Writing Desire-Led Ads with AI: The Complete Workflow and the 6 Mistakes That Kill Scaling

The research is done. The desire is identified. The market size is validated. The intensity is right. The seasonal timing is good.

Now comes the ad.

The goal of this article is to give you the complete workflow for turning that research into copy — and to document the six mistakes that undo all of it. Because the desire can be exactly right, and the creative process can still fail if it's built on bad habits around AI, bad habits around creative iteration, or bad habits around when to look for a new desire versus when to keep testing the current one.

Start with the workflow. Then read the mistakes section carefully — it's where most scaling efforts break down.

The role of AI in desire-led ad creation

AI handles the copywriting. The human handles the desire identification.

This distinction matters enormously. AI is excellent at generating multiple versions of copy once it knows the desire to target. It can produce 6 variations of a video script and 6 variations of an image headline in minutes. It can apply a structural framework from one category to a completely different product. It can iterate across desires fast.

What AI cannot do reliably is identify which desire to go after. That requires the research methods from Article 5 — survey responses, Reddit threads, TikTok view counts, competitor reviews. AI doesn't have access to that real-time data about what's actually resonating with real people. It can make educated guesses, but those guesses are based on training data, not on what your specific customers are actually saying right now.

You pick the desire. AI writes the copy. Never reverse this.

Asking AI to decide what desire to target is asking it to do the part it's worst at. Asking AI to write five versions of copy once you've already told it the desire is asking it to do what it's best at.

The complete workflow: TrendTrack to final copy

This is the full sequence, end to end.

Step 1: Pull the structural framework from TrendTrack

Go to TrendTrack and pull ads from a completely different niche than yours — not a competitor. Find their top-performing video ad and top-performing static image ad. The metric to use is longevity: if an ad has been running for two months or more, it's almost certainly performing. Brands don't keep spending on ads that don't work.

Capture the structure in plain text:

• How does the video open? (First 3 seconds)
• What problem or desire does it call out immediately?
• How does it build tension?
• When does the product enter?
• What's the call to action?

Do the same for the image ad:

• What's the headline?
• What desire or pain does it reference?
• What's the visual concept?

Write this out in a document before touching AI. You want to be able to articulate the structure clearly before asking AI to apply it.

Step 2: Give the structure and the desire to AI

The prompt to Claude or ChatGPT should include three things:

1. The structural framework you captured in Step 1 (describe it or paste the original ad)
2. Your product (describe it specifically — not just "a supplement," but exactly what it does, what format it comes in, what the mechanism is)
3. The specific desire you want to target (from your research)

Be explicit. "Here's the structure from this testosterone ad. Here's my product: a breath spray that eliminates bad breath at the source rather than masking it. Here's the desire I want to target: professional confidence — the fear of losing clients or opportunities because of breath."

Step 3: Generate the first version

AI produces the first version of the video script and the image headline. Review it for two things:

1. Does it actually target the desire you specified, or did it drift toward the product?
2. Does the structure hold? (Opening call-out, tension build, product introduction, call to action)

A concrete example, starting from Morris Men's testosterone ads and applying the structure to a breath spray product targeting the professional confidence desire:

Original testosterone video structure: "I almost started [major intervention] until I found this."

Breath spray version: "I lost a client and I didn't realize it was my breath until months later. As you get close, you shake hands, you lean in to close and if something's off, they don't tell you. They just don't call you back."

Original testosterone image headline: "Your grandfather had more testosterone at 60 than you at 35."

Breath spray image version: "Your mints have been lying to you for 5 minutes at a time."

Both versions carry the same structural DNA — the contrast, the indictment, the revelation — but speak to a completely different desire for a completely different product.

Step 4: Scale across desires

Once you have the first version, ask AI for six more versions, each targeting a completely different desire. Give it explicit desire labels so it doesn't just rephrase the same angle with different words.

Using the breath spray example, the same structural framework produces:

Desire: Social Anxiety / Self-Consciousness: "For years I covered my mouth every time I talked to someone. I thought that was just me. You start standing a little further away. Talking a little quieter. Turning your head. Nobody knows you're doing it but you."

Desire: Family Connection: "My grandmother told me my breath smelled like campfire. She wasn't wrong. I love my cigars. But you start noticing people lean back. Your wife. Your kids."

Three completely different desires. Three completely different audiences. The same product. The same underlying structure. All written in minutes once the framework and desires were given to AI.

Step 5: Rewrite manually where needed

AI doesn't always nail it perfectly. Read every output out loud. Anything that sounds manufactured, vague, or off-voice — rewrite it. The goal is for the copy to sound like something a real person would actually say to another real person, not like a marketing prompt. Specific, concrete, emotionally honest.

Remove any em-dashes AI tends to insert. Simplify any overworked phrases. Make sure the opening three seconds of each video version land hard.

Step 6: Brief your creative team with the desires, not just the scripts

When you brief your creative team, include the desire — not just the script. The team shooting or designing the creative needs to know what emotional state they're trying to capture, not just what words to say. The desired state should be visible in the visual, not just mentioned in the copy.

The 7-step framework for every ad you write

Every desire-led ad, regardless of format, follows this structure:

Step 1: Call out the current state — where the buyer is right now, the problem, the pain, what's wrong.

Step 2: Show the desired state — what they want their life to look like, who they want to be, what they want to achieve.

Step 3: Introduce the product as the bridge — how it takes them from where they are to where they want to be.

Step 4: Provide specific proof — a number, a result, a testimonial, a before/after.

Step 5: Handle the objection — what's the main reason someone wouldn't buy immediately? Address it.

Step 6: Visualize the desired state — don't just say it, show it in the visual component of the creative.

Step 7: Clear call to action — what do you want them to do right now?

Not every ad needs all seven steps at full length. A 15-second video might compress steps 1-3 into the first 8 seconds and use 5 and 7 in the last 7. A static image might just be steps 1 and 2 in the headline and the visual. But the underlying logic — current state, desired state, product as bridge — should be present in every format.

The 6 mistakes that kill scaling

Understanding the desire framework is half the job. The other half is avoiding the habits that undo it. These six mistakes are specific, common, and expensive.

Mistake 1: Going after a Must Be Nice desire

This is the most foundational error. Building an entire ad strategy around something people think would be nice to have — but don't urgently want — means fighting the buyer's psychology at every step.

The ads look fine. The creative is decent. The targeting seems right. But conversion rates are flat, CPAs are high, and nothing scales. The problem isn't the ad. It's that the desire behind the ad doesn't have the intensity to produce action.

The fix: Before finalizing any desire to go after, score it on the intensity framework from Article 3. If it's Level 1 or Level 2, find the Level 3 or Level 4 version of it before spending.

Mistake 2: Ignoring seasonal relevance

Running a dry skin product hard in July. Running weight loss ads in October with no urgency lever. Running back-to-school ads in November.

The desire exists in the abstract, but the seasonal intensity isn't there right now. The buyer you're targeting has this desire as a background thought, not an active urgency. You're asking them to buy something they won't need for months.

The fix: Check Google Trends before scaling any desire. Understand when it peaks, build toward that window, and hold back when the seasonal timing is wrong. Covered fully in Article 4.

Mistake 3: Thinking one desire is enough

Finding one desire that works and never testing another. This is comfortable but limiting. The ceiling on any single desire is the size of the market for that desire. Once you've captured the willing buyers in that pool, growth stalls.

The unlock — the one that took the paint by numbers brand from $130,000 to $1.1 million a month — came from finding a completely different desire for the same product. Not a variation on the same desire. A different one entirely.

The fix: Keep a desire backlog. Every desire you identified in research but haven't tested yet goes on the list. When growth stalls, you don't restart the research process from scratch — you pull the next desire from the backlog and test it.

Mistake 4: Copying competitor ads without understanding the desire

Seeing a competitor ad that looks like it's performing — high engagement, comments, lots of activity — and ripping the creative structure without understanding what desire it's built on.

The creative might perform for the competitor because they've already built brand recognition in that desire space. Or because their product serves that desire better than yours does. Copying the format without understanding the underlying desire positioning produces ads that look similar but don't carry the same relevance.

The fix: When you see a competitor ad that's performing well, first ask: what desire is this built on? Is it Must Be Nice, Annoyance, Keeps Me Up at Night, or Bleeding Neck? Is there a different desire you could go after with better positioning? Use TrendTrack for inspiration on structure, but bring your own desire research to determine what you actually say.

Mistake 5: Letting AI pick the desire

Asking Claude or ChatGPT: "What desire should I target for this product?"

AI will give you an answer. It will sound confident. It will be plausible. It will be wrong in ways you can't verify without real market research.

AI's training data doesn't include your customers' recent survey responses. It doesn't include the Reddit thread from last month where 80 people described their pain in visceral detail. It doesn't include the TikTok view counts showing which desire is currently capturing massive attention. It makes inferences from general knowledge — which is useful for many things, but not for desire identification.

The fix: Do the research first. Identify the desire yourself. Then tell AI exactly what desire you want to go after, and let it write the copy. The research methods in Article 5 are the inputs. AI is the execution layer. Keep those roles clear.

Mistake 6: Changing creatives when you should be changing desires

"Our ads aren't working. We need new creatives."

This is the default response to declining performance — and it's often wrong. If you've tested multiple creative formats, hooks, and angles for the same desire and none of them are working, the problem probably isn't the creative. It's the desire.

New creatives for a wrong desire produce new results that look just like the old results. The CPA doesn't move. The ROAS doesn't improve. Because the constraint isn't execution — it's positioning.

The fix: Before briefing a new round of creatives, ask: has this desire been genuinely maxed out? Have we been at the same revenue level for six or more months despite good creative testing? If yes, this is a desire audit moment. Pull up the desire backlog. Find the next one to test.

When to look for a new desire (the exact triggers)

Four specific signals that indicate it's time to stop iterating and start finding a new desire:

You've been stuck at the same revenue for six or more months. Not declining — stuck. Stable revenue despite active creative testing is a sign that you've saturated the desire.

CPA keeps climbing no matter what you do. You've tried new hooks, new formats, new audiences. CPA just climbs. This is the marginal buyer problem — you've already gotten the easy buyers in this desire space, and the remaining ones are harder and more expensive to reach.

Google Trends shows the desire hitting a seasonal low. This one is temporary — hold back now and come back when the season turns. But if you're stuck and the desire is also at a seasonal low, the timing is doubly wrong.

Multiple strong creative concepts for the same desire have underperformed. Not one bad test. Multiple genuine attempts — different angles, different hooks, different emotional territories within the desire — and none of them moved the needle. The desire itself is the constraint.

When any two of these are true simultaneously, stop creative testing and start desire research.

The complete desire-to-ad process: all steps together

For reference, the full sequence from start to finish:

1. Break down your product using Feature → Performance → Benefit → Desire (do this manually)
2. List every desire your product can speak to — don't filter yet, aim for 8-10
3. Research each desire using the 6 methods: customer surveys, YouTube/TikTok view counts, Reddit threads, competitor reviews, TrendTrack
4. Score each desire on market size, current intensity, and seasonal relevance
5. Prioritize by starting with the desire that scores highest across all three
6. Pull a structural framework from TrendTrack in a completely different niche
7. Brief AI with the framework, the product, and the specific desire — let it generate copy across multiple versions
8. Rewrite manually anywhere the AI output sounds vague, over-produced, or off-voice
9. Build and launch the creative with the desire visible in both the copy and the visual
10. Monitor for the six mistake patterns — if scaling stalls, diagnose before adding more creative spend

The desire is the foundation. Everything else — creative format, targeting, campaign structure, budget — is built on top of it. Get the desire right and everything downstream becomes easier. Get it wrong and no amount of optimization fixes it.

The checklist

• Never ask AI to pick the desire — that's your job, done through research
• Always brief AI with three things: structural framework, product description, and specific desire
• Generate at least 6 versions per desire, each with explicit desire labels so they don't all end up sounding the same
• Rewrite AI output manually anywhere it sounds like a marketing prompt instead of a real person
• Maintain a desire backlog — every validated desire you haven't tested yet, ready to pull when scaling stalls
• Diagnose before briefing new creatives — ask whether the problem is creative execution or desire saturation
• Run the 6-mistake audit before each new creative cycle: right desire? right intensity? right season? right structure?
• Show the desired state visually, not just in copy — the visual should make someone see themselves in the right side of the bridge
• Stop creative testing when two or more stall signals are present simultaneously — shift to desire research instead

← [Back to Playbook Index](README.md)