually pushing advertisers towards AI optimisations ever since.

Fast forward to 2026, and Meta doesn't want us micromanaging campaigns anymore. They want us to set a goal and let the algorithm work, but critically, this has come at the point when their AI is genuinely better at processing large volumes of data and making optimisation decisions at scale than any human could.

But, this doesn't mean ads are now ‘automated’.

It means the type of skill required has changed. Advertisers are no longer audience tweakers, they’re now strategists who feed good inputs into a smarter system.What's automated is the optimisation within a campaign. The tactics, not the strategy.

Meta's AI handles:

  • Bid adjustments throughout the day

  • Which ad variations to show to which audiences

  • When to spend your budget

  • Audience expansion based on your customer data

Meta's AI does not handle:

  • Whether your offer is compelling

  • Whether your creative is good

  • Whether you're targeting the right customer

  • Whether your landing page converts

Think of it like this, AI is a really good autopilot, but someone still has to decide where the plane is going, do the maintenance checks on the aircraft, ensure it’s fully fuelled, cleaned and crewed after each flight, and be on board to oversee the journey.

Creating a breadth of content for Meta Ads

WHY YOU NEED PLENTY OF CREATIVE

With Meta's move towards AI optimisation, creative has become the primary lever we can control.

AI optimises using the options you give it. If you give it five mediocre images, AI will show the least mediocre one more often, but it can't turn a mediocre image into a great one.

If you give it ten variations of the same copy angle, AI will optimise delivery. But it can't create a new angle if you haven't provided one.

The maths isn’t complicated

  • 3 images + 2 headlines = 6 possible combinations

  • 5 images + 5 headlines = 25 possible combinations

  • 10 images + 10 headlines = 100 possible combinations

More combinations means the system has more options to test, more options means faster learning, and faster learning often means better performance.

This is why agencies keep pushing creative production.

Not because they're lazy, trying to upsell services or don’t know what they’re doing (all things we’ve heard from new brands when questioning why their agency has been leaning heavily on the creative strategy), it’s because it's what the platform increasingly rewards.

Brands that invest in creative at the same level they did five years ago are likely to see performance plateau. Not because the ads don't work, but because AI can only work with what you give it… and it likely needs more.

At £2k/month spend, mediocre creative might not matter too much.

At £10k/month, it can become your ceiling.

RUNNING ADS IS STILL A SPECIALIST SKILL

With AI increasingly in the driving seat, you’d be forgiven for thinking your agency fee is no longer an essential. However, we’d argue that running ads in 2026 is more of a specialist skill than it’s ever been.  

Five years ago, the specialist skills were:

  • Knowing how to layer audiences

  • Understanding bid strategies and CPC targets

  • Manually pausing underperforming ad sets

  • Testing incrementally

Today, the specialist skills are:

  • Understanding what inputs feed the AI

  • Knowing when to trust automation and when to intervene

  • Creating enough creative variation to give the system meaningful choices

  • Reading data to spot genuine issues rather than normal fluctuations

  • Understanding customer data well enough to feed it back into Meta correctly

The job hasn’t become easier, it’s just evolved.

Let’s look at some of the things AI is doing in your campaigns.

SMART BIDDING

Throughout the day, Meta's algorithm adjusts how much it bids for impressions based on the likelihood someone will convert.

Someone who looks like a buyer? Higher bid.

Someone who looks like a browser? Lower bid.

This happens continuously, across enormous volumes of auctions.

Why it's powerful:

A human simply cannot process and react to that amount of information in real time.

ADVANTAGE+ AUDIENCE

You upload customer data such as email addresses and purchase information.

Meta's AI then identifies people who behave similarly.

Why it's powerful:

The system learns from behavioural patterns rather than relying solely on interests and demographics, and these perform better.

CREATIVE ROTATION

You upload multiple images, videos, headlines and copy variations.

Meta tests combinations and increases delivery to stronger performers.

Why it's powerful:

The system can evaluate hundreds of combinations simultaneously.

…. AND WHAT IT ISN’T DOING

Meta's AI can optimise execution, but it can’t decide what your business should be trying to achieve.The algorithm doesn't know whether your next priority is acquiring new customers, increasing repeat purchases, improving profitability, launching a new product, or gaining market share.

You still need to determine:

  • Are we targeting new or existing customers?

  • Are we driving awareness or sales?

  • Are we prioritising efficiency or growth?

  • Which products should we push?

  • Which customer segments matter most?

  • What does success actually look like?

Back to the airplane analogy above, Meta's AI is great at helping you reach a destination, however, it doesn’t know where your business should be heading in the first place.

Even with autopilot on, you still need someone in the driving seat

THE PROBLEMS AI WON’T SOLVE

One of the biggest misconceptions about Meta's AI is that it can compensate for weak marketing fundamentals.

It can't.

When AI-driven campaigns underperform, the cause is usually not the automation itself, it's that the system is being fed poor inputs, limited data, or mixed signals.

Here are the most common issues we see.

YOUR CONVERSION TRACKING IS OFF

Meta's AI learns from the outcomes you measure.

If purchases are being double-counted, key events aren't firing correctly, or tracking is incomplete, the system will optimise towards the wrong outcomes. Before making campaign decisions, ensure your tracking is accurate and reliable.

When it comes to AI and data, the rule is rubbish in, rubbish out.

YOUR CREATIVE HAS GONE STALE

Meta can only optimise from the creative assets you provide. If your audience has seen the same ads for weeks or months, performance naturally declines.

AI can't invent new creative angles, it can only identify the strongest option from what's available. Regular creative refreshes remain one of the biggest performance levers available to advertisers.

YOU’RE RESTRICTING THE ALGORITHM TOO MUCH

Some advertisers still try to define the perfect audience manually. The problem with that is that today’s optimisation works best when it has room to learn. Excessively narrow targeting often limits performance by preventing Meta from finding valuable customers outside your assumptions. It will also throttle your campaigns quicker when you start to scale.

YOUR LANDING PAGE CREATES FRICTION

Getting the click is only half the job. If the landing page experience doesn't match the promise made in the ad, conversion rates suffer. Meta’s AI can optimise delivery, but it can’t repair a poor customer journey or landing page experience.

YOU DON’T HAVE STRONG FIRST-PARTY DATA

Customer lists, purchase data and conversion signals help Meta understand who your best customers are. The better the data you provide, the more effectively the system can identify similar prospects.

YOUR BUDEGT DOESN’T GENERATE ENOUGH SIGNALS

Meta’s AI improves through data. If a campaign generates very few conversions each week, learning happens more slowly and performance can be inconsistent.

Smaller budgets can work, but campaigns that produce limited conversion volume often take longer to optimise and scale.

Where does your Meta Ads budget need to go?

HOW THIS CHANGES YOUR BUDGET ALLOCATION

For founders already running AI-led campaigns, budget allocation needs to reflect how the platform now works.

CREATIVE BUDGET

This should stay the same or increase.

You need more creative variation than you did a few years ago because creative is now one of the main ways you influence performance.

More angles, more formats, more hooks. More product-led messaging and more customer insight. Not more for the sake of it, but more meaningful options for the system to test.

TESTING BUDGET

Testing hasn’t disappeared, it’s just changed shape.

The platform now handles more of the micro-testing inside campaigns, but you still need budget for bigger strategic tests

  • New offers

  • New creative angles

  • New product priorities

  • New landing pages

  • New customer segments

  • New messaging themes

Meta’s AI can tell you which option performs better, but it can't tell you which ideas are worth testing in the first place.

MANAGEMENT BUDGET

Management is no longer just about making small adjustments inside Ads Manager. The value is now in strategy, creative direction, tracking management and data quality, measurement and interpretation.

The best operators are not just reacting to campaign performance, they're shaping the inputs that determine that performance.

MEDIA SPEND

AI can make scaling more efficient, but only when the foundations are working. If tracking is clean, creative is strong and the offer converts, increasing spend can happen with more confidence.

But if those foundations are weak, more spend usually just exposes the problem faster.

THE BOTTOM LINE FOR FOUNDERS

Meta's AI is not something most founders are preparing for, it's something that’s been active in their accounts for some time. The question is whether it's being used to its full potential.

The AI is excellent at processing huge amounts of data, testing combinations and identifying patterns at a scale, at a rate that humans can’t touch.

But it still relies on people to provide the things it cannot create on its own

  • Strategy

  • Creative direction

  • Accurate data and solid tracking set up

  • Commercial judgement

  • A clear understanding of what success actually means

The brands that are performing best in 2026 are asking 

  • Are we giving it the right data?

  • Are we producing enough strong creative?

  • Are we clear on the commercial goal?

  • Are we interpreting the results properly?

  • Are we improving the inputs, not just watching the outputs?

AI has transformed the way Meta campaigns are managed, creating opportunities that didn't exist a few years ago. However, despite the advances in automation, success in 2026 is far from "set and forget". The brands seeing the strongest results are those combining AI with clear strategy, strong creative, clean data, and ongoing optimisation.

If you'd like a second opinion on whether your current Meta setup is fully optimised for the AI tools, book in for an ads audit. We’ll review your Meta side tracking, audience data, creative, and the broader set up. 

Next
Next

Plans to Scale in Q3? Here’s What to Fix in your Shopify Store First