Meta Platforms (NASDAQ: META) is quietly reshaping how advertisers build and test their creative assets, issuing updated best practices tied to how its ad ranking models evaluate content.
The company has been advising advertisers that its models will not rank an ad highly if it shares the same visual look or opening hook as other creative from the same advertiser.
Olivia Cripps, global paid media director at social agency Billion Dollar Boy, said Meta has made clear that its engines will treat visually similar ads as fundamentally the same piece of content.
“If you had a video of me, and then also a video of me, and this one has a text that says, ‘Text A,’ and this one says, ‘Text B,’ that would actually be considered the same creative,” Cripps said.
The clarification has helped Billion Dollar Boy structure its creative testing process for clients, allowing the agency to approach ad variation with greater strategic precision.
Meta has also been offering more granular advice to agency partners, including guidance to mix video and static assets and to vary messaging tone by using different hooks, according to Ankit Jadav, associate director of paid social at Rain the Growth Agency.
Meta spokesperson Alisha Swinteck said the company’s guidance is deliberately broad and outcome-focused, rather than designed to help advertisers engineer around any specific model’s behavior.
“We don’t get into the specifics of how our ranking models represent individual creative assets. Our guidance to advertisers is deliberately outcome-based, like ‘diversify your creative, test, and let the system optimize,'” Swinteck said in an email.
Swinteck added that creative diversification is a “longstanding best practice as we see it deliver measurable results,” reinforcing that the advice is not a reaction to any single model change.
During Meta’s second-quarter earnings, CFO Susan Li said the company altered its models for understanding user behavior to better analyze organic activity and match users with relevant ads.
Those model changes, combined with the deployment of Meta’s recommendation model Gem, drove an 8.3% increase in ad clicks, Li said on the earnings call.
Meta is also testing creative insights metrics within its Insights API, which delivers ad performance data directly to advertisers and agency partners.
Swinteck confirmed that a “small number of agencies/GCR resellers” have been participating in tests of those creative insights metrics since June of this year.
The broader push reflects Meta’s continued investment in its advertising infrastructure, as the company works to align marketer behavior more closely with how its increasingly sophisticated models assess and rank creative content.
