AI DIAGNOSTIC PLAYBOOK FOR UA TEAMS
25 mobile growth problems, 100+ checks, and AI prompts that help you find what actually changed
Most UA teams don’t need more AI prompts. They need a faster way to understand what actually changed. CPA jumped. ROAS dropped. A winning creative started fading. A cheap GEO brought users who never converted. This playbook gives you 25 UA diagnostic cards to separate real signals from noise — before you pause, scale, refresh, or report.
The UA AI Diagnostic Playbook
25 mobile growth problems, 100+ checks, and AI prompts that help you find what actually changed
Most UA teams don’t need more AI prompts. They need a faster way to understand what actually changed. CPA jumped. ROAS dropped. A winning creative started fading. A cheap GEO brought users who never converted. This playbook gives you 25 UA diagnostic cards to separate real signals from noise — before you pause, scale, refresh, or report.
The UA AI Diagnostic Playbook
25 mobile growth problems, 100+ checks, and AI prompts that help you find what actually changed
Most UA teams don’t need more AI prompts. They need a faster way to understand what actually changed. CPA jumped. ROAS dropped. A winning creative started fading. A cheap GEO brought users who never converted. This playbook gives you 25 UA diagnostic cards to separate real signals from noise — before you pause, scale, refresh, or report.
How to Use This Playbook
Find the growth symptom that matches your problem
Copy the “Better AI ask”
Paste your real campaign data, notes, and recent changes
Review the signals AI checks
Let AI support the diagnosis — but keep humans in charge of budget, partners, and final decisions
01
02
03
04
05
25 UA AI Diagnostic Cards
Performance Problems
[ Growth symptom ]
[ Best tool ]
[ Weak AI ask ]
[ Better AI ask ]
Signals to check
[ Decision rule ]
CPA jumped in the last few days
ChatGPT
Why did CPA increase?
Separate this CPA increase into auction, creative, funnel, tracking, GEO mix, and source quality causes. Rank each cause by evidence and missing data.
CPI, CTR, CVR, target event rate, spend shift, GEO/ source mix, recent changes
Do not pause until tracking, volume,
and source/GEO
shifts are checked.
01
CPI increased but downstream quality improved
ChatGPT
Why did CPI get worse?
Check if higher CPI is actually a bad
enough to justify it. Compare cost, events, retention, and ROAS.
CPI, CPA, ROAS, retention, trial/ deposit/purchase rate
Do not optimize CPI alone if quality improved.
02
ROAS dropped after scaling
ChatGPT
Why did ROAS drop?
Compare pre-scale and post-scale
performance.
Identify whether the drop came from audience expansion, GEO shift, creative fatigue, bid pressure, or cohort lag.
Spend increase, ROAS by cohort, GEO, source, creative, frequency
Do not cut too fast if cohorts are still immature.
03
Installs increased but events dropped
ChatGPT
Why are users not converting?
Check whether the install growth came from lower-quality sources, new GEOs, different creatives, store mismatch, or tracking changes.
Install volume, event CVR, source mix, GEO mix, creative promise, event setup
Treat cheap volume as risky until downstream events prove quality.
04
CTR improved but CPA got worse
Claude + ChatGPT
Why is this ad not converting?
Check whether the install growth came from lower-quality sources, new GEOs, different creatives, store mismatch, or tracking changes.
CTR, CVR, CPA, store page message, onboarding, audience intent
Do not call a creative a winner only because CTR is high.
05
Volume suddenly dropped
ChatGPT
Why did volume drop?
Check whether the install growth came from lower-quality sources, new GEOs, different creatives, store mismatch, or tracking changes.
Spend, impressions, reach, frequency, bids, approvals, audience size
Check delivery and budget limits before changing creatives.
06
One GEO suddenly underperforms
ChatGPT
Why is this GEO bad?
Check whether the install growth came from lower-quality sources, new GEOs, different creatives, store mismatch, or tracking changes.
CPI, CPA, CVR, retention, language, creative angle, source mix
Do not kill a GEO before checking localization and sample size.
07
Creative Problems
[ Growth symptom ]
[ Best tool ]
[ Weak AI ask ]
[ Better AI ask ]
Signals to check
[ Decision rule ]
A winning creative is fading
ChatGPT + Claude
Is this creative tired?
Separate this CPA increase into auction, creative, funnel, tracking, GEO mix, and source quality causes. Rank each cause by evidence and missing data.
CTR trend, CPA trend, frequency, spend, audience, bid changes
Refresh before replacing if the core angle still works.
08
CTR is slowly declining
Claude
How do we improve CTR?
Check if higher CPI is actually a bad
enough to justify it. Compare cost, events, retention, and ROAS.
CTR by day, first 3 seconds, frequency, audience, format
Do not change the whole concept if only the hook is weak.
09
New creatives keep losing to old winners
Claude
Why are new creatives bad?
Compare pre-scale and post-scale
performance.
Identify whether the drop came from audience expansion, GEO shift, creative fatigue, bid pressure, or cohort lag.
Hook, promise, proof, pacing, CTA, audience fit
New creatives should evolve the winning logic, not abandon it.
10
Hook works in one GEO but fails in another
Claude
Localize this creative.
Check whether the install growth came from lower-quality sources, new GEOs, different creatives, store mismatch, or tracking changes.
GEO, language, cultural context, CTR, comments, store CVR
Do not localize words only; localize the reason to care.
11
Creative refresh made performance worse
ChatGPT + Claude
Why did the refresh fail?
Check whether the install growth came from lower-quality sources, new GEOs, different creatives, store mismatch, or tracking changes.
Old vs new metrics, script, visuals, audience, placement
Roll back if the refresh removed the original winning signal.
12
Many creatives get clicks but weak users
Claude + ChatGPT
Why are users low quality?
Check whether the install growth came from lower-quality sources, new GEOs, different creatives, store mismatch, or tracking changes.
CTR, event CVR, retention, reviews, store page, ad claim
Avoid scaling clicky creatives that damage downstream quality.
13
Market / Source Problems
[ Growth symptom ]
[ Best tool ]
[ Weak AI ask ]
[ Better AI ask ]
Signals to check
[ Decision rule ]
Cheap GEO brings weak users
ChatGPT
Is this GEO good?
Separate this CPA increase into auction, creative, funnel, tracking, GEO mix, and source quality causes. Rank each cause by evidence and missing data.
CPI, CPA, retention, purchase/deposit/trial rate, fraud signals
Cheap installs are not growth unless quality follows.
14
Partner traffic looks “too good”
ChatGPT
Is this partner good?
Check if higher CPI is actually a bad
enough to justify it. Compare cost, events, retention, and ROAS.
Install volume, event depth, retention, fraud, source transparency
Do not trust low CPI without downstream validation.
15
One source scales but quality drops
ChatGPT
Should we keep scaling this source?
Compare pre-scale and post-scale
performance.
Identify whether the drop came from audience expansion, GEO shift, creative fatigue, bid pressure, or cohort lag.
Spend curve, CPA, retention, sub-source, GEO, placement
Scale only the segments that preserve quality.
16
New market starts strong, then collapses
ChatGPT + Perplexity
Why did this market stop working?
Check whether the install growth came from lower-quality sources, new GEOs, different creatives, store mismatch, or tracking changes.
Reach, frequency, CTR, CPI, CPA, competition, localization
Do not assume early market performance is stable.
17
Fraud signals are unclear
ChatGPT
Check whether the install growth came from lower-quality sources, new GEOs, different creatives, store mismatch, or tracking changes.
CTIT, retention, event timing, device patterns, source split
Escalate suspicious patterns, but don’t accuse without evidence.
18
Data / Tracking Problems
[ Growth symptom ]
[ Best tool ]
[ Weak AI ask ]
[ Better AI ask ]
Signals to check
[ Decision rule ]
Metrics changed after app update
ChatGPT
Why did performance change?
Separate this CPA increase into auction, creative, funnel, tracking, GEO mix, and source quality causes. Rank each cause by evidence and missing data.
App release date, event firing, funnel CVR, campaign changes
Check product/tracking before blaming media.
19
MMP and ad platform disagree
ChatGPT
Which number is right?
Check if higher CPI is actually a bad
enough to justify it. Compare cost, events, retention, and ROAS.
Attribution windows, time zones, event definitions, SKAN, delays
Use one source of truth for decisions.
20
Attribution window makes channels look unfair
ChatGPT
Compare these channels.
Compare pre-scale and post-scale
performance.
Identify whether the drop came from audience expansion, GEO shift, creative fatigue, bid pressure, or cohort lag.
Click/view windows, conversion lag, channel type, goal
Do not compare channels without attribution context.
21
Event volume is too low to decide
ChatGPT
Should we pause this?
Check whether the install growth came from lower-quality sources, new GEOs, different creatives, store mismatch, or tracking changes.
Spend, installs, events, days live, variance
No budget decision without minimum data threshold.
22
Decision Problems
[ Growth symptom ]
[ Best tool ]
[ Weak AI ask ]
[ Better AI ask ]
Signals to check
[ Decision rule ]
Team wants to scale after one good day
ChatGPT
Should we scale?
Separate this CPA increase into auction, creative, funnel, tracking, GEO mix, and source quality causes. Rank each cause by evidence and missing data.
Days live, event volume, variance, ROAS lag, budget jumpDays live, event volume, variance, ROAS lag, budget jump
One good day is not a scaling signal.
23
Team wants to pause after one bad day
ChatGPT
Should we pause?
Check if higher CPI is actually a bad
enough to justify it. Compare cost, events, retention, and ROAS.
3–7 day trend, volume, changes, source mix, tracking
One bad day is not a pause signal.
24
AI gives a confident recommendation
Claude
Is this recommendation right?
Compare pre-scale and post-scale
performance.
Identify whether the drop came from audience expansion, GEO shift, creative fatigue, bid pressure, or cohort lag.
Missing data, assumptions, confidence, business context
Never accept AI confidence as evidence.
25
Bonus: 25 Copy-Paste AI Asks for UA Teams
[ Copy-paste AI ask ]
The UA AI Diagnostic Playbook
25 mobile growth problems, 100+ checks, and AI prompts that help you find what actually changed
Most UA teams don’t need more AI prompts. They need a faster way to understand what actually changed. CPA jumped. ROAS dropped. A winning creative started fading. A cheap GEO brought users who never converted. This playbook gives you 25 UA diagnostic cards to separate real signals from noise — before you pause, scale, refresh, or report.
Final Checklist
Before you change budget, pause a source, or call a creative a winner
AI will not replace UA judgment.
But it can make the first diagnosis faster, cleaner, and easier to challenge.
Use this playbook to ask better questions, catch weak signals earlier, and avoid acting on noisy data.
Is the data volume big enough?
01
Are date ranges consistent?
02
Did tracking or attribution change?
03
Did budget, bids, or audience change?
04
Did GEO/source mix change?
05
Is there cohort lag?
06
Are we optimizing the right KPI?
07
Did AI explain its logic?
08
Did a human UA owner approve the final action?
09
Did AI show what evidence is missing?
10
Final Checklist
Before you change budget, pause a source, or call a creative a winner
Check this first:
-
Is the data volume big enough?
-
Are date ranges consistent?
-
Did tracking or attribution change?
-
Did budget, bids, or audience change?
-
Did GEO/source mix change?
-
Is there cohort lag?
-
Are we optimizing the right KPI?
-
Did AI explain its logic?
-
Did AI show what evidence is missing?
-
Did a human UA owner approve the final action?
Use promo code AI in the form below to claim your $1,000 bonus for your next UA test.
Creative Problems
Performance Problems
[ Growth symptom ]
[ Best tool ]
[ Weak AI ask ]
[ Better AI ask ]
Signals to check
[ Decision rule ]
A winning creative is fading
ChatGPT + Claude
Is this creative tired?
Identify whether this looks like creative fatigue or a media buying change. Compare early vs recent performance and suggest refresh options that keep the winning angle.
CTR trend, CPA trend, frequency, spend, audience, bid changes
Refresh before replacing if the core angle still works.
08
CTR is slowly declining
Claude
How do we improve CTR?
Diagnose whether the opening hook, visual pattern, offer, audience saturation, or format is losing attention. Suggest 5 refresh directions.
CTR by day, first 3 seconds, frequency, audience, format
Do not change the whole concept if only the hook is weak.
09
New creatives keep losing to old winners
Claude
Why are new creatives bad?
Compare new concepts against the old winner. What core promise, emotion, format, or proof point did the new versions lose?
Hook, promise, proof, pacing, CTA, audience fit
New creatives should evolve the winning logic, not abandon it.
10
New creatives keep losing to old winners
Claude
Why are new creatives bad?
Compare new concepts against the old winner. What core promise, emotion, format, or proof point did the new versions lose?
Hook, promise, proof, pacing, CTA, audience fit
New creatives should evolve the winning logic, not abandon it.
11