The hook is the single line that determines whether someone stops scrolling or keeps going. On Facebook and Instagram, you have roughly 1-3 seconds to earn attention. If the hook fails, the rest of your ad — the copy, the creative, the offer — never gets seen. AI tools like ChatGPT can generate dozens of high-quality hook variations in minutes, but only if you prompt them with the right frameworks and real customer data.
Below is the complete playbook: five proven hook frameworks, the exact prompts to generate each one, 20+ ready-to-use templates, real ecommerce examples of good vs. bad hooks, and an A/B testing system to find your winners fast. If you're spending money on ads, this is where your ROAS improves first.
Why the Hook Matters More Than Anything Else
Every ad is a funnel. The hook is the top. If nobody stops scrolling, your click-through rate tanks, your cost per click rises, and your CPA climbs regardless of how good your landing page is.
Meta's algorithm reinforces this. Ads with higher engagement rates (driven by hooks that stop the scroll) get shown to more people at lower costs. A strong hook doesn't just get attention — it actively reduces your ad costs by signaling to the algorithm that people want to see your content.
The problem: writing hooks is hard. Most ecommerce founders default to generic openers like "Tired of [problem]?" or "Introducing [product name]" — hooks that feel invisible in a feed full of content fighting for the same eyeballs. AI changes the equation by letting you generate and test at volume.
The 5 Hook Frameworks That Work for Ecommerce Ads
Every effective ad hook falls into one of these five categories. Each triggers a different psychological response. The best ad accounts test across all five frameworks, then double down on what works for their audience.
| Framework | Psychological Trigger | Best For | Example |
|---|---|---|---|
| Problem-Agitate | Pain recognition | Products that solve a frustration | "Your kitchen sponge is a bacteria farm. Here's what to use instead." |
| Curiosity Gap | Information gap theory | Unique or unexpected products | "The reason your coffee tastes bitter has nothing to do with the beans." |
| Bold Claim | Pattern interrupt | Products with strong differentiation | "This $12 serum outperformed a $180 department store brand in blind tests." |
| Social Proof | Herd behavior / trust | Products with real customer traction | "47,000 dog owners switched to this food last month. Their vets noticed." |
| Before / After | Transformation desire | Visual result products (skincare, fitness, home) | "My garage went from disaster to showroom in one weekend. $89 total." |
Framework 1: Problem-Agitate Hooks
How it works: Name a specific pain point your customer already feels, then twist the knife by making it feel more urgent or more disgusting than they realized. The product becomes the relief.
Why it stops the scroll: People are wired to pay more attention to threats and problems than to benefits. A problem-agitate hook hijacks that instinct.
ChatGPT Prompt for Problem-Agitate Hooks
Copy this prompt and replace the bracketed inputs with your product details:
You are a direct-response copywriter for ecommerce Facebook ads. Generate 10 problem-agitate ad hooks for [PRODUCT]. The target customer is [TARGET CUSTOMER] who struggles with [PAIN POINT]. Each hook should: (1) name a specific, visceral problem in the first line, (2) make it feel worse or more urgent than the reader realized, (3) be under 15 words, (4) use conversational language — no corporate buzzwords. Here are real customer complaints about existing solutions: [PASTE 3-5 NEGATIVE REVIEWS OF COMPETITORS].
Key detail: Pasting real customer complaints is what makes the output specific. Without them, you get generic hooks. With them, the AI picks up real language and real frustrations.
Framework 2: Curiosity Gap Hooks
How it works: Reveal just enough to create an open loop — a question the reader can't answer without engaging with the ad. The gap between what they know and what they want to know is what holds attention.
Why it stops the scroll: The brain treats open loops like unfinished tasks. It wants closure. A curiosity gap hook creates a micro-obsession that can only be resolved by watching or reading the ad.
ChatGPT Prompt for Curiosity Gap Hooks
You are a direct-response copywriter. Generate 10 curiosity gap hooks for a Facebook ad selling [PRODUCT] to [TARGET CUSTOMER]. Each hook should: (1) hint at a surprising fact, counterintuitive insight, or little-known cause without revealing the answer, (2) be under 15 words, (3) make the reader think "wait, what?" Product benefits: [LIST 3-5 KEY BENEFITS]. Common misconceptions in this space: [LIST 2-3 THINGS PEOPLE GET WRONG].
Framework 3: Bold Claim Hooks
How it works: Make a statement so specific or surprising that the reader stops to evaluate whether it's true. Bold claims work because they create a pattern interrupt — the brain expects generic ad speak and gets something concrete instead.
Important: Bold claims must be truthful and substantiable. The hook earns the click; the ad body and landing page prove the claim. Unsubstantiated claims destroy trust and can violate ad platform policies.
ChatGPT Prompt for Bold Claim Hooks
You are a direct-response copywriter for ecommerce ads. Generate 10 bold claim hooks for [PRODUCT]. Target customer: [TARGET CUSTOMER]. Each hook should: (1) make a specific, surprising, and truthful claim about the product, (2) include a concrete number, comparison, or timeframe where possible, (3) be under 15 words, (4) feel like something you'd stop scrolling to verify. Here are real product differentiators: [LIST WHAT MAKES YOUR PRODUCT DIFFERENT]. Here are real results customers report: [PASTE 3-5 POSITIVE REVIEWS WITH SPECIFIC OUTCOMES].
Framework 4: Social Proof Hooks
How it works: Lead with evidence that other people — preferably a lot of them — already trust, use, or love the product. Humans look to others when uncertain. Social proof hooks shortcut the evaluation process.
ChatGPT Prompt for Social Proof Hooks
You are a direct-response copywriter. Generate 10 social proof hooks for a Facebook ad selling [PRODUCT] to [TARGET CUSTOMER]. Each hook should: (1) lead with a number, a group, or a specific customer outcome, (2) feel like a news headline about real demand, (3) be under 15 words. Real data to use: [NUMBER OF CUSTOMERS, UNITS SOLD, REVIEWS, STAR RATING, REORDER RATE, WAITLIST SIZE, or any other traction metric you have].
Only use real numbers. Fabricated social proof is easy to spot and violates ad platform policies. If you don't have impressive numbers yet, use qualitative proof: "My dermatologist asked what I changed" is social proof without needing 50,000 customers.
Framework 5: Before / After Hooks
How it works: Show the gap between the customer's current state and the state your product delivers. The bigger and more specific the gap, the more compelling the hook.
ChatGPT Prompt for Before / After Hooks
You are a direct-response copywriter for ecommerce. Generate 10 before/after hooks for [PRODUCT]. Target customer: [TARGET CUSTOMER]. Each hook should: (1) contrast a specific "before" state with a specific "after" state, (2) make the transformation feel achievable, (3) be under 15 words, (4) focus on the outcome the customer cares about, not the product features. Real customer results: [PASTE 3-5 REVIEWS DESCRIBING THEIR TRANSFORMATION OR RESULTS].
Know your numbers before you scale.
Better hooks mean more clicks — but clicks only matter if the unit economics work. Plug in your ad spend, AOV, and margins to see your real return on ad spend before scaling.
Open Free ROAS Calculator →20+ Hook Templates You Can Use Today
These templates work across niches. Replace the bracketed sections with your product specifics. Each template is tagged with its framework so you can test across categories.
| # | Framework | Template | Example (Skincare Brand) |
|---|---|---|---|
| 1 | Problem-Agitate | "Your [thing] is [disgusting/broken fact]." | "Your pillowcase is sabotaging your skin every night." |
| 2 | Problem-Agitate | "Stop [common habit]. It's making [problem] worse." | "Stop washing your face twice a day. It's making breakouts worse." |
| 3 | Problem-Agitate | "You're spending $[X]/month on [category] and still [failing]." | "You're spending $80/month on skincare and still waking up with dull skin." |
| 4 | Problem-Agitate | "If [common product] actually worked, you wouldn't need [number] of them." | "If moisturizers actually worked, you wouldn't need 6 of them." |
| 5 | Curiosity Gap | "The reason [problem] has nothing to do with [obvious cause]." | "The reason your skin looks tired has nothing to do with sleep." |
| 6 | Curiosity Gap | "[Profession] have known this about [topic] for years. Now you can too." | "Estheticians have known this about retinol for years. Now you can too." |
| 7 | Curiosity Gap | "I asked [expert] the #1 mistake in [category]. The answer surprised me." | "I asked a dermatologist the #1 skincare mistake. The answer surprised me." |
| 8 | Curiosity Gap | "There's one ingredient in your [product] that's doing the opposite of what you think." | "There's one ingredient in your moisturizer that's actually drying your skin." |
| 9 | Bold Claim | "This $[low price] [product] replaced my $[high price] [competitor]." | "This $14 serum replaced my $190 La Mer moisturizer." |
| 10 | Bold Claim | "[Result] in [short timeframe]. No [common requirement]." | "Visible glow in 5 days. No 12-step routine." |
| 11 | Bold Claim | "We tested [product] against [competitor]. [Specific result]." | "We tested our serum against 4 prestige brands. It scored highest on hydration." |
| 12 | Bold Claim | "[Number]% of customers reorder within [timeframe]." | "73% of customers reorder within 60 days." |
| 13 | Social Proof | "[Number] [people] switched to [product] last [timeframe]." | "12,000 women switched to this serum last quarter." |
| 14 | Social Proof | "[Star rating] stars across [number] reviews. Here's why." | "4.8 stars across 3,200 reviews. Here's why." |
| 15 | Social Proof | "My [person they trust] asked what I changed." | "My dermatologist asked what I changed. It was one product." |
| 16 | Social Proof | "Sold out [number] times. Back in stock for [duration]." | "Sold out 3 times. Back in stock for 48 hours." |
| 17 | Before / After | "Before: [specific bad state]. After: [specific good state]." | "Before: foundation to leave the house. After: bare skin confidence." |
| 18 | Before / After | "[Time period] ago I [bad state]. Now [good state]." | "3 months ago I wouldn't go out without makeup. Now I don't own any." |
| 19 | Before / After | "I went from [X] to [Y] in [timeframe] with [product]." | "I went from weekly breakouts to clear skin in 6 weeks with one serum." |
| 20 | Before / After | "My [thing] used to [bad]. Now it [good]. $[price] total." | "My skin used to flake by noon. Now it glows at midnight. $14 total." |
| 21 | Problem-Agitate | "Nobody talks about the [hidden cost/problem] of [common solution]." | "Nobody talks about the hidden irritants in 'clean' skincare." |
| 22 | Curiosity Gap | "The [product category] industry doesn't want you to know this." | "The skincare industry doesn't want you to know this about SPF." |
| 23 | Bold Claim | "One product. [Number] problems solved. $[price]." | "One serum. 4 skin problems solved. $14." |
| 24 | Social Proof | "[Specific person/role] recommended this. [Number] people listened." | "A TikTok esthetician recommended this. 50,000 people listened." |
Usage tip: Don't pick one template and hope it works. Pick 4-5 across different frameworks, generate variations with AI, and test them against each other. The data tells you which framework your audience responds to — then you scale that framework.
Good vs. Bad AI Hooks: Real Examples
The difference between a hook that stops the scroll and one that gets ignored usually comes down to specificity. Here are side-by-side comparisons across multiple ecommerce niches:
| Niche | Bad Hook (Generic) | Good Hook (Specific) | Why the Good One Works |
|---|---|---|---|
| Dog Food | "Tired of your dog's stomach issues?" | "Your dog's kibble has 4 ingredients banned in Europe." | Specific number + surprising fact = pattern interrupt |
| Coffee | "Love great coffee? Try ours!" | "Your coffee is stale. Here's how to tell in 3 seconds." | Challenges assumption + offers quick payoff |
| Fitness | "Get fit with our resistance bands!" | "A physical therapist designed these bands. Your gym doesn't sell them." | Authority + exclusivity |
| Skincare | "Achieve glowing skin with our serum." | "I replaced 4 products with this $14 serum. My aesthetician noticed." | Specific savings + third-party validation |
| Home | "Organize your home today!" | "I organized my entire pantry in 20 minutes. $35 total." | Specific time + specific cost = believable |
| Baby | "The best bottle for your baby." | "3 pediatricians switched their own babies to this bottle." | Expert social proof + personal stakes |
| Fashion | "Shop our new collection!" | "This $39 jacket got mistaken for Aritzia three times last week." | Specific comparison + social validation |
| Supplements | "Boost your energy naturally." | "I stopped drinking coffee for 30 days. This is what I used instead." | Personal story + curiosity about the alternative |
The pattern is clear: bad hooks use vague language ("great," "best," "amazing") and talk about the product. Good hooks use specific details (numbers, timeframes, named alternatives) and talk about the customer's experience. AI generates both kinds — your prompt determines which.
How to A/B Test AI-Generated Hooks
Generating 20 hooks is the easy part. Finding the winner requires a systematic testing process. Here's the framework that works for ecommerce brands running Facebook and Instagram ads:
Step 1: Generate Volume
Use the prompts above to generate 15-20 hooks across all five frameworks for a single product or campaign. Don't edit heavily at this stage — you want raw volume and variety.
Step 2: Score and Shortlist
Rate each hook on three criteria:
- Specificity (1-5): Does it include a concrete detail, number, or named reference? Generic = 1. Hyper-specific = 5.
- Emotional pull (1-5): Does it trigger curiosity, fear, desire, or surprise? Flat = 1. Visceral = 5.
- Brand fit (1-5): Does it match your voice and audience? Off-brand = 1. Sounds like your best customer = 5.
Take the top 5 scorers. Those are your test candidates.
Step 3: Build the Test
In Meta Ads Manager, create one campaign with one ad set. Inside that ad set, create 3-5 ads — identical creative, identical body copy, different hooks only. This isolates the hook as the variable.
- Budget: Allocate enough for each ad to get 1,000+ impressions (usually $10-$20/day per ad)
- Duration: Run for 3-5 days minimum before making decisions
- Metric to watch: Hook-through rate (for video) or CTR (for static). Then track downstream to CPA and ROAS
Step 4: Read the Data
After 3-5 days, rank your hooks by CPA (not CTR alone — a clickbait hook with high CTR but no conversions is worse than useless). Kill the bottom performers. Scale the top 1-2.
Step 5: Iterate on Winners
Take your winning hook and go back to ChatGPT:
This hook won our A/B test: "[WINNING HOOK]". Generate 10 variations that preserve the same angle and emotional trigger but use different wording, structure, or specifics. Keep them under 15 words.
Now test the variations against the original. This is how you find your "best possible version" of a winning angle rather than just the first version that worked.
Common AI Hook Mistakes (and How to Fix Them)
AI makes predictable mistakes when generating hooks. Know them in advance and you'll save hours of bad output:
- Generic pain points. "Tired of [problem]?" is the default AI opening. Fix: add a constraint in your prompt — "Do NOT start any hook with 'Tired of' or 'Are you'."
- Product-first language. AI defaults to describing the product instead of the customer's experience. Fix: tell the AI "focus on the customer's situation, not the product features."
- Buzzword soup. "Revolutionary," "game-changing," "premium quality" — words that mean nothing. Fix: add "no adjectives that can't be measured or proven" to your prompt.
- Too long. AI hooks often run 20-30 words. A scroll-stopping hook needs to land in under 15 words — often under 10. Fix: add a hard word count limit to every prompt.
- Missing the audience. If you prompt with just a product name, AI writes for a generic audience. Fix: always include target customer demographics, psychographics, and real language from their reviews.
The Prompt Structure That Produces the Best Hooks
Every effective hook prompt has the same five components. If you're getting weak output, one of these is missing:
| Component | What to Include | Why It Matters |
|---|---|---|
| Role | "You are a direct-response copywriter for ecommerce Facebook ads" | Sets the writing style and format expectations |
| Framework | Specify: problem-agitate, curiosity gap, bold claim, social proof, or before/after | Forces structured output instead of random copy |
| Customer data | Real reviews, real complaints, demographics, psychographics | Specificity in = specificity out. This is the #1 quality lever |
| Constraints | Word limit, banned phrases, tone requirements | Prevents generic filler and keeps hooks tight |
| Volume | "Generate 10-15 variations" | More options = better shortlisting. AI's best hook is rarely its first |
How to Use Hooks Across Ad Formats
The hook changes format depending on where it appears, but the framework stays the same:
- Video ads (Reels, Stories): The hook is the first 1-3 seconds — spoken, shown as text overlay, or both. If using AI for scripts, instruct it: "The first sentence must work as a standalone scroll-stopper even without audio."
- Static image ads: The hook is the headline text on the image and/or the first line of primary text. Test putting the hook on the image itself (not just in the caption) — in-image text gets seen first.
- Carousel ads: The hook is the first card. If someone doesn't stop on card one, cards two through five don't exist. Treat the first card as a standalone ad.
- Primary text (caption): The hook is the first line before the "See more" fold. On mobile, only the first ~125 characters of primary text are visible without expanding. Your hook must land within that window.
Scaling: From One Winning Hook to a Full Creative System
Once you find a hook framework that works for your audience, build a system around it:
- Week 1: Test 5 hooks across all 5 frameworks. Identify top 2 performers by CPA.
- Week 2: Generate 10 variations of each winner. Test variations against the originals.
- Week 3: Pair the winning hook with different creatives (video vs. static, different imagery, different CTAs). The hook stays constant; the creative changes.
- Week 4: Refresh. Even winning hooks fatigue. Go back to AI, generate a new batch using the same framework but different angles and details.
This cycle keeps your ROAS healthy by preventing creative fatigue — the silent killer of ad accounts. When CTR drops and CPA rises, it's almost always stale hooks and creatives, not audience exhaustion.
Tracking Hook Performance
The metrics that tell you whether your hooks are working:
| Metric | What It Measures | What "Good" Looks Like | Where to Find It |
|---|---|---|---|
| Hook Rate (video) | % who watch past 3 seconds | Above 30% | Meta Ads Manager → Video Engagement |
| CTR (all) | % who click the ad | Above 1.5% for link clicks | Meta Ads Manager → Performance |
| CPC | Cost per click | Varies by niche — lower is better relative to your baseline | Meta Ads Manager → Performance |
| CPA | Cost per purchase | Below your max CPA (AOV × margin) | Meta Ads Manager → Results |
| Thumb-stop ratio | 3-second views / impressions | Above 25% | Custom metric (calculate manually) |
The metric that matters most is CPA. A hook with a lower CTR but a lower CPA is a better hook — it's attracting fewer but more qualified clicks. Use our ROAS calculator to see how CPA improvements translate to actual profit.
Frequently Asked Questions
Can AI write ad hooks that actually convert?
Yes. AI tools like ChatGPT can generate dozens of hook variations in minutes using proven frameworks. The key is feeding the AI specific inputs — customer pain points, product benefits, competitor angles, and real review language — rather than generic prompts. AI handles volume and variation; you handle selection and brand voice editing.
What is the best AI tool for writing ad hooks?
ChatGPT (GPT-4) and Claude are the strongest general-purpose options. They produce better output than dedicated ad copy tools when given structured prompts with real customer data. Specialized tools like Jasper and Copy.ai offer templates but are less flexible for framework-driven hook generation.
How many hook variations should I test per ad?
Test 3-5 variations per ad concept. Generate 15-20 with AI, shortlist the top 5, then run them as separate ads in the same ad set. Give each at least 1,000 impressions and 2-3 days before making cut decisions.
What makes a good ad hook for Facebook and Instagram?
A good hook does one of five things in 1-3 seconds: calls out a specific pain point, creates a curiosity gap, makes a bold claim, leverages social proof, or shows a before/after contrast. The hook must be specific to your audience — generic openers get scrolled past. Specificity is what stops the thumb.
How do I prevent AI-generated hooks from sounding generic?
Feed the AI three things: (1) real customer reviews with their exact language, (2) specific product benefits with numbers, and (3) competitor angles to differentiate against. Then add constraints: "conversational language," "no buzzwords," "under 12 words." The more specific your inputs, the less generic the output.
Should I use AI hooks for video ads or just static image ads?
Both. For video, the AI hook becomes your opening line in the first 1-3 seconds. For static, it's your headline or first line of primary text. Video hooks need to be even shorter and more punchy. Generate for both formats and test which delivers the lower CPA for your audience.