Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.utari.ai/llms.txt

Use this file to discover all available pages before exploring further.

Ad Creative Diagnosis helps JeremyAI make the creative call before writing more ads. Instead of acting like a generic ad copy generator, this worker looks at the offer, audience, funnel, proof, creative format, and performance data, then identifies what is most likely constraining the ad right now. Use it when you want JeremyAI to answer questions like:
  • Is this really a hook problem, or is the issue proof, mechanism, lead quality, offer-message fit, or the funnel?
  • What should we keep from a winning ad before making variants?
  • Why is this ad getting cheap clicks but poor buyers?
  • Is performance down because of creative fatigue, audience mismatch, or a downstream conversion problem?
  • What should the next ads, scripts, hooks, image concepts, or creative briefs actually test?
The worker should leave you with a clear operating answer: what is wrong, what to keep, what to change, what to test next, and what data would change the diagnosis.

Watch the walkthrough

Use the worker walkthrough to see how to choose Ad Creative Diagnosis, give JeremyAI the right ad and funnel context, and read the creative diagnosis it returns.

When to use Ad Creative Diagnosis

Use this worker when:
  • an ad is spending but the lead quality is weak
  • a winner is starting to fatigue and you need a refresh path
  • you have a winning ad and want controlled variants without killing what works
  • you have losing ads and need to know whether the hook, body, proof, format, CTA, or funnel is the issue
  • you need UGC scripts, founder-led scripts, image-ad concepts, primary text, hooks, or creative briefs
  • cold traffic is not converting and you need to decide between direct response and content-first education
  • CTR looks good but booked calls, shows, sales, or customer quality are weak
  • you need to diagnose whether cheap leads are actually the wrong leads
  • you are not sure whether to create more ads or fix the VSL, webinar, landing page, booking flow, or sales process first
Use it whenever the next creative move could affect spend, lead quality, or campaign direction.

Example use cases

Ad Creative Diagnosis can handle creative review, iteration, production, and testing questions. Common use cases include:
  • Winning ad iteration: identify why the winner works, preserve the core mechanism, and create controlled variants.
  • Losing ad diagnosis: decide whether the ad failed from attention, belief, proof, offer fit, CTA, format, or downstream mismatch.
  • Creative fatigue: interpret rising frequency, falling CTR, rising CPC/CPL, and stable funnel conversion to decide whether to refresh creative.
  • UGC and short-form scripts: create native scripts without inventing fake customer stories or unsupported results.
  • Image-ad concepts: define the visual, overlay text, primary text, headline, CTA, and reason each concept should work.
  • Offer-message fit: compare the ad promise to the page, VSL, webinar, call, checkout, or actual delivery.
  • Content-first strategy: decide when the audience needs education before a direct-response ask.
  • Retargeting ads: write ads that prove the next step, such as the call, webinar, audit, demo, or checkout.

Get the strongest diagnosis

JeremyAI can give a better creative diagnosis when it has the actual ad context. You do not need perfect data, but you should bring as much of this as you can:
  • The offer: what you sell, price point, target customer, and delivery model.
  • The next step: VSL, webinar, booked call, checkout, DM, lead magnet, application, or purchase.
  • The creative: current ad copy, script, screenshot, image brief, video transcript, or hook.
  • The audience: cold, warm, retargeting, broad, niche, in-market, or needs-convinced.
  • The metrics: spend, CPM, frequency, CTR, CPC, CPL, CPA, ROAS, cost per 3-second view, cost per 15-second view, hold rate, booked calls, shows, sales, or lead quality.
  • The proof: case studies, testimonials, screenshots, claims, objections, average results, ranges, caveats, or compliance constraints.
  • The decision: whether you want to scale, refresh, write variants, kill an ad, change the funnel, or test a new angle.
Ask JeremyAI to:
  • diagnose the most likely creative constraint
  • say what is probably not the issue
  • identify what to keep from the current ad
  • produce the next creative assets
  • explain how to test them
  • state what data would change the diagnosis
The goal is not to get more random ads. The goal is to know what the next creative move should be and why.

Use it in rounds

You do not need to solve the whole account in one prompt. Use Ad Creative Diagnosis in rounds:
  1. First pass: give the ad, offer, audience, funnel destination, and whatever metrics you have.
  2. Evidence pass: add screenshots, Meta CLI output, CRM notes, call notes, VSL metrics, landing-page data, or sales feedback.
  3. Creative pass: ask for hooks, scripts, primary text, image concepts, or variants based on the diagnosis.
  4. Testing pass: ask how to structure the test and what metric should decide the winner.
  5. Follow-up pass: come back after new data and ask whether the diagnosis changed.
This is especially useful when the first answer says the problem is not actually creative. The worker may tell you to fix the VSL, funnel, qualification, follow-up, or sales process before writing more ads.

Before you start

You can start with a manual prompt. The worker does not need every integration connected to be useful. For a stronger diagnosis, bring:
  • the ad or creative you want diagnosed
  • the offer type and delivery model
  • the audience and platform
  • the funnel destination
  • current performance data
  • downstream quality data when available
  • proof or claims you are allowed to use
  • the exact output you want, such as hooks, scripts, image ads, or a test plan
Do not paste access tokens, passwords, private keys, or other secrets into the chat. Ad Creative Diagnosis should never ask for those secrets. If it does, stop and contact support.

Start an ad creative diagnosis

  1. Open JeremyAI.
  2. Go to Workers.
  3. Select Ad Creative Diagnosis.
  4. If you want the worker to use connected data, confirm the relevant integrations are connected or have screenshots, exports, or key numbers ready.
  5. Click Start Chat.
  6. Paste the ad, creative, metrics, and offer context.
  7. Ask the worker to diagnose the constraint and tell you what to test next.
If Ad Creative Diagnosis is not visible, refresh JeremyAI and confirm you are in the correct workspace. If it still does not appear, try opening JeremyAI in an incognito or private browser window. If the worker appears there, clear your browser cache and cookies for JeremyAI, then sign in again. If it still does not appear after that, contact support or your Utari team contact.

Start with a strong creative prompt

A strong prompt gives JeremyAI the offer, audience, creative, numbers, and decision you need to make. Use this structure: I need an ad creative diagnosis for [business / offer].

Offer:
[What we sell, price point, target customer, and delivery model]

Funnel destination:
[VSL / webinar / booked call / checkout / DM / lead magnet / application / purchase]

Audience:
[Cold / warm / retargeting / broad / niche / in-market / needs-convinced]

Current creative:
[Paste ad copy, hook, script, image concept, video transcript, or screenshot notes]

Performance:
- Spend:
- CPM:
- Frequency:
- CTR:
- CPC:
- CPL / CPA:
- Cost per 3-second view:
- Cost per 15-second view:
- Booked calls / shows / sales:
- Lead quality notes:

Proof and claims we can use:
[Testimonials, screenshots, exact results, ranges, caveats, credibility proof, mechanism proof, or say "no proof yet"]

Decision:
[Should we scale, refresh, kill this, write variants, test a new angle, change the funnel, or diagnose lead quality?]

Please diagnose the most likely creative constraint, tell me what to keep, tell me what to change, create the next assets to test, and explain what data would change your diagnosis.\
Shorter prompts can still work. Good short prompts include:
  • “This ad gets cheap leads but nobody buys. Is this a hook problem or a lead quality problem?”
  • “This winner is fatiguing. What should we refresh without killing what works?”
  • “Give me five controlled variants of this winning ad without changing the core mechanism.”
  • “Write three UGC scripts for this offer, but do not invent customer stories or results.”
  • “Create five image-ad concepts and leave claim slots where we need real proof.”
Avoid asking only for “more hooks” when you have performance data. If the hook is already earning attention, the problem may be proof, mechanism, offer fit, page fit, or lead quality.

Example prompt

Use this format when you have one winning ad and need stronger variants: I need an ad creative diagnosis.

Offer: high-ticket fitness coaching for busy dads.
Funnel: ad to training, then booked call.
Audience: cold Meta traffic.

Winning ad:
"If you are a busy dad who has tried keto, fasting, and 5am workouts but still cannot stay consistent, this is for you. We built a 4-day plan that works around your calendar instead of asking you to become a different person. Watch the training and see how it works."

Performance:
- CPL is stable
- close rate is solid
- no major fatigue yet

I want 10 new variants, but I do not want random ads. Diagnose why this is working, tell me what not to change, then write controlled variants.

If any variant needs proof, a result, a time commitment, scarcity, or a delivery detail I did not provide, put it in brackets instead of inventing it.\
The worker should respond with a direct creative call, not just a list of ads: Your winner is working because it names the failed attempts, validates the busy-dad reality, and positions the mechanism as a plan that fits his life instead of forcing him to become a different person. Do not touch that core mechanism.

The variants should change one major variable at a time: the pain angle, objection, proof direction, or emotional frame.\
Then it should create usable variants and include claim slots where needed, such as: Variant: Social Proof Angle
"If you are a busy dad who has tried keto, fasting, and 5am workouts but still cannot stay consistent, this is for you. [Insert real number] dads have finally stuck with a plan because it was designed for their calendar, not against it. Watch how it works."\

What the worker should give you

A useful Ad Creative Diagnosis answer should include:
  • the likely creative constraint
  • the evidence or assumption behind the call
  • what is probably not the issue
  • what to keep from the current ad
  • what to change first
  • the next assets to test
  • the testing structure
  • the metrics that should decide the winner
  • any claim slots or proof needed before launch
  • what data would change the diagnosis
If the data is thin, the worker should still give a provisional call. It should state its assumptions, give the most useful first-pass direction, and ask only the few questions that would actually improve the next answer. Useful follow-up prompts include:
  • “What are you assuming that could be wrong?”
  • “Which number would most change this diagnosis?”
  • “What should I not change yet?”
  • “Is this a creative issue or a funnel issue?”
  • “Turn this into a 7-day creative test plan.”
  • “Rewrite the variants with claim slots instead of invented proof.”
  • “Give me the mechanism version because we do not have result proof yet.”

Proof and claim slots

Ad Creative Diagnosis should not invent proof. If you have not supplied a number, result, timeframe, sample size, delivery detail, scarcity claim, guarantee, testimonial, or operational detail, the worker should either leave it out or put it in brackets. Good claim slots look like:
  • [insert real number]
  • [insert verified result]
  • [insert actual timeframe]
  • [insert real proof point]
  • [insert actual time commitment]
  • [insert true scarcity reason]
  • [insert real customer story]
  • [insert verified before/after]
This matters most for proof-first, social-proof, urgency, price, retargeting, specificity, and UGC-style ads. If you do not have result proof yet, ask for mechanism-led ads instead: We do not have verified customer results yet. Use mechanism proof, demonstration proof, or credibility proof. Leave any result claims as bracketed slots.\ For UGC-style scripts, provide a real customer story if you want testimonial-style copy. Otherwise, the worker will use founder, educator, creator, or demo-style scripts without inventing results. Write three UGC-style scripts for a $97 AI ad-auditing training for coaches and agencies.

Speaker mode: founder/educator.
Do not write fake first-person customer results.
Do not describe this as software or done-for-you. It is a training that teaches the diagnostic framework and gives prompts.\

Connect useful data sources

Connect or provide the smallest set of source-of-truth systems needed for the creative decision. You do not need to connect every app. Useful sources include:
  • paid traffic data: Meta CLI output, platform exports, screenshots, or campaign notes
  • creative assets: ad copy, scripts, video transcripts, image concepts, thumbnails, or creative folders
  • funnel analytics: landing-page conversion, VSL watch data, webinar registration, checkout, or booking data
  • CRM and pipeline data: lead quality, show rate, close rate, sales notes, buyer quality, or refunds
  • call and meeting data: sales call notes, objections, transcripts, or recordings
  • proof sources: testimonials, screenshots, case studies, customer messages, dashboard exports, or documented results
  • manual tracking: Google Sheets, Airtable, Google Docs, Notion, or uploaded exports
If multiple apps cover the same function, use the one your team treats as the source of truth.

Meta CLI data

For Meta ad data, use Meta CLI output for now. Useful fields include:
  • campaign, ad set, and ad names
  • spend
  • impressions
  • frequency
  • CPM
  • CTR
  • CPC
  • leads
  • CPL
  • conversions
  • CPA
  • ROAS when available
  • cost per 3-second view
  • cost per 15-second view
  • hold rate or thumb-stop metrics when available
  • booked calls, shows, sales, or lead quality by ad when available
For creative diagnosis, downstream quality matters. Cheap leads do not automatically mean good ads, and high CTR does not automatically mean qualified demand.

How the worker stays focused

Ad Creative Diagnosis is configured to stay focused on the creative decision. It should not drift into a broad business audit unless the data shows the ad is not the real constraint. The worker should:
  • diagnose before writing
  • preserve what is working
  • avoid random creative testing
  • test one major variable at a time when possible
  • separate testing from scaling
  • call out when the issue is funnel, lead quality, sales, or offer-message fit
  • use proof slots instead of invented claims
  • preserve the offer type and delivery model
  • give a practical test plan after creating assets
You do not need to configure the worker before using it. Start with the ad context, add the numbers or screenshots you have, and ask for the creative call.

Triggers

Triggers let Ad Creative Diagnosis run automatically. They do not change the worker itself or connect new data sources. Use a trigger only when the task is repeatable. Good trigger use cases include:
  • weekly creative fatigue review
  • weekly winning-ad iteration review
  • recurring review of Meta creative performance
  • creative surplus planning before fatigue hits
  • retargeting creative review after a campaign update
For one-off diagnosis, use Start Chat instead. For a scheduled trigger:
  1. Click Create new.
  2. Choose Scheduled Trigger.
  3. Select Ad Creative Diagnosis as the worker.
  4. Give the trigger a clear name and optional description.
  5. Set the schedule and timezone.
  6. Choose the execution method.
  7. Write the trigger instructions.
  8. Create the scheduled task.
Example scheduled trigger prompt: Run an ad creative diagnosis for the last 7 complete days.

Use the connected Meta data or pasted Meta CLI output if available.

Identify signs of creative fatigue, lead quality issues, hook problems, proof/mechanism problems, offer-message mismatch, or funnel mismatch.

For each active ad, say whether to keep, refresh, pause, or use as a control.

Create 3 to 5 controlled next-test ideas for the highest-priority creative constraint.

If data is missing, say what is missing and give the best provisional diagnosis from the available data.\
Before turning on a trigger, run the same instruction manually in chat. After the trigger is active, check the first few runs and pause it if the output is incomplete, too broad, or using the wrong data.

If the answer is not useful yet

You usually do not need support to improve a weak creative diagnosis. Give the worker better context and ask it to be more decisive. Try prompts like: Pick one creative constraint. Do not give me a list.\ What should I keep from this ad before changing anything?\ Use the metrics I gave you. Is this hook, proof, lead quality, or funnel fit?\ Do not invent proof. Rewrite these with bracketed claim slots.\ This is a workshop, not software. Rewrite the ads so the delivery model is accurate.\ If the ad is not the real problem, say that directly and tell me what to fix first.\ If the worker cannot access a source directly, paste the key numbers, upload an export, or share a screenshot with sensitive information hidden.

When to contact support

Contact support if:
  • Ad Creative Diagnosis is not visible after refreshing, confirming you are in the correct workspace, trying an incognito or private browser window, and clearing browser cache and cookies
  • a connected integration is not available to the worker after reconnecting it
  • a trigger is enabled and configured correctly but does not run
  • you cannot tell whether the worker should have access to a specific data source
When sending screenshots, hide access tokens, private account IDs, customer data, payment information, customer emails, and other sensitive information. The Workers walkthrough shows how focused JeremyAI workers can be created and configured. If you are using the preconfigured Ad Creative Diagnosis community worker, you do not need to recreate it.