Creative fatigue, measured. Most ads stop before they wear out, and most refreshes are not new

KEY TAKEAWAYS

1. Fatigue hits people, not ads. Ad-level averages hide it, so watch cost per result by cumulative impressions.
2. Most ads lose the auction before they fatigue. An early drop is a delivery question, not a refresh signal.
3. A refresh counts only if the platform sees a different creative. Change the visual core, keep the idea if it works.
4. Small accounts cannot see a 20% cost rise in noise. Act on doubling, or pool data across weeks.

Two things changed in the last two months. On 26 August 2026 Meta added a Creative diversity column to Ads Manager that rates each ad set Low, Medium or High. On 5 October TikTok announced Smart+ Creative Selection, which pools brand, creator and AI-made assets in one approval flow. TikTok had already said in July that its new auto-add option exists to reduce ad fatigue. Both platforms now treat fatigue as a creative supply problem. The evidence mostly agrees, with two corrections. Fatigue is rarer than most accounts assume. And most refreshes do not count as new.

Fatigue happens to people, not to ads

Meta's own analysts published the clearest study of the mechanism in May 2023. They modelled clicks against the number of times each person had already seen the same creative, with controls for audience saturation and predicted click rate. Click likelihood fell roughly as (N+1) to the power of -0.43, where N is prior exposures. That works out to about 26% lower on the second view, 38% lower on the third and about 45% lower on the fourth.

Three details in that study matter more than the curve.

• The unit is the creative, not the ad. Ads that share visual elements are experienced as the same thing, so they fatigue together.
• Frequency hides it. Frequency is an average. In their example, an ad set averaged 3, while people reached on day three had already seen the ad 7 times.
• Repetition was never helpful for direct response. They found no early boost from a second or third view. Clicks and conversions got steadily more expensive.

Across all Meta impressions, the viewer had already seen the same creative 4.2 times on average in the prior 30 days. Over 19% of impressions came after more than five prior exposures. In a split test with about 26,000 cases, adding new creative to fatigued ad sets lifted conversion rate by about 8% in the most fatigued cases.

Why ad-level numbers barely move

If each person's click likelihood halves within five views, you would expect ads to collapse. They mostly do not. The Interconnections benchmark, published 12 August 2026, followed 368 Meta creatives from 15 direct-to-consumer brands between April and August. It indexed each creative to its own first 100,000 impressions.

CPM stayed within about 3.1% of its start. The two datasets do not conflict. Meta measured one person's response to repeat views. Interconnections measured the whole ad, which keeps reaching people seeing it for the first time and blends them with repeat viewers. The ads that reach a million impressions are also the ones that held up. So the per-person decay shows up mostly as cost, and only a little as CTR. If you watch CTR alone, you will see fatigue late or not at all.

The decay curves rest on 57 ads from 13 brands, and frequency could not be measured. Read the shape, not the exact percentages.

Most ads never get the chance to fatigue

The same benchmark has a more useful finding. The median creative ran 18 days and delivered 10,665 impressions. 80% never reached 100,000 impressions, and 93% never reached 500,000. The top 10% of creatives took 68% of spend.

That means most ads that "died" were never shown enough to wear out. They lost the auction to a sibling. A drop in the first few days is a delivery or learning question. Fatigue needs repeat exposure, and repeat exposure needs time and spend. Interconnections also found early CTR useless as a forecast. Among 86 creatives, the best early CTR third went on to fewer lifetime impressions than the worst.

When does repetition stop hurting?

Some research says strong ads barely wear out. Chen, Yang and Smith's 2016 experiment found ads that were both original and relevant showed little sign of wear-out, while weaker ads followed the classic rise and fall. TikTok and System1 showed 40 ads to 8,000 heavy TikTok users at four exposure levels in 2025. Ads with strong early branding had 19% less attention decay at heavy exposure, and ads rated 4 stars or more kept building brand equity across repeats.

This does not overturn Meta's result. The outcomes differ. Chen and System1 measured attitudes and brand equity in controlled exposure. Meta measured clicks in live auctions. For a performance campaign, assume every repeat costs something. For brand work, a strong ad can carry more frequency than a weak one, and frequency makes a weak ad worse.

What counts as a new creative?

This is where most refreshes fail. Meta's analysts count ads that share visual elements as one creative. Jon Loomer, who has tested this at length, says colour or single-word changes no longer count as diverse. The new column reports the same thing. Search Engine Land, MarketingConcurrent and a Japanese practitioner note all describe Meta's wording as an estimate of how visually varied the images and videos in an ad set are. Several reported Low ratings on accounts that mixed formats.

Common Thread Collective lists five scored dimensions, including messaging theme and hook type, but cites no Meta source. The reported Meta wording is visual, so treat message variety as useful but unscored.

TikTok works the same way in its automation. Its Smart Creative help page says the system favours the most visually distinct videos, judged by scene similarity. It watches new videos for 3 to 5 days on no more than 10% of the ad group budget, then pauses any that show fatigue and queues replacements.

TikTok and System1 add a qualifier. They found that fresh executions of one strong idea held up better than one unchanged asset. So keep a winning concept, but change what the viewer actually sees: a new scene, person, setting or format. A new headline on the same video is the same creative. For how to budget a set of genuinely different concepts, see how many ad concepts your budget can test. Meta's static-to-video tool is a format change worth trying, with the limits covered in our static versus video piece.

Which numbers have no source?

Three common claims fall apart when traced.

• Frequency 2.5 or 3.5. Good Morning Co traced these to practitioner blogs with no study behind them. Interconnections could not find an origin either.
• TikTok fatigues in 48 to 72 hours. Influee's July 2026 piece gives no data or method. TikTok's own 3 to 5 day watch window is the only platform figure.
• A fixed shelf life in days. Day counts mix spend levels. A $50 a day ad and a $5,000 a day ad age at very different speeds.

Meta's own labels are more concrete. As Loomer documents, the Delivery column shows Creative Fatigue when cost per result is at least double its past level, and Creative Limited when it is higher but less than double.

How much data does a fatigue call need?

Our calculation, not a platform figure. Suppose you compare an ad's cost per result in two windows of equal spend. To tell a real change from noise at the usual 80% power and 5% error rate, you need roughly this many conversions in each window:

Most accounts cannot see a 20% rise for a single ad. That is a reason to use Meta's doubling rule as the trigger when volume is low, and to pool several weeks when it is not.

A worked example (hypothetical)

A skincare brand runs one video ad at a $12 CPM and a $40 cost per conversion. Its first 100,000 impressions cost $1,200 and bring 30 conversions. Over the next 150,000 impressions, it spends $1,800 and gets 38 conversions, so cost per conversion is about $47. That is an 18% rise, close to the benchmark's 250,000 to 500,000 band.

The rise is consistent with early fatigue, but 30 and 38 conversions are far too few to prove it. The ad is still profitable. The sound move is to keep it running, add one visually different concept to the ad set, and let delivery shift. If cost per conversion reaches $80 over 33 or more conversions, the ad is clearly fatigued. Replace it with a new concept, not a recut.

Which fix fits what you see?

A seven-step fatigue check

1. Export ad-level daily data. Include impressions, spend, CPM, clicks and conversions. Note reach where you have it.
2. Index each ad to its own first 100,000 impressions. Smaller accounts can use the first 30 conversions instead.
3. Plot cost per result against cumulative impressions, not days. Mark where CPM moved, so you do not confuse auction prices with fatigue.
4. Rule out delivery problems. Ignore ads under a few days old and ads that never won spend.
5. Check the size of the rise against your volume. Use the table above before calling a 20% change real.
6. Read the Creative diversity column and group near-duplicates. Count concepts the way the platform does, by what the viewer sees.
7. Refresh by concept. Add one visually different creative per ad set, give it a week, and check the hook metrics alongside cost.

Fatigue is real, but it is slower and rarer than most refresh calendars assume. The work is telling it apart from delivery, and then giving the system something it can see as new. If you want help reading this across Meta and TikTok as part of your wider visibility work, join the waitlist.

Questions

How do I know if my ad has creative fatigue?

Compare its cost per result now with its cost over its own first 100,000 impressions, with CPM roughly stable. A rise across cumulative impressions points to fatigue. A collapse in the first few days usually does not, because fatigue builds with repeat exposures. Meta's Delivery column shows Creative Fatigue when cost per result is about double its past level.

What frequency means an ad is fatigued?

No traced study supports the common 2.5 or 3.5 thresholds. Meta's analysts point out that frequency is an average, so an ad set at 3 can include people who have seen the ad 7 times. Use cost per result by cumulative impressions instead, and treat frequency as context.

Does changing the headline or colour refresh a fatigued ad?

Probably not. Meta's analysts treat ads that share visual elements as the same creative for fatigue, and its Creative diversity column scores visual variety. TikTok's Smart Creative favours videos that differ by scene. Change what the viewer sees, such as the scene, person or format.

Do TikTok ads fatigue faster than Meta ads?

Nobody has published data that shows it. Claims of 48 to 72 hours on TikTok come without a method. TikTok's own Smart Creative watches new videos for 3 to 5 days before pausing fatigued ones. Judge both platforms by the same rule, cost against the ad's own baseline.

Sources

1. Analytics at Meta, Creative fatigue. How advertisers can improve performance by managing repeated exposures medium.com
2. Interconnections, The Meta Creative Fatigue Benchmark, Edition 1 theinterconnections.com
3. Jon Loomer, Creative fatigue. What it is and how to prevent it jonloomer.com
4. Jon Loomer, How I create ads now jonloomer.com
5. Search Engine Land, Meta Ads. Why creative diversity matters more than volume searchengineland.com
6. MarketingConcurrent, Meta Ads Manager introduces Creative Diversity score marketingconcurrent.nl
7. Common Thread Collective, Meta's new Creative Diversity score is live in Ads Manager commonthreadco.com
8. Good Morning Co, Meta Ads update September 2026 goodmorningco.com
9. Good Morning Co, What is creative fatigue in Meta Ads? Frequency 2.5 vs 3.5 goodmorningco.com
10. TikTok Ads Help, About Smart Creative ads.tiktok.com
11. TikTok Ads Help, About Smart+ campaigns ads.tiktok.com
12. TikTok for Business, TikTok product preview. What's new for brands in Q3 2026 ads.tiktok.com
13. TikTok Newsroom, TikTok unveils AI-powered updates for advertisers newsroom.tiktok.com
14. TikTok for Business, Creativity pays dividends ads.tiktok.com
15. System1, The long and the short (form) of it system1group.com
16. Chen, Yang and Smith, The effects of creativity on advertising wear-in and wear-out, Journal of the Academy of Marketing Science ideas.repec.org
17. Influee, TikTok ad fatigue influee.co
18. AppsFlyer, Creative fatigue appsflyer.com

NEXT STEP

See where your brand stands.

Your Digital Authority Score, who AI recommends for your target questions, and the placements that close each gap. Your first report is free.

Get your free report →