Meta's Creative Diversity Rating Says Low: What It Measures and What to Change

A new column appeared in Meta Ads Manager this year: creative diversity, rating each ad set Low, Medium or High. Then a great deal of advice appeared around it, most of it built on numbers Meta has never published. The metric is worth acting on, but only once you separate what Meta documents from what the ecosystem has inferred — because those two lists point at different production changes, and one of them costs a lot more than the other.

The short answer

Meta's creative diversity metric rates each ad set Low, Medium or High based on how visually similar its images and video thumbnails are to one another. Meta labels it estimated and in development, and it is a reporting metric rather than a documented delivery penalty. It reads visual sameness, not creative volume: twenty ads built from the same opening frame can rate Low while five genuinely different concepts rate High. Raise it by varying the things the metric can see — first frame, format, subject, setting, composition — and raise performance by varying the angle behind them. Those are not the same list.

What the metric is, precisely

Creative diversity arrived in Ads Manager reporting during 2026 and was widely noticed from August onward. You find it under Columns, then Customize columns, then Creative diversity. It reports at ad set level with three values — Low, Medium, High — and Meta describes it as estimated and in development.

What it estimates is how visually diverse the images and videos in an ad set are. The comparison is between your creatives, not against a benchmark, and reporting on the metric consistently describes it as reading how alike your images and video thumbnails look. Meta's stated purpose is to surface creatives that have become visually similar, in service of reducing fatigue and supporting longer-term performance.

That definition carries two implications most write-ups skip. First, this is a measure of sameness, not of quantity: an ad set with thirty assets can rate Low and an ad set with six can rate High. Second, because the signal is visual and thumbnail-weighted, a change that transforms your message but leaves the opening frame intact may not move the rating at all — and a change that alters the frame but says the same thing may move it while doing nothing for results.

What Meta documents versus what the ecosystem claims

This is the table to read before you act on anything else you have seen about this metric. A striking amount of the guidance circulating since August states specific numeric thresholds that are not in Meta's documentation, and in several cases originate in third-party analytics tools that compute their own similarity scores.

Claim	Status	What to do with it
The metric rates ad sets Low, Medium or High on visual similarity of images and video thumbnails	Documented by Meta in Ads Manager and its help centre	Treat as fact
It is estimated and in development	Meta's own label	Treat as fact, and as a reason not to optimise to it mechanically
Its purpose is flagging visual sameness to reduce fatigue	Meta's stated intent	Treat as fact about intent, not as a promise of causation
"Similarity above 40% means you need more variety; above 60% triggers retrieval suppression"	Third-party tooling thresholds, not Meta metrics. Different vendors compute similarity differently	Ignore the specific numbers. The directional idea — near-identical creatives get collapsed — is plausible; the percentages are not portable
A Low rating directly suppresses delivery or raises CPM	Not documented. An inference from how retrieval and ranking systems deduplicate	Do not brief it to a client as a fact. Act on it because creative fatigue is real, not because the column is
"Seed consolidated ad sets with 15 to 25 diverse creatives"	Vendor and agency recommendation	Reasonable for a high-spend e-commerce account. Wrong advice for a business spending £3,000 a month, where it means twenty cheap assets instead of five good ones
Videos with similar opening sequences are treated as near-identical	Reported by practitioners as reflecting Meta's guidance; consistent with a thumbnail-weighted visual signal	Act on it. It is the cheapest high-leverage change available

The mistake to avoid

Optimising directly to a metric Meta labels "in development" is how accounts end up with a wall of visually scattered, strategically identical ads. The rating is a smoke alarm, not a scoreboard. If it says Low and your cost per result is stable and your frequency is flat, you have a diagnostic worth noting and no emergency.

The diversity axes, ranked by cost

Here is the part no ranking page for this query currently provides: which specific production changes plausibly move a visual similarity estimate, which change performance, and what each one costs to produce. The two middle columns diverge more than people expect, and that divergence is the whole practical insight.

Axis to vary	Moves a visual similarity signal?	Moves performance?	Marginal production cost
First 1–3 seconds / opening frame	Strongly — thumbnail-weighted	Strongly — this is where scroll-stop happens	Very low if planned at the shoot; a re-edit otherwise
Format (short video, long video, carousel, static, UGC-style)	Strongly	Moderately — different placements favour different formats	Low from existing footage
Subject or on-screen person	Strongly	Strongly	Medium to high — usually needs a second shoot or a second creator
Setting and background	Strongly	Weakly on its own	Low to medium — one extra location or corner of the same location
Composition and framing (wide vs close, vertical native vs cropped)	Strongly	Moderately	Very low if shot deliberately; impossible to fake in post
Colour grade and palette	Moderately	Weakly	Very low — and the classic cosmetic shortcut, so use it as a garnish not a strategy
Angle, hook or value proposition	Weakly, unless it changes what is on screen	Strongly — the single biggest performance lever	Low to plan, and the one thing that cannot be outsourced to an editor
Headline and primary text only	No	Weakly	Near zero — which is why accounts are full of it
Text overlay on an otherwise identical frame	Barely	Weakly	Near zero — the most common form of fake diversity

Read the top rows and the bottom rows together and the strategy writes itself. The cheapest genuine wins are the opening frame, format and framing, because all three are decided at the shoot and cost almost nothing when they are planned rather than retrofitted. The expensive win is a new subject. The illusion is everything in the bottom two rows: swapping headlines and stamping different text over the same frame produces a lot of ad IDs and very little diversity, visual or strategic.

A one-shoot plan that produces a High rating

Most accounts do not have a creative volume problem. They have a pre-production problem: the shoot was planned to produce one film, and everything after that is variations of one frame. Planning divergence in advance costs almost nothing and changes both the rating and the results.

Write four angles before you write any scripts. Not four hooks for one angle — four different reasons a person buys. Problem, outcome, objection, comparison is a serviceable default set.
Shoot each angle with a deliberately different opening. One opens on a face mid-sentence, one on the product in use, one on the problem state, one on text-free motion. That single rule does more for the rating than any editing decision.
Vary framing across angles, not within them. Wide for one, tight for another. Two setups in the same room count as two settings for a visual signal and cost you fifteen minutes.
Capture stills and carousel frames on the day. Format diversity from one shoot is free; cropping video frames afterwards is not the same thing and usually looks it.
Shoot vertical natively. Cropped 16:9 reads as the same composition as its parent asset, so you pay for an asset that adds nothing to diversity.
Keep a stability slot. Reserve a share of spend for proven performers and accept that the slot pulls your rating down. That is a deliberate trade, not a failure.

The economics here are the same ones that govern what a brand video actually costs: the cost of an additional usable asset collapses when the shoot was planned for it and stays high when it was not. Diversity is simply another deliverable that has to be specified in pre-production. Our one-shoot, month-of-content approach is the same discipline applied to organic.

When Low is the right answer

A rating is not a verdict, and there are several situations where chasing High would actively cost you money.

Situation	Act on a Low rating?	Reasoning
Retargeting or bottom-funnel ad set with a narrow audience	No	Small audiences fatigue on frequency, which you can already see directly. Consistency often helps recall here
Single-SKU account early in testing	Not yet	You need a validated angle before you multiply production against it
Prospecting ad set, stable spend, rising frequency and falling CTR	Yes, urgently	This is the fatigue pattern the metric exists to surface, and the numbers agree with it
Scaling ad set where spend has doubled in a month	Yes	Diversity requirements scale with reach; what sufficed at one budget will not at three times that
Account where every asset is the same frame with different text	Yes	The rating is telling you something true that you could have found by opening the ads tab
Rating is Low but cost per result and frequency are both flat	Note it, don't react	An in-development estimate is not a reason to spend money this month

The broader point is that this metric is most useful as an early warning that arrives before your performance data does. Fatigue shows up in frequency and click-through rate eventually; a visual sameness estimate can flag the setup that will cause it. That makes it a planning input, not a KPI. How you then test the new creative is a separate discipline with its own rules — concepts before variations, kill criteria fixed in advance — which is the ground our Meta creative testing framework covers, and which this metric does not replace.

What we'd do about it

Add the column, sort your prospecting ad sets by it, and use it to prioritise which shoot brief gets written next. Do not rebuild an account around it, do not quote vendor similarity percentages to your client as Meta data, and do not mistake twenty assets for twenty ideas. If the account is spending real money against four visually identical ads, the metric has earned its place by telling you so in one screen.

Frequently asked questions

What is Meta's creative diversity metric?

It is a reporting metric in Meta Ads Manager that rates each ad set Low, Medium or High based on how visually similar its images and video thumbnails are to one another. You add it under Columns, then Customize columns, then Creative diversity. Meta labels the metric as estimated and in development, and describes its purpose as surfacing creatives that have become visually similar in order to reduce fatigue and support longer-term performance.

Does a Low creative diversity rating hurt delivery or raise CPM?

Meta does not document creative diversity as a delivery penalty. It is presented as a reporting metric. The widely repeated claim that a Low rating suppresses delivery is an inference from how ranking systems deduplicate near-identical creatives, and the specific similarity percentages quoted around it come from third-party analytics tools rather than from Meta. Creative fatigue is a real cost; the column is a diagnostic for it, not a documented auction mechanic.

How do I improve my creative diversity rating?

Vary the things a visual signal can see. The cheapest high-leverage changes are the opening one to three seconds of video, the format mix across video, carousel and static, and the framing and composition. Changing the on-screen subject or setting also moves it but costs more. Changing only headlines, primary text or a text overlay on the same frame does almost nothing, which is why accounts full of those variants still rate Low.

Is creative diversity the same as having more ads?

No. The metric measures sameness, not volume. An ad set with thirty assets built from one opening frame can rate Low, while an ad set with six genuinely different concepts can rate High. Adding cosmetic variants increases the number of ad IDs without increasing diversity in either the visual or the strategic sense.

How many creatives should an ad set have?

There is no documented Meta number. Vendor and agency guidance commonly suggests seeding consolidated ad sets with 15 to 25 diverse creatives, which is reasonable advice for a high-spend e-commerce account and poor advice for a small budget, where it produces twenty weak assets instead of five strong ones. Scale the count to your spend and reach, and prioritise distinct angles over asset count.

Where does creative diversity appear in Ads Manager?

In the reporting columns at ad set level. Open Columns, choose Customize columns, and add Creative diversity. Advertisers began noticing the column from August 2026 onward, and it carries Meta's in-development label, so treat the value as directional.

The takeaway

Meta has given you a free diagnostic for a real problem, wrapped in a lot of advice built on numbers it never published. Use the column the way it was designed: sort prospecting ad sets by it, treat Low plus rising frequency and falling click-through as a signal to brief a shoot, and treat Low with flat costs as a note rather than an emergency. When you do produce, spend the effort on the axes that are cheap and effective at once — different openings, different formats, different framing, planned before anyone starts recording — and stop paying for variants that change the text over the same frame. Those have never fooled the audience, and they do not fool the metric either.

Meta's creative diversity rating says Low. Here's what it measures and what to change