Meta’s GenAI Ad Tools: Creative or Destructive?

It was in October last year that Meta launched its generative AI toolkit for advertisers that allowed them to create backgrounds, expand images and create versions of ad text based on their original copy. Now, they’ve broadened it by allowing advertisers to create variations of the image with AI-inspired ideas. 

Does this signal the end of ad copy writing and creatives? Before we explore this question, let’s see what exactly Meta has come up with in its latest update. A post on Facebook notes that advertisers can get full image variations for the photo that includes riffs that update its subject or the product being showcased. 

Creative changes or changing creativity?

So, what does this mean? Meta elaborates with an example of someone promoting coffee beans via a cup containing the beverage. The new tools can create variations of this creative such as changing the background scenery, adjusting the coffee cup, and even allowing the “creator” to share text prompts that can fine-tune the creative variations. 

Of course, what it doesn’t tell us is how generative AI would understand the expected outcome of the ad and change its smart ideas accordingly. So, would it be able to convince the user to visit a local coffee shop? Probably not, but it can get her to buy the coffee cup itself by offering variations of the product that may not even exist. 

We spoke to some designers and creative writers of a Bangalore-based agency and their response was epic! Most advertisements are overstatements and now Meta is simply adding to the chaos by offering tools that can be abused by the advertisers by offering products that may not even exist.

Could the toolkit further the cause of fakes?

The example provided by Meta for demonstration purposes suggests that the coffee cup can be transformed into multiple colors, shown from different angles, and containing swirls of foaming milk that are distinct in each example. Now, if this isn’t overstating what would be, asks a senior creative designer at the agency. 

When the creative designing is being done by a team, they usually sit across the table, understand the customer’s needs, and come up with a representation that is closest to reality as well as descriptive of what they want the audience to take away. “We don’t see such tools adding much value, other than reducing a bit of creative time on our workflows,” they say. 

However, the company seems quietly confident that the presence of strong guardrails would prevent any misuse of the toolkit, especially those that are inappropriate or low quality. The post says there are “pre-guardrails that filter out images that gen AI models do not support and “post-guardrails” that remove generated text and image content that does not meet the quality metrics or can be deemed inappropriate by Meta’s yardsticks. 

Meta is pitching it as a support to creativity

Meta was also careful not to overstate the use cases by noting that its goal was “to help you at every step of your journey, whether that’s improving ad performance by helping you develop creative variations, automating certain parts of the ad creation process, or increasing your credibility and engagement through Meta Verified.”

The feature has already started rolling out and in the weeks ahead, advertisers would be able to prompt the AI engine to tailor image variations as well. It will also allow advertisers to add text overlays on their AI-generated images with a dozen most popular font typefaces that a user can pick from. 

Some additional features are rolling out too

The company also announced that its image expansion feature launched for Feeds will now be available for Reels across both Facebook and Instagram. Advertisers can now use AI to adjust their image assets to fit across multiple aspect ratios represented by Reels and Feed. The idea seems to be to help advertisers spend less time repurposing their creative assets, which is something they would welcome. 

Besides these use cases, the new AI toolkit can also be used to generate versions of an ad headline as well as the primary text. This capability is being tested to go with sound like a brand’s voice and tone where previous campaigns can be used as reference. The text generation capabilities are also being moved to Meta’s next-gen Llama3 model. 

While these new features will roll out gradually before the end of 2024, Meta also noted that it plans to expand its subscription service Meta Verified to new markets, having already tested it in Australia, New Zealand, and Canada. It will offer four tiers in the subscription plan with the base features of a verified badge, account support and impersonation guards.

Further tiers will include tools for profile enhancements, creating connections and smarter ways to access customer support. The rollout to Argentina, Mexico, Chile, Peru, France, and Italy would happen over the coming weeks, Meta said while also notifying that the subscription service would soon get expanded to WhatsApp.