The best hooks are discovered, not invented.
The hooks that convert come from the exact words your customers used to describe their problem before any marketer touched their thinking. Not paraphrases. Not summaries. The verbatim phrases real people reach for when they explain why they downloaded your app, what it changed for them, and what life looked like before it.
The jobs-to-be-done methodology calls these “golden nuggets.” We use the same term. The hard part used to be finding them.
Hannah Parvaz covered the qualitative research method in her episode of the Mobile User Acquisition Show. Her approach: schedule 1:1 customer interviews, listen for the word that keeps repeating, follow it to the underlying motivation. One year of that work at Curio, an audio journalism app, kept surfacing the word “seem.” Users didn’t want to be more educated. They wanted to seem more interesting. That became the line “Become the most interesting person in the room.” Regardless of origin or attribution, that is what a golden nugget does when you find one.
The methodology was right. In practice, running it properly requires someone whose full-time job is customer research. A UA manager with a testing roadmap to ship cannot schedule 20 interviews, transcribe them, run affinity mapping sessions, and still be producing creative by end of sprint. Most teams try it once, get one useful insight, and then never find the time to do it again.
AI changes that entirely.
Page Contents
The feel test
Before any workflow, a quick experiment. Read these two sets of hooks for the same cashback app and notice what you feel:
| Brief-room hook | Golden nugget hook |
|---|---|
| Save money every time you shop. | I opened my app at checkout and saw $4.80 back. I almost cried. |
| Every purchase gives back. | My husband thought I was making it up. I showed him the deposit. |
| Your spending, rewarded. | I get paid to go to Target. I get paid to go to Walmart. I get paid to exist. |
The first column could belong to any cashback app, any country, any year. A branding team wrote it. It is grammatically clean, logically correct, and completely interchangeable with every competitor in the category.
The second column lands differently. “I almost cried” is not something a copywriter invents. “My husband thought I was making it up” requires a specific moment: a conversation, a phone screen held up, a specific deposit amount. “I get paid to exist” is the kind of escalation that only happens when someone is genuinely delighted and speaking without a filter.
When someone sees a hook built from language like theirs, they don’t feel marketed to. They feel recognized. That recognition is the difference in thumb-stop rate.
Generic AI copy is the average of everything written about the category. Golden nuggets are specific because they come from one person.
The AI workflow for finding golden nuggets
Your customers have already written their golden nuggets. They are sitting in your App Store reviews, your Google Play reviews, your support tickets, and your survey responses. The job is not to generate them. It is to retrieve and cluster them, and AI makes that fast enough to run every creative sprint.
Step 1: Source raw reviews
Pull 200-500 reviews from three sources: your own App Store and Google Play listings, your top two or three competitors’ listings, and any in-app survey responses or support tickets where users describe why they downloaded or why they stayed.
Include recent negative reviews too. Friction language is often the most evocative: the exact description of a user’s before-state contains the emotional charge that makes a before-after hook land.
Step 2: Feed into AI, cluster by Emotion, Context, and Audience
Paste the reviews into Claude, or any capable model, with this prompt:
You are helping me find golden nuggets: the verbatim customer phrases that reveal real emotional motivation, not surface-level satisfaction.
Read through all these reviews carefully. Then do the following:
1. CLUSTER by three dimensions:
Emotion: What emotional state was driving the download? Look for anxiety, frustration with an alternative, aspiration, a life event (first job, move, new relationship), or shame about the before-state. Name the emotion precisely, not generically.
Context: What was happening in their life when they downloaded? What had they tried before? What problem were they trying to solve?
Audience: What does their language tell you about who they are? Their life stage, their relationship to the category, their self-image.2. FOR EACH CLUSTER, extract:
The 5 most verbatim phrases that capture why they downloaded or what changed for them. Copy their exact words. Do not paraphrase, summarize, or clean up their grammar.
Any phrase that describes a specific physical behavior, object, or ritual from their before-state. These are your highest-value nuggets.
Any phrase that describes how they felt before the app existed in their life.3. FLAG with an asterisk any phrase that passes this test: could only one real person have said this? If yes, it is a golden nugget.
Do not write theme summaries. Do not interpret motivations. Do not suggest hook ideas. Return only the clusters, their labels, and the verbatim phrases.
The Emotion, Context, and Audience dimensions map directly to the creative brief structure we use for all mobile ad creative briefs. Running the clustering through the same lens means the golden nuggets slot straight into brief inputs with no translation step.
Step 3: Extract verbatim phrases from each cluster
From each cluster, the verbatim phrases are your brief. Not the AI’s summary. Not a theme label. The words a real person wrote.
Here is what this looked like for a fintech budgeting app. One cluster of reviews about manual tracking produced the phrase: “I used to print out a spreadsheet and keep it in my pocket.” That sentence is more brief-ready than any positioning document. It is specific (printed, pocket, not an app), behavioral (a real workaround someone actually did), and it puts you inside the problem. No brief room produces that line. You find it or you don’t have it.
The phrase became a hook. The campaign ran the “before” visual of someone unfolding a paper spreadsheet at the checkout counter. The specificity of the image came entirely from the customer’s words.
Step 4: Map golden nuggets to hook formats
Once you have verbatim phrases clustered by Emotion, Context, and Audience, map each phrase to a core hook format. The brief writes itself.
| Golden nugget phrase | Emotion cluster | Hook format |
|---|---|---|
| “I get paid to go to Target. I get paid to exist.” | Delight / escalating vindication | Direct benefit, scales as “I get paid to ___” |
| “I used to print out a spreadsheet and keep it in my pocket” | Frustration / relief | Confession / before-after |
| “I almost cried at the checkout” | Surprise / disproportionate delight | Moment of discovery |
| “My husband thought I was making it up” | Vindication / social proof | Skeptic-to-believer arc |
What we see: Golden nugget phrases carry their hook format inside them. The emotional register is already present. The format is the container you choose, not the emotion you invent.
What we learn: The brief is a retrieval problem, not a creative problem. The insight already exists in your reviews. AI makes the retrieval fast enough to run every sprint instead of once a year.
Why the second column feels different
There is a test for whether a phrase is a genuine golden nugget or polished copy that approximates one. Ask: could only a real person say this?
“Save money on shopping”: any model trained on the internet generates this automatically. It is the average of everything said about the category.
“I used to print out a spreadsheet and keep it in my pocket”: no model generates that unprompted. It requires the memory of a specific person doing a specific workaround on a specific day. The physical detail (printed, folded, pocket) is the tell. Generic AI copy is smooth because it is an average. Golden nuggets have texture because they come from one person’s life.
That texture is what makes the hook stop a thumb. When a user sees it, they don’t consciously notice the specificity. They just feel the recognition: “That’s me.” That feeling is the metric you’re building toward, and it is the part that cannot be fabricated. For more on the role of AI in creative at scale, see how to scale creative production with AI without losing quality.
Running the sprint in practice
This is a two-hour sprint, run once per creative cycle or before entering a new audience segment.
| Sprint step | Time | Output |
|---|---|---|
| Pull 200+ reviews (own app + top 3 competitors) | 30 min | Raw review corpus |
| AI clustering prompt (Emotion, Context, Audience) | 15 min | Structured clusters |
| Extract 3-5 verbatim phrases per cluster | 20 min | Golden nugget bank |
| Map phrases to hook formats | 20 min | Brief: 10-15 hook candidates |
| Write hook variants from golden nuggets | 30 min | Production-ready hook set |
The output replaces your brief. The hooks go directly to the creative strategist for production, anchored in real customer language rather than invented positioning. If you are running dynamic creative optimization, a golden nugget bank is also the cleanest source for variant generation: each phrase is a different emotional entry point, which means your DCO variants test genuinely distinct hypotheses rather than surface-level copy swaps.
Frequently asked questions
What are golden nuggets in mobile ad creative?
Golden nuggets are the verbatim phrases customers use to describe their problem or desire before any marketer has shaped their thinking. The term comes from the jobs-to-be-done methodology. In mobile ad creative, golden nuggets produce hooks with unusual specificity because they come from real user experience, not a brief room.
How do you find golden nuggets for mobile app ads?
The fastest method is AI-assisted review clustering. Pull 200-500 App Store and Google Play reviews from your app and top competitors. Paste into an AI model, ask it to cluster by emotional state, context, and audience, and extract the five most verbatim phrases per cluster. Those phrases are your golden nuggets.
What is the difference between a brief-room hook and a golden nugget hook?
Brief-room hooks are invented and generic. Golden nugget hooks are specific because they come from one person’s lived experience. The test: could only a real person say this? “Save money on shopping” is the category average. “I used to print out a spreadsheet and keep it in my pocket” is one person’s actual workaround. That specificity creates recognition, not persuasion.
Can AI generate golden nuggets?
No. AI clusters and surfaces golden nuggets from your review corpus. It cannot fabricate the raw material. The physical specificity comes from real users. AI’s role is retrieval and clustering, not invention.
How often should we run a golden nugget sprint?
Once per creative cycle. A full sprint takes about two hours and produces 10-15 hook candidates ready for production.
What is the Emotion, Context, and Audience framework?
It is the clustering framework RocketShip uses in all creative briefs. Emotion is the user’s state when they downloaded. Context is what they were doing. Audience is who they are. Clustering reviews through this lens means golden nuggets map directly to brief inputs with no translation step.
Related reading
