Creative limited or Creative fatigue on Meta: how many new ads to brief each week

Your best ad is wearing out while you read this. Meta and TikTok both say so in their own help pages, and both leave out the part a production team needs most: how many new ads to have ready, and when.

The platform rules are short. The production schedule that follows from them is where accounts win or lose the month.

What do Creative limited and Creative fatigue mean in Meta Ads Manager?

Both are delivery statuses. Meta's help page, About creative fatigue recommendations in Meta Ads Manager, says they appear in the Delivery column for an ad set or ad "when we believe that your audience has seen the same ad too many times". Meta looks at all recent exposures of the ad's image or video, then compares cost per result with the ads you ran before.

• Creative limited: cost per result is higher than for your past ads, but less than double.
• Creative fatigue: cost per result is at least twice what your past ads cost.

Two details change how you read the label. The check only runs for ad sets with a single creative, so an ad set that rotates several ads will not show it. And the label arrives after cost has already risen, which means you have been paying more for a while by the time you see it.

Jon Loomer, who has documented these statuses in detail, argues that refreshing ads on a fixed calendar is a myth. He has a point. The trigger for launching new ads is a decline you can measure. Producing them is a different job, and it has to start before the decline.

What does TikTok say about refreshing creative?

TikTok's creative best practices for performance ads are more specific than most agency advice. They suggest "between 3-5 different creatives per ad group and 3-5 diversified ad groups per campaign", with the content proposition in the first 3 seconds and the hook in the first 6. On refresh, they say to replace creatives "when delivery results exhibit a consistently declining trend, or when daily new users are low".

There is no seven-day rule in there. The difference from Meta is that TikTok asks you to keep several creatives live at once, which gives the system something to shift spend to while you make the next batch.

Why does the delivery system need more creative than it used to?

In December 2024 Meta's engineers published a post on Andromeda, the retrieval engine behind Advantage+. It was built for "exponential growth in volume of eligible ads" and reported a 6% improvement in recall and an 8% improvement in ads quality on selected segments. The same post says more than a million advertisers used Meta's generative AI tools to create more than 15 million ads in a month, and Meta estimates that advertisers using image generation saw 7% more conversions.

Those are Meta's numbers about Meta's system. The direction is still clear: far more creative competes in each auction, and the system is built to pick from a large pool.

Meta's own guidance on creative diversification adds the rule that matters for production. Ads that look or feel alike are treated as variations of the same creative. Ten recoloured versions of one ad give the system one option. Kenneth Dosanjh set out the full argument, with budgets and ad counts per platform, in Creative is the new targeting.

Is it creative fatigue, or did something else change?

Check before you brief. A price change, a slow landing page or broken tracking all push cost per result up, and no new ad fixes them. Kevin Dosanjh's guide to investigating creative fatigue on Meta covers the full method: compare equal periods on impressions, reach, clicks and purchases, and keep a hypothesis log.

The tree below turns that diagnosis into a brief. Answer it for one ad set.

The last branch is the one most teams skip. When click-through falls while people who do click still buy, the product and offer are fine, and the first seconds of the ad have stopped working.

How many new ads does a weekly cadence need?

Here is how the numbers worked in one published account. IndianBites, an Indian restaurant and takeaway in Copenhagen, needed fresh food creative every week and could not shoot it. The IndianBites case study records a weekly batch of 12 variants built on a style system tuned to their existing hero imagery. Over 11 weeks that came to 142 ads from 18 hero concepts, each tested in 6 to 10 variant cuts, with losers cut within 72 hours. The winning cohorts reached 2.4x return on ad spend; that is a cohort figure, and the case study says it is not account-wide.

Run the arithmetic on it. 142 ads over 11 weeks is about 13 new ads a week. 18 concepts over 11 weeks is between one and two new concepts a week, and each concept carried several openings.

That shape matters more than the total. One concept a week is fragile, because a concept that fails leaves the account with nothing new to test. Two concepts a week, each cut into several openings, gives the system real variety while you learn which reason to buy earned the spend.

Planning in-house? Start from TikTok's range. Three ad groups with 3 to 5 creatives each means 9 to 15 creatives live at once. The number to brief each week is whatever replaces the ones your media buyer turned off last week, plus one new concept in reserve.

How do you brief the next batch before the cost curve bends?

If fatigue shows up as a cost curve, production has to start before the curve bends. In practice that means four habits.

1. Brief on launch day. The next batch is briefed the day the current one goes live, so it is approved before the first decline.
2. Write a hypothesis, then the request. "The opening is unclear, show the product in the first second" can be tested. "New colours" cannot.
3. Hold everything else steady. Keep product, lighting and style locked across batches, so the only thing that changes between variants is the thing you are testing.
4. Name concepts and variants apart. Label each file with its concept and opening, so your media buyer can see which reason to buy is working.

That is the split of work. The media buyer keeps the account. The producer keeps the account fed.

AI Vidia delivers AI video ads in weekly 9:16 batches; the Performance Retainer covers 40 finished videos a month. To see whether a batch beats your current winner before you commit, the Control Test runs 10 variants against your best ad for 14 days in your own account.

Your next step takes about 15 minutes. Open Ads Manager, find your top-spending ad set with a single ad, and check its Delivery status and cost per result against the previous 14 days. Then run it through the tree above.