Meta is building, testing and introducing a range of AI tools to help marketers run more effective ads.

The AI Sandbox

The newly-announced AI Sandbox project, which is currently in testing with a select group of advertisers, includes a few different ways that advertisers can use AI. Meta wants to use the sandbox to learn what features work best for brands and then make sure they’re easy to use moving forward.

Meta

“To start, we are building tools like text variation, background generation and image outcropping to do things like make an ad’s text more engaging or improve parts of its creative.”

Meta mentioned three tools there – so let’s take a deeper look at them:

Text Variation generates multiple versions of text that highlight the key message in an advertiser’s copy. An advertiser can then choose the one they like best, or pick different messages for different audiences.

Background Generation creates background images from a text input, which will let marketers diversify their creative assets and try various backgrounds more quickly.

Image Cropping adjusts creative to different aspect rations such as Stories or Reels, which will let marketers make multiple versions of the same creative fit across different platforms without spending ages in an editing suit.

Meta says that it will “gradually expand” these tools to more advertisers in July and plan to add them to some products later in the year. So, if you’re reading this with a smile while imagining how much time it’ll save you, it may pay to dial back expectations! 

Where do these fit within Meta Advantage?

If you’re someone who is excited about AI in advertising, then Meta has a tool you can use right now. It’s called Meta Advantage, and it uses AI and machine learning to optimise campaign results, personalise ads and save brands money.

More marketers are waking up to the possibilities that Meta Advantage offers. For example, there are three times as many advertisers using the feature now than there were six months ago. That’s pretty spectacular growth!

Meta

“Today, we are introducing new Meta Advantage features that will allow advertisers to take advantage of AI quicker, give them more flexibility in the creative they use and help them measure what’s working.”

The new features include:

Switch manual campaigns to Advantage+ in just one click by going to Ads Manager and clicking on “duplicate” within an ad campaign.

Using video creative in catalogue ads. Previously marketers were limited to static product images, but now they will be able to use video. Meta will use AI and machine learning to show your video to people in Feed, Stories, Watch and Reels and marketers across manual or Advantage+ shopping campaigns.

Rolling out Performance Comparisons to let marketers compare existing manual sales campaigns with Advantage+ campaigns and understand how automation can positively impact a campaign.

Improving performance with Advantage+ audience will give advertisers an updated way to reach people and drive conversions.

“Instead of using an advertiser’s audience targeting inputs as hard constraints, such as Men, 18-35 years, who like baseball, advertisers who use Advantage+ audience will add their audience inputs as suggestions to guide who sees an ad.”

Meta says the feature will allow its ad system to find people outside a hard constraint to target and convert. However, it will honour hard constraints for things like age limits or shipping locations.

Meta’s evolving AI capabilities

Meta is building its AI capabilities quickly – which is no surprise given it invests billions of dollars a year into its AI infrastructure. Meta is using larger, more complex models in its ad systems that share learnings which each other to enhance advertising and deliver better results.

Meta

“For example, previously on Instagram, we needed one model for optimizing clicks on Story ads and another model to optimize conversions or sales for ads on Reels. Now, with more advanced AI modeling that optimizes across all surfaces (Feed, Story, Explore, and Reels), it allows us to transfer learnings across multiple objectives at the same time.”