How do I know an ad is fatiguing, and what do I do about it?

An ad works because it is new to the people who see it. Meta shows it to the people most likely to respond first, then to everyone else, then to the first group again. By the third or fourth time the same person has seen the same creative, they stop looking. Frequency goes up, click-through goes down, and the cost of each result climbs, with nothing else changed.

It is not a mystery and it is not a sign the idea is wrong. It is a sign the surface is worn out.

Three columns, read together over a week:

• Frequency. Average times each person has seen the ad. Above 3 to 4 within a week for a prospecting campaign, fatigue is likely. Retargeting tolerates more.
• CTR (link). Compare this week with the ad's first full week. A fall of 20 to 30% while frequency rises is the signature.
• Cost per result. Rising while CTR falls and frequency rises. Any one of these alone can have another cause; all three together do not.

Do not read a single day. Meta's delivery is lumpy and a Tuesday dip is not fatigue.

Because their advertisers keep making versions. In Dovey's index of the Meta Ad Library on September 24, 2026:

• Buildify, "Build apps with AI": 341 days, 129 versions.
• Peak Footwear, "Walk pain free or get money back": 401 days, 24 versions.
• Feelingirl, a Black Friday shaper ad: 280 days, 21 versions.
• Fitbod, form-tip videos: 176 days, 18 versions.
• RatePunk, "Real deals, no travel hype": 120 days, 13 versions.
• Try Strawberry, brightening cream: 301 days, 12 versions.

A version, in Meta's count, is another ad using the same creative and text: a different audience, a different placement, a small change. The idea is constant. The surface is rotated. That is the whole method.

The ads with one version that run 400+ days do exist (Tempo, 565 days; Syncfusion, 469) and they tend to be evergreen offers in low-frequency categories. If you are not one of those, plan on versions.

In order of cost, cheapest first, and stop when the numbers recover:

1. The hook. New first line, same everything else. Ten minutes. Alevia runs the same body under "Bloating? Try Amla.", "Fatigue? Try Amla." and "Amla vs Coffee".
2. The first frame or the picture. Same copy, new visual. For video, a new opening three seconds.
3. The offer framing. Not a bigger discount, a different frame: "free shipping" instead of "10% off", the guarantee instead of the price.
4. The format. The still becomes a carousel, the talking head becomes a demo.
5. The angle. Problem-led instead of offer-led. This is a new ad, not a refresh, and it should be tested as one.

Keep the winning idea until the idea itself stops working, which takes much longer than the creative.

Before the numbers move. A working routine for a small account:

• Every week, one new version of each winner (a new hook or a new visual).
• Every two weeks, one new angle to test against the winner.
• Every month, retire what has been below the account average for two weeks.

The brands above with 12 to 129 versions are on a schedule like this. Dovey makes the new versions: ask for a remix of the winner in your brand and it comes back in three sizes on request, and you approve the ones you like.

• A seasonal offer ending (the Black Friday ad that stayed live into spring).
• An audience that was small to begin with: frequency climbs fast and refreshing the creative will not fix it. Widen the audience.
• A landing page change. Check the page before the ad.
• A competitor launching against you. Check the Ad Library.