The short answer
Creative fatigue is when results slide because the same people keep seeing the same ad: frequency up, click-through down, cost per result up, in that order. It comes slower than most teams think. The median winner on our board has been live 120 days, and problem-aware ads last longest: 15% of them have passed a year, against 7% of unaware ads.
What creative fatigue is, and how fast it comes
Creative fatigue is a performance drop caused by repetition: the people the ad reaches have seen it too many times. Meta describes it the same way, and Ads Manager can surface creative fatigue recommendations on ads where it detects the pattern.
The common mistake is expecting it too early. Many teams refresh on a calendar (every two weeks, every month) whether or not the ad is tired. The ads that keep getting funded don't behave that way. In our October 9, 2026 snapshot of 2,443 live Meta ads with $50k+ in estimated spend, the median ad has been live 120 days, and 14% have passed a year. We cover the full distribution in how long winning ads run. This article is about the other side: how to tell when one of yours is actually wearing out, and what to do about it.
The signals, in the order they show up
Fatigue tends to show in a sequence. Watching for the sequence, not one number, keeps you from killing an ad on a bad day.
| Order | Signal | What it looks like | What else could cause it |
|---|---|---|---|
| 1 | Frequency rising | Same people seeing the ad more often, week over week | Budget increase on a small audience |
| 2 | Hook rate or CTR falling | Fewer people stop or click as frequency climbs | A placement shift (more Audience Network or Reels) |
| 3 | CPM rising for that ad | The ad needs more to win the same auctions | Seasonal competition (Q4, sale events) |
| 4 | Cost per result rising | The business symptom you actually feel | Landing page, offer, stock or tracking problems |
If cost per result rises while frequency and CTR are stable, it is probably not fatigue. Check the post-click experience and your event tracking before touching the creative. Broken signal shows up as rising CPA too; see Conversions API and signal quality.
Two practical notes. Look at these per ad, not per ad set; one tired ad can hide inside a healthy ad set. And compare weeks, not days, because daily numbers swing too much to read.
Which ads last longest
Some ads hold up far better than others. The board shows clear differences by the reader the ad is written for:
| Awareness level | Ads | Still under 90 days | Past a year |
|---|---|---|---|
| Problem-aware | 491 | 27% | 15.3% |
| Most aware | 205 | 29% | 13.2% |
| Solution-aware | 496 | 35% | 12.1% |
| Product-aware | 126 | 29% | 10.3% |
| Unaware | 825 | 35% | 7.0% |
Unaware ads, the ones that open with a story or a surprise, are the least likely to last a year. That makes sense: a surprise works once per person. Problem-aware ads name something the viewer lives with every day, and that line keeps landing.
The same pattern shows in how the hook works. Ads that open with a scientific mechanism are the least durable (7.6% past a year, median 102 days). Ads that open with offer pressure run a median 143 days, and plain problem-to-solution openers have 15% past a year.
7%
of unaware ads have passed a year live
15%
of problem-aware ads have
102 days
median run for mechanism-led hooks
143 days
median run for offer-led hooks
The practical reading: if your account leans on unaware stories and mechanism reveals, expect to refresh more often. If it leans on problem-aware and offer ads, expect them to last, and don't swap them out early.
How winners get refreshed
The board also shows when brands add versions to an ad. Here is the share of ads carrying two or more versions, by age:
| Days live | Ads | Share with 2+ versions |
|---|---|---|
| 30 to 89 | 701 | 28% |
| 90 to 179 | 1,037 | 34% |
| 180 to 364 | 367 | 26% |
| 365 or more | 338 | 16% |
Versions peak in months three to six. That is when a winner is scaled, and also when fatigue first bites, so brands add new cuts and openers around the body that works. Ads that survive past a year mostly run as one version: by then they are stable evergreens that don't need propping up.
Who has a better life, cats or dogs?
- Est. spend
- $9M
- Days live
- 567
- Format
- Video, 124s
- Variants
- 6
An unaware ad that beat the odds: over 18 months live, refreshed with six versions instead of being replaced.
Did anyone order a spill-proof and machine washable rug?
- Est. spend
- $2.2M
- Days live
- 1121
- Format
- Video, 39s
Over three years on a single version. A product demo in disguise: it shows the rug surviving a spill, which doesn't get old the way a twist does.
This one easy mistake will cost you thousands when buying a used car.
- Est. spend
- $3.3M
- Days live
- 116
- Format
- Video, 168s
- Variants
- 8
Eight versions at 116 days: the mid-life refresh pattern. A sibling ad with "trick" instead of "mistake" is also live.
Refresh or replace
| What you see | Do this |
|---|---|
| CTR falling, the ad still converts well when people click | Refresh the hook: new first three seconds, new thumbnail, same body |
| CTR stable, conversion rate falling | Not the creative. Check the landing page and offer |
| A refreshed hook also slides within two weeks | Replace the concept: new angle or new awareness level |
| An old winner lost spend to a new one but is still profitable | Leave it running. It may reach people the new one doesn't |
| Frequency high on a small audience | Widen the audience before changing creative |
An early-warning routine
- Track three numbers per ad, weekly: frequency, CTR (or hook rate for video), cost per result.
- Flag an ad when two of three move the wrong way for two weeks in a row. One bad week is noise.
- Have the refresh ready before you need it. For every scaling ad, keep two new openers briefed. Our guide to iterations vs new concepts shows how to write them.
- Refresh first, replace second. Add the new hook next to the original, don't swap it out.
- Plan new concepts on a cadence that matches your mix. Accounts heavy in unaware stories need more new ideas than accounts built on problem-aware ads. See creative velocity.
When you need fresh openers, Ad Radar's search over video transcripts shows how other brands reopened the same idea. A query for your product's problem returns every winning hook that names it.
Figures marked as estimated spend come from Ad Radar's model of engagement on public Meta Ad Library ads. They are estimates, labeled as such, and are best used to rank ads against each other.