TL;DR
Meta quietly added a Creative Diversity column to Ads Manager that scores each ad set Low, Medium, or High. The score reflects a real mechanism: Meta's ad system groups visually and thematically similar ads into a single creative entity, and one entity gets one slice of reach no matter how many variants you made. Ten product shots with different headlines register as one creative. The fix isn't more variants, it's genuinely different concepts: different personas, different angles, different formats. Variation is not diversity.
Your ten ads are one ad
Picture the scene. You made ten versions of the ad. Different headline, different crop, different background color. You felt productive. You put them in an ad set and waited for the algorithm to find ten pockets of buyers.
Meta looked at them and saw one ad.
That is not an opinion anymore. It's a score, sitting in your Ads Manager right now. Late in the summer of 2026, Meta added a Creative Diversity column to Ads Manager reporting (find it under Columns, then Customize columns). It rates each ad set Low, Medium, or High. Meta labels it "in development" and "estimated," which means the formula will change, but the direction is set and it isn't reversing.
Why this exists: the machine reads your creative now
To understand the score, you need to understand what changed under the hood. Meta's ad system no longer leans on your audience settings the way it used to. Its retrieval engine (Meta calls it Andromeda) reads your creative directly to decide who should see it. The creative is the targeting now. That's the part everyone repeats.
Here's the part people miss: when your ads look and sound alike, the system groups them into a single creative entity. Ten product shots with small tweaks register as one creative, not ten. Your reach gets capped at a fraction of the available audience segments, no matter what you spend. You can't buy your way out of sameness. The system has already decided your ten ads are one idea, and one idea gets one audience.
This is why some accounts can spend heavily and still feel like they're talking to the same room. They are. They built one room.
What the score actually measures
The score looks at real dimensions of variety:
- Content style. UGC-style video versus polished product shot versus demo versus founder talking to camera. These read as different things to the system.
- Messaging theme. Benefit-led versus social proof versus urgency versus comparison. Same product, different reason to care.
- Hook type. Question versus stat versus before-and-after versus contrarian claim. The first two seconds are a dimension of their own.
- Format. Video versus static versus carousel. Format changes how the creative is consumed, so it counts.
- Who's on screen. Different faces, different settings. A dermatologist explaining ingredients and a customer showing results are different entities.
Ten ads in an ad set does not guarantee a High. If all ten open on the same frame with the same person saying the same line, Meta scores that Low and treats the set as one creative for reach purposes. And here's the uncomfortable corollary: a Low score isn't a vanity metric. It's a reach cap wearing a label.
Variations are not diversity
Let's be precise about what doesn't count, because this is where most accounts fail:
- Same image, different headline. Not diversity.
- Same video, different crop. Not diversity.
- Same concept, different background color. Not diversity.
- Same hook, different thumbnail. Not diversity.
- Same script read by a different creator, shot the same way. Probably not diversity either, though it might fool a human reviewer.
The test we use: would a viewer notice the difference in the first two seconds, and would the ad appeal to a different person? A new persona with a new problem is a new concept. A new headline is a new coat of paint.
How to build a genuinely diverse ad set
Stop planning variants. Plan concepts along three dimensions:
1. Persona. Who is the ad for? List three to five distinct buyer types. The busy parent, the skeptical comparison shopper, the gift buyer, the bargain hunter. Different people, different triggers.
2. Angle. Why would that person buy? Social proof, problem and solution, demonstration, comparison, transformation, offer, objection handling. Each angle is a different reason to care, which means a different audience the system can find.
3. Format. UGC-style video, founder to camera, product demo, static image, carousel, testimonial compilation. Format changes consumption, so it changes reach.
A worked example: a skincare brand selling one moisturizer could run a dermatologist explaining ingredients (expert persona, education angle, talking head), a customer in her fifties showing her routine (mature skin persona, transformation, phone video), a guy who "never used moisturizer" (male persona, objection handling, humor), a static with a five-star review (social proof, image), and a carousel comparing it to a pricier brand (comparison). Same product, five concepts, five different audiences the system can match. That's what a High looks like.
The volume trap, updated
You've heard that competitive brands now run dozens of assets a month and leaders run hundreds. That's true, and it's also where this score bites hardest: volume without variety is just expensive sameness. A hundred ads that all appeal to the same customer archetype are a hundred attempts to win the same person. A hundred ads across genuinely different concepts are pathways to different people.
Before you brief your next batch of creative, ask the question the score is asking: if Meta grouped these by what they actually are, how many ads do we really have? If the answer is one, you don't have a testing problem. You have a thinking problem.
FAQ
Where do I find the Creative Diversity score? In Ads Manager, go to Columns, then Customize columns, and look for Creative diversity. It's labeled "in development" and "estimated," so treat the exact formula as provisional.
Does a Low score hurt delivery, or is it just informational? The score reflects a real grouping mechanism: similar ads are treated as one creative entity for reach. A Low score means your ad set's reach is effectively capped at what one concept can win, regardless of spend.
How many concepts should an ad set have? There's no magic number, but aim for genuinely different concepts across persona, angle, and format rather than a variant count. Five truly different concepts will outperform twenty paint jobs.
Will Advantage+ creative enhancements fix a Low score? No. Automated enhancements (backgrounds, crops, text variations) create variants of the same concept. They don't create new concepts. The diversity has to come from your briefing, not the machine.
Is this related to Andromeda? Yes. The retrieval engine reads creative to match ads with people, and it groups similar creative into single entities. The diversity score is the visible surface of that mechanism.
Written by the MetaMaxd team. We run Meta ads for brands that are tired of agency theater: no vanity metrics, no "hacks," just accounts engineered to make money. If your ad sets are full of ads and empty of reach, that's what our audits are for.
