How to write a cold-traffic ad creative brief and 3 ad scripts with ChatGPT or Claude
Pick one market, segment and avatar from ads that already sell. Then fill in the 17 reasons a stranger buys (the content, hope, value and impulse sales), and have ChatGPT or Claude turn that brief into 3 ad scripts at 3 risk levels that are ready to film. You end with a saved ad brief, 3 scripts and a source for every claim.
From the playbook Builds a cold-traffic ad brief and 3 ad scripts that make the 4 sales by Christian Martin ✓ · Updated Nov 19, 2026
What you need
- The offer the ad is for: name, what's inside it (parts, modules, ingredients or features) and how it's delivered
- Your price options and your guarantee, plus any real numbers such as your refund rate
- Any real reason to buy now: a bonus, a bundle, a deadline, limited spots or stock, or a price going up
- Your own ads that worked, or your best-selling page, if you have them
- Access to the Meta Ad Library (free) to find competitor ads that have run a long time
- Buyer words: reviews, testimonials, support emails, sales-call notes or Reddit threads where your buyers talk
- A ChatGPT or Claude chat you can keep open for the whole job
The steps
Load your premises
Paste your offer facts into the chat: brand, offer name, what's inside, attributes (how it's made or delivered, how long results take, who it's not for, any limits), price options, guarantee, what buyers gain by buying now and what they lose by waiting. Ask the AI to list these 8 premises back to you and to point out anything that's missing. Check one thing: if you have no real reason to buy now, it should write [real reason to buy now] as a blank and not make up a deadline.
Pick one market
Ask the AI for 3 to 5 markets this offer could serve, each named by a desire and not a product category ("owners who want more booked calls", not "owners who want better software"). Then do the research yourself. Look up your own best ads and pages, or search the Meta Ad Library for competitors in each market and note which ads have run longest. Paste what you found and ask it to pick ONE market: the one your best sellers speak to or, if the offer is new, the one with the most long-running competitor ads. Cut any market with fewer than 3 competitors.
Pick the segment and build the avatar
Paste the winning ads, reviews and buyer words you collected. Ask for the segment, meaning people who share the desire plus a condition, identity or behavior, and tell it to copy a segment that already works unless you want a bet. Then ask for the avatar in 2 layers. Psychographics: primary desire, problem, secondary desires, pain points, dreams, fears. Identity: demographics, behavior (which ads, videos, news and trends they follow) and perspective (beliefs, enemies, allies, their words). Check that each line says where it came from. Read it and ask yourself, "Is this the person I want to sell to?" Fix it before going on, because the whole brief rests on this pick.
Find the mechanism (elements 7 to 10)
Ask for 4 things. Point of contact: what the avatar already blames, and what the winning ads in this segment blame. Problem mechanism: the real cause they're most ready to accept, built on what they already believe. Function mechanism: the action that has to happen to fix it, and why what they already tried didn't do it. Solution mechanism: what does that action, and why it fixes the problem at the root. Push back if it teaches a brand-new idea when it could connect dots the avatar already believes.
Fill the content elements (1 to 6)
Ask for 5 to 10 topics the avatar cares about right now (searches, news, trends) and pick one. Then ask for the format (the look of reference ads in this market, the narrative style, the characters), the concept (a one-sentence logline plus the angle), the copy hook and the visual hook. Tell it to start the copy hook from a hook structure that already works in your market and fill it with the avatar's own words. Then ask for identification lines that make the avatar think "that's me".
Fill the offer elements (11 to 14)
Ask for the conditions: why the avatar can't get the mechanism on their own or from the cheap option. Then ask for the USP, written as proof, one condition at a time, that your offer meets each one. For features, tell it to keep only those that move one of 4 levers: odds (why it works this time), dream, effort or time. Any feature that moves none of them stays out. For anchoring, ask for the price per day or per result next to what living with the problem costs in money, feelings and time.
Fill the close (15 to 17)
Ask for risk reversal that shows why your guarantee is more than enough, for example a guarantee period longer than the time results take, and not just "we have a guarantee". Then ask for what they gain now, option by option, and what they lose by waiting, using only the real deadlines, spots, stock or price changes from your premises.
Review the brief table
Ask for all 17 elements in a table with 3 columns: element, the bullet points the ad will say, and the source. Guesses should be marked [guess]. Ask it to name the 2 or 3 weakest elements. Add what you know that makes them stronger, such as a customer story, a real number or a result, and have it update the table.
Pick a reference ad and split it
Find a long-running winning ad in your market, either your own or a competitor's from the Meta Ad Library, in the format from your brief. Paste its transcript or describe it. Ask the AI to split it into mini-blocks by the job each one does: hook, identification, belief break, mechanism reveal, product, offer, close. The reference sets only the order and rhythm. What gets said comes from your brief.
Write 3 ads at 3 risk levels
Ask for 3 scripts. Conservative: proven market, segment, mechanism and concept, with only the hook or format changed. Moderate: 1 or 2 bets, such as a new concept or a new look. Aggressive: several bets (new topic, format and concept). For each one, get the bets it makes, the visual hook for the first 3 seconds, the copy hook word for word, the full script with element numbers such as [8 Problem mechanism], scene-by-scene visuals and characters, the call to action and where the link goes, the length (60 to 120 seconds), and the primary text and headline. Check that the 3 ads differ in what they bet on, not just in the first line.
Run the claims check
Ask the AI to list every number and claim in each ad (spoken, in the text or on screen) with its source. Any number with no source becomes a marked [guess] or a blank for you to fill. Then check each ad: the 4 sales come in order (content, hope, value, impulse), the price and product come after the mechanism, every element is either in the script or left out with a stated reason, and no blank is ever a spoken line.
Save the brief for the next batch
Copy the target, the 17-element table, the 3 scripts and the list of open blanks into one document called "Cold-traffic ad brief: [offer name]". Keep the market, segment, avatar, mechanism, conditions, USP and buyer words where you'll paste them into the next chat, so your next ad batch starts from this brief and not from zero.
A prompt to start with
You are a creative strategist who writes cold-traffic ads. A cold-traffic ad doesn't sell the product. It makes 4 sales before the product: the Content Sale (they watch to the end), the Hope Sale (they see what stopped them before, the action that fixes it and the mechanism that does that action), the Value Sale (this product is the only place to get that mechanism under the right conditions, at a price far below what the problem costs) and the Impulse Sale (waiting is riskier than buying). Each sale is made with reasons, the 17 elements: 1 Format, 2 Topic, 3 Concept, 4 Visual hook, 5 Copy hook, 6 Identification, 7 Point of contact, 8 Problem mechanism, 9 Function mechanism, 10 Solution mechanism, 11 Conditions, 12 USP, 13 Features (each must move odds, dream, effort or time), 14 Anchoring, 15 Risk reversal, 16 What they gain now, 17 What they lose by waiting. My premises: - Brand: [brand name] - Offer: [offer name] - What's inside: [parts, modules, ingredients, features] - Attributes: [how it's made or delivered, how long results take, who it's not for, limits] - Price options: [prices and bundles] - Guarantee: [guarantee and any real number, e.g. refund rate] - Gain by buying now: [bonuses, bundles, extras, or "none"] - Lose by waiting: [real deadline, spots, stock, price rise, or "none"] What already sells: [my winning ads, best pages, or competitor ads that have run a long time in the Meta Ad Library] Buyer words: [paste reviews, testimonials, support emails, Reddit threads] Work in this order and stop where I say: 1. List 3 to 5 markets (each named by a desire, not a product), pick ONE based on what already sells, then pick a segment and build the avatar (psychographics + identity + their words). Say where each pick came from. STOP and ask me: "Is this the person you want to sell to?" 2. After I confirm, fill in all 17 elements as the bullet points the ad will say, in a table with a source next to each. Mark anything unsupported as [guess]. Name the 2 or 3 weakest elements and ask me what I know that would make them stronger. STOP. 3. After I answer, write 3 ad scripts: conservative (only the hook or format changes), moderate (1-2 bets) and aggressive (several bets: new topic, format, concept). For each one, give the bets, the visual hook, the copy hook word for word, the full script with element numbers in brackets, scene-by-scene visuals, the CTA, the length (60-120 sec), the primary text and the headline. Product, price and offer come only after the mechanism. Never make up a deadline, number or result, and never write a blank as a spoken line. 4. Under each ad, list every number and claim with its source.
Paste it into ChatGPT or Claude and put your own facts in the [brackets].
An example of the finished work
COLD-TRAFFIC AD BRIEF: Calm Leash Course (Steady Paws, a made-up online dog-training business)
PREMISES
Brand: Steady Paws. Offer: Calm Leash Course, a 21-day online program for leash-reactive dogs. Inside: 21 daily video lessons, a printable distance ladder, 2 video reviews of your walk by a certified trainer, and a members' forum. Attributes: 10-15 minutes a day, no prong or shock collar, most owners see calmer walks in 3-6 weeks, not for dogs with a bite history (they need an in-person trainer). Price: $197 one time, or 3 payments of $69. Guarantee: 60 days, full refund. Gain now: a bonus live Q&A for this month's group. Lose by waiting: enrollment closes on the 30th and the next group opens in [date: owner to confirm].
TARGET
Market: dog owners who want calm, normal walks. (Source: 4 of Steady Paws' 5 longest-running ads and its top page speak to this.)
Segment: owners of adopted rescue dogs who bark and lunge at other dogs. (Source: 3 competitors in the Meta Ad Library have run rescue-dog ads for 6+ months.)
Avatar: a woman, 30-50, US, lives in a suburb or apartment, adopted her dog 1-3 years ago, walks at 5 a.m. to avoid other dogs, follows rescue pages and dog-training creators on Instagram. Her words: "he's not aggressive, he's scared," "reactive," "over threshold." Fear: she'll have to give him back, or he'll hurt someone. (Source: 112 reviews and 2 Reddit threads in r/reactivedogs.)
BRIEF (the 17 elements, short form)
1 Format: a trainer talking to camera in a park, with B-roll of a real walk (from 2 reference ads)
2 Topic: "Why your rescue barks at dogs but not people" (a top Reddit question)
3 Concept: the trainer walks a reactive dog past another dog and counts out loud how far away the dog stays calm. Angle: "He's not protecting you. He was already over the edge before he saw that dog."
4 Visual hook: a dog lunging at the end of a leash, then a freeze-frame with a red ring around the owner's tight grip on the leash
5 Copy hook: "This is what happens to your rescue when you walk him at 5 a.m. to avoid other dogs" (a proven structure in this market)
6 Identification: crossing the street, walking in the dark, apologizing to strangers
7 Point of contact: "something happened to him before I got him" (owners blame his past; competitor ads say the same)
8 Problem mechanism: stress builds up over the day, so by the walk he's already near his limit, and one dog tips him over
9 Function mechanism: lower his stress before the walk, then teach a new response at a distance where he can still think. Corrections hide the warning but leave the stress, and treats fail because an over-the-limit dog won't eat.
10 Solution mechanism: the Threshold Walk: 3 quiet "decompression" days, then a distance ladder that closes the gap a few steps at a time
11 Conditions: needs daily structure, a distance you track, and someone who spots the early signs you miss
12 USP: a 21-day plan (structure) + a printable ladder (tracking) + 2 trainer video reviews (someone who spots the signs)
13 Features: odds = trainer reviews; dream = coffee-shop walks by week 6; effort = 15 min a day; time = 3-6 weeks, because built-up stress takes weeks to come down
14 Anchoring: $197, or about $9 a day over the 21 days, next to $100-150 per private session [guess: owner to confirm local rates], 5 a.m. walks and the guilt
15 Risk reversal: the 60-day guarantee covers nearly 3 times the 21-day plan
16 Gain now: the live Q&A bonus for this group only
17 Lose by waiting: enrollment closes on the 30th
Weakest spots: 14 (local trainer rates are a guess), 10 (needs a customer result) and 17 (the next open date isn't set).
ADS
Ad 1, Conservative ("The 5 a.m. walk"): proven concept and format, new hook only. 75 seconds. Primary text: "If you walk your rescue in the dark so you won't see other dogs, read this." Headline: "Calm walks in 21 days, no prong collar."
Ad 2, Moderate ("Count the steps"): new concept, the live distance count on camera. 90 seconds.
Ad 3, Aggressive ("Street test"): new topic (the dog-park debate), new format (street interviews with owners) and new concept. 110 seconds.
CLAIMS CHECK (Ad 1)
"3-6 weeks": premises. "60-day guarantee": premises. "$100-150 a session": [guess]. "No prong collar": premises. "Enrollment closes the 30th": premises.
Open blanks: next group's date, local trainer rates, one named customer result.
The business in this example is made up.
Or hand the whole job to the playbook
Christian Martin wrote these steps as a playbook on Iggy. It runs in your own ChatGPT or Claude, and Iggy gives it the facts of your business, so you do not explain them each time. In the free webinar, we set Iggy up together, in real time.
Mistakes to avoid
- Asking the AI to "write me an ad" with no briefBuild the 17-element brief first. The brief decides what gets said, and a reference ad decides only the order and rhythm. Without the brief, you get the average ad on the internet.
- Opening with the product, the price or the offerMake the content and hope sales first: hook, identification, then the mechanism. Bring in the product, price and offer only after the avatar believes there's a real cause and a real fix.
- Naming the market by product categoryName it by a desire: "owners who want calm walks", not "people who want a dog-training course". Then pick the one market your best sellers or competitors' longest-running ads already speak to.
- Guessing the avatar from your headBuild it from winning ads, reviews, Reddit threads and support emails, with a source next to each line. Spend the most time on their behavior, meaning what they watch, read and follow, because that's where new hooks and concepts come from.
- Teaching cold traffic a brand-new theoryStart from what they already blame (the point of contact), and give a problem mechanism that proves that suspicion right. Connect dots they already believe.
- Listing every featureKeep only features that move one of 4 levers: odds, dream, effort or time. A feature that moves none of them stays out of the ad.
- Making up urgency or numbersUse only real deadlines, spots, stock or price changes. If there's none, leave a blank and don't say it in the script. Every number, including ones shown on screen, needs a source or a [guess] mark.
- Writing 3 versions that differ only in the hookMake the 3 ads differ in what they bet on. Conservative changes only the hook or format, moderate makes 1 or 2 bets (concept or look), and aggressive bets on a new topic, format and concept.
Questions and answers
What is an ad creative brief?
It's one page that says who the ad talks to and what it says, before anyone writes a script. For cold traffic, this guide's brief has 3 parts: the target (one market, one segment, one avatar) and the 17 elements of ad psychology, each with a source. The scripts are then written from the brief.
What are the 4 sales in a cold-traffic ad?
The content sale (they want to watch to the end), the hope sale (they see the real cause, the fix and the mechanism that does it), the value sale (your product is the best or only place to get that mechanism, at a price far below what the problem costs) and the impulse sale (waiting is riskier than buying). They come in that order, all before the price.
Can I use this ad brief template for a service, coaching or software?
Yes. "What's inside" becomes the parts the buyer gets, such as sessions, modules or features. "Attributes" becomes how it's delivered and how fast results come. For B2B, the avatar's demographics become role, company size, revenue and industry.
Where do I find ads that already work in my market?
Start with your own best ads and pages. For competitors, search the free Meta Ad Library by brand or keyword and look for ads that have run for months, because brands keep paying for ads that sell. Ads from past years count too, since the avatar's pains and fears hardly change.
What if I have no real reason for people to buy now?
Leave [real reason to buy now] as a blank and leave elements 16 and 17 out of the script, with a note saying why. A fake deadline hurts trust. Consider adding a real one, such as a bonus for this month's buyers or a group that closes on a date.
How long should a cold-traffic video ad be?
About 60 to 120 seconds is common for this kind of ad. That's enough time to make all 4 sales. The first 3 seconds (the visual hook and copy hook) decide whether anyone stays.
Which of the 3 ads should I run first?
Run the conservative one first, since it reuses what already works. A scaling brand runs a few conservative ads, a lot of moderate ones and some aggressive ones. When one bet wins, it pays for the ones that didn't.
Do I have to redo the whole brief for every new batch of ads?
No. Save the market, segment, avatar, mechanism, conditions, USP and buyer words. In the next batch, reuse them and spend your time on new topics, formats, concepts and hooks, where the bets pay off.