Ad Fatigue
Overview
Ad Fatigue is a Juno AI agent. It reviews every active ad across your connected platforms, tests each one against three
fatigue conditions, and flags the ads losing efficiency. For each flagged ad it shows the evidence behind the call and
drafts a recommended fix with an expected outcome you then verify.
Navigation: Juno > AI Agents > Ad Fatigue
Use it to check
| Question | Where to look | | ------------------------------------------- |
--------------------------------------------------------------------- | | How many ads are decaying right now | Fatigued
Ads summary card | | Which ads need creative attention this week | The card grid below the summary | | Why an ad was
flagged | Issues Detected chips on the card | | How large the decline is | View Detailed Analysis, then the condition
panels | | Whether audience saturation is the cause | Frequency panel inside the detailed analysis | | What to do about
it | AI Recommendation block at the foot of the analysis | | How fresh the finding is | Detection date on the card, and
the date range in the analysis header |
The Summary Cards
| Card | What it counts | | ------------------ | ---------------------------------------------- | | Total Ads Analyzed |
Every active ad the agent reviewed in this run | | Fatigued Ads | Ads tripping enough conditions to be flagged | |
Healthy Ads | Ads clearing the conditions |
Track the fatigued share over time rather than against an outside benchmark. A ratio climbing week over week means
creative production is falling behind spend.
Reading an Ad Card
| Element | What it tells you | | ---------------------- |
------------------------------------------------------------------------------------------------------------------------------------
| | Creative preview | The image or video running. No Preview Available is normal for Google Search and Shopping ads,
which carry no single creative image. | | AI FIX badge | A recommendation is ready inside the detailed analysis | |
Issues badge | Number of fatigue conditions tripped | | Ad name | The ad as named in the platform | | Campaign line |
The parent campaign, with a funnel label such as TOFU or BOFU where your naming carries one | | Platform chip | Facebook
Ads or Google Ads | | Issues Detected | The conditions tripped, for example CTR Decline and CPM/CPC Fatigue Pattern | |
Detection date | The day this finding was produced | | View Detailed Analysis | Opens the evidence and the
recommendation |
The Three Fatigue Conditions
| Condition | What it measures | What trips it | | ----------------------- |
-------------------------------------------------------------------------- |
--------------------------------------------------------------------------- | | CTR Decline | Click-through rate for the
current week against the recent baseline | Current week falling below both the last 4 weeks and the longer run average |
| Frequency | Average number of times one person saw the ad | Frequency climbing while performance falls | | CPM/CPC
Fatigue Pattern | Cost per thousand impressions and cost per click against their own history | Both costs rising
together |
The analysis header reports how many tripped, for example Fatigue Conditions Detected (2 of 3). Each panel carries its
own label, CRITICAL for a tripped condition and HEALTHY for one holding.
An ad flagged on two conditions is not twice as urgent as one flagged on two different conditions. Read which two before
ranking your work.
The Detailed Analysis
The header names the ad, the campaign, the analysis window with its length in days, the impressions behind the numbers,
and the platform. Check the window first. A finding built on a closed window is history, not a live signal.
Each condition panel shows the same four figures.
| Figure | What it tells you | | ------------------ | ---------------------------------------- | | Current Week | The
most recent full week of performance | | Last 4 Weeks | The recent baseline | | All-Time Avg | The longer run average
for the ad | | Performance Change | The size of the move |
Read the current week against the last 4 weeks for the short-term move, and against the all-time average to see whether
the decline is structural or a bad week.
Frequency reads in the opposite direction from the others. A falling frequency alongside a falling click-through rate
points away from audience saturation and toward the creative itself.
The cost panel adds the historical averages beside the current figures, so cost per thousand impressions and cost per
click each sit next to their own baseline.
Reading the Patterns Together
| Pattern | Read | | ----------------------------------------------------------------------- |
-----------------------------------------------------------------------------------------------------------------------------------------------------------------
| | Click-through rate down, frequency up, cost per thousand impressions up | Creative fatigue on a saturating audience.
Refresh the creative and widen the audience. | | Click-through rate down, frequency flat or falling, costs up | The
creative is losing attention, or the auction got more expensive. Audience size is not the binding constraint. | |
Click-through rate steady, costs up | Auction competition rather than fatigue. Check seasonality and competitor pressure
before touching the creative. | | Click-through rate down on a Search or Shopping ad | Query mix, competition, or
landing page relevance. Search inventory does not wear out the way a repeated image does, because each impression
answers a new query. |
The AI Recommendation
The recommendation block carries three parts.
| Part | What it holds | | --------------- | ---------------------------------------------------------------- | | What
To Do | The specific change, including budget handling and a test window | | Why | The metric movement behind the
recommendation | | Expected Impact | The outcome the agent anticipates if you run the change |
Expected Impact is a hypothesis, not a forecast. Run the change as written, hold budget where the recommendation says to
hold it, and measure against the baseline named in the Why section. Do not paste Expected Impact into a client report as
a commitment.
The recommendation is a draft for a human to approve. Nothing pauses, duplicates, or launches on its own.
[Patch point: confirm whether the AI FIX badge triggers any action inside the ad platform, and whether findings also
arrive in the AI Inbox as tasks]
How to Use Ad Fatigue to Make Decisions
1. Rank by spend rather than by issue count. Most flagged ads carry the same two issues, so the issues badge does not
separate a $50 ad from a $5,000 ad.
2. Open the detailed analysis before acting. The chips name the condition and the panels size it.
3. Check frequency before widening an audience. Widening a lookalike against a falling frequency spends effort on the
wrong problem.
4. Separate Meta creative flags from Search and Shopping flags. The first calls for new creative, the second usually
calls for a query and bidding review.
5. Change one thing at a time and keep the original running long enough to compare, following the test window in the
recommendation.
6. Record the result against the baseline named in the analysis, so the next review starts from evidence rather than
from memory.
What to Check and When
| When | What to check | | ------------------------- |
---------------------------------------------------------------------------------- | | Weekly | The Fatigued Ads count
and the detection dates behind it | | Before a creative sprint | The flagged list, to brief the team on evidence rather
than on instinct | | Two weeks after a refresh | Whether the refreshed ad now clears the conditions | | Monthly | The
fatigued share of total ads analyzed, as a read on creative supply | | In every business review | Which recommendations
were run and what happened, rather than the flag count alone |
Troubleshooting
| Symptom | Likely cause | Fix | | ------------------------------------------------------------- |
--------------------------------------------------------------------------------------------------- |
--------------------------------------------------------------------------------- | | No Preview Available on a card |
The ad is a Google Search or Shopping ad with no single creative image | Expected. Identify the ad by name and campaign.
| | Detection date is older than a few days | The agent runs on a schedule rather than on demand | Check the schedule
under Task Scheduler, then re-run if your platform supports it | | An ad you expect to see is missing | It cleared the
conditions, or it fell outside the analysis window | Confirm the ad was active during the window shown in a detailed
analysis | | Performance Change does not reconcile with the All-Time Avg | The change compares the current week with the
recent baseline rather than with the all-time average | Read the panel figures directly rather than recomputing from the
average | | An ad is flagged with normal frequency | Fatigue is measured on three conditions, and two are enough | Read
which two tripped before assuming audience saturation | | Recommendation names a campaign or audience you no longer run
| The analysis window closed before your change | Re-run the agent, then act on the newer finding | | Every flagged ad
shows the same two issues | Common when one platform dominates the account | Rank by spend and by funnel stage instead
of by issue type |
Common Questions
Which baseline does Performance Change use?
The panel reports the current week, the last 4 weeks, and the all-time average side by side, and the change figure
tracks the recent baseline. [Patch point: state the exact denominator, since the figure reconciles with the last 4 weeks
rather than with the all-time average]
Why is an ad called fatigued when frequency is normal?
Flagging needs two of the three conditions, and frequency is only one of them. An ad with a falling click-through rate
and rising costs meets the bar on its own. In that case the creative is losing attention rather than reaching the same
people too often.
Do Google Search ads suffer creative fatigue?
Not in the same way. A repeated image wears out an audience. A search ad answers a new query each time, so a falling
click-through rate there usually points at query mix, competition, or relevance. Check the search terms report before
rewriting the ad.
Does the agent change anything in my ad account?
No. Every recommendation waits for a person to run it.
How often does the agent run?
On a schedule rather than on demand. The detection date on each card and the window in the analysis header tell you how
current the finding is. [Patch point: state the default cadence and where to change it]
Related Articles
- Getting Started with LayerFive
- AI Inbox
- Creative Reporting
- Meta Performance Dashboard
- Google Performance Dashboard
- Attribution Analytics
- Glossary of Terms