The skill
Install once in Claude. It reads your Ads Manager or TikTok export by creative ID and flags the creatives that are fatiguing, with the evidence behind each call.
What's inside
Three files, one job: see the decay first
Built for performance marketers, not analysts.
Install guide
Step by step setup, turn on code execution, upload the skill, toggle it on. Includes a no-install option for anyone who cannot add skills.
Plan B prompt
The same logic as a plain prompt. Paste it into any Claude chat with your export and run the scan without installing anything.
How it works
Three steps, from a creative-level export to a refresh plan.
Step 1Export at the creative level
Pull your ad performance from Meta or TikTok with a creative ID column and per-creative metrics, hook rate, CTR, frequency, First-Time Impression Ratio, CPM and CPA.
Step 2Claude scans it
It sets each creative's baseline from its first three to five days, flags the leading signals that hold for three or more consecutive days, and classifies each creative as healthy, watch, refresh now, or confirmed fatigue.
Step 3Get a status table and a plan
You receive a per-creative status table with the signals that tripped, a refresh and testing plan matched to your spend, and a list of any fields that were missing.
There is no live connection to your ad account. You bring the export, the skill does the analysis.
How to detect creative fatigue before ROAS drops
Creative fatigue is the predictable decay in response when the same ad is shown too many times. Most teams catch it from CPA or ROAS, which only move after the money is spent. The signals that warn you turn days earlier, at the creative level, and reading them is what this skill does.
What creative fatigue is, and why it matters now
Creative is now the dominant performance lever, close to half of incremental sales in large studies, and after Apple's tracking changes the algorithm reads engagement with your creative to decide who sees it. That makes creative the targeting mechanism, not just the message. Meta's own research found conversion likelihood drops about 45 percent by the fourth exposure, with no beneficial wear-in to wait for, so a tired creative is both a weaker ad and a weaker targeting signal.
Leading indicators versus lagging indicators
CPA and ROAS are lagging indicators. By the time they move, the fatigue is days old. The signals that turn first are hook or thumbstop rate, CTR against the creative's own launch baseline, First-Time Impression Ratio in prospecting, frequency, and CPM creep, and they typically break three to five days before CPA does. The catch most teams miss is that fatigue lives at the creative level, not the ad set, so the same creative can be exhausted while any single ad's frequency still looks fine. You read it by creative ID.
How the skill flags fatigue
The skill sets each creative's baseline from its first three to five days out of learning, then watches the leading signals against that baseline, flagging only when a signal holds for three or more consecutive days so a single noisy day does not trigger a false alarm. It classifies each creative as healthy, watch, refresh now, or confirmed fatigue, using thresholds like a 25 to 35 percent CTR decline or frequency above four in prospecting. Before you produce anything, it runs the fatigue versus saturation test, since a fresh creative that restores CTR in the same audience means fatigue, while recovery only with a new audience means saturation, and the two need opposite fixes. It also flags when a rising CPA is probably not creative at all.
Refresh cadence and testing velocity
The durable fix is shipping more concepts faster, not refreshing in a panic. The skill sets a cadence near the platform lifespan, roughly 10 to 14 days on Meta and about weekly on TikTok, but lets the leading signals trigger action rather than the calendar. It sizes a weekly testing target matched to your spend, around 20 to 30 new creatives per 100,000 dollars of monthly spend, and tracks your winner rate against the benchmark, since a rate well under 4 percent means your problem is concept diversity, not volume. It rotates gradually, keeping winners and adding about 30 percent new into existing ad sets, so you do not reset the learning phase.
Using AI without accelerating fatigue
AI makes variation cheap, which is real leverage, but the rigorous evidence shows it reaches parity, not dominance, and it homogenizes output. If the whole market generates from the same models, creatives converge and fatigue arrives faster. The skill uses AI for hook and angle variants and library tagging while reserving humans for net-new concepts, and it treats library diversity as something to measure, cutting AI volume when hooks start to converge. For the full evidence base, including Meta's decay curve and the studies on AI homogenization, read Creative Fatigue Is a Targeting Problem Now, and AI Is Making It Worse.