Ad creative brief generator: AI prompts that turn reviews into Meta and TikTok briefs

Every paid social agency has the same Tuesday. A client's ads are fatiguing, the media buyer wants new creative by Friday, and the strategist is three tabs deep in Trustpilot, copying review lines into a Google Doc and scrolling the Meta Ad Library for whatever the competitors are running this week.

That research is the most important part of the job. It's also the part that eats the week, and it's why most agencies cap how many clients one strategist can carry.

So we built a blueprint for an AI ad creative brief generator that does the digging for you. You feed it a landing page, competitor ads and customer reviews. It hands back personas, whitespace angles, persona-first hooks, concepts by format, copy variants, a UGC shot list and a testing matrix with ad names ready for Ads Manager.

Get the full blueprint: the AI Ad Creative Brief Generator pro prompt includes the complete Lovable master prompt (23,000 characters), the build steps, the tool setup, pricing guidance and a cold pitch you can send to agencies today.

Why creative briefs are the real bottleneck

Media buying got automated. Advantage+ campaigns, broad targeting and Meta's own delivery system now do most of what a buyer used to do by hand with interest stacks. What's left for the agency to win on is the creative: which angle, which hook, which format, for which person.

And creative only works when it's built on real customer language. The best hooks usually start in a 2-star review where someone wrote "I just wanted it to arrive before I ran out", and a strategist noticed no other brand in the category talks about running out.

Finding that line takes hours. You read a few hundred reviews. You tag pains and objections. You open every competitor's Ad Library page, note which ads have been running for months (long-running ads are a strong hint something's working), and sort them by angle in your head. Then you write the brief.

Agencies charge real money for this. Industry pricing for performance creative retainers in 2026 runs roughly $5,000 to $8,000 a month at a starter tier and $12,000 to $15,000 at scale, according to Darkroom's agency cost breakdown. The strategy work inside that retainer is where the margin leaks out, because it doesn't scale with headcount the way editing does.

What the creative brief generator does

The tool is a web app built in Lovable, run by the agency once per client campaign. A strategist works through five short input steps and then hits "Build brief". Here's what happens on each side of that button.

The inputs

The offer. Paste the landing page URL and the app reads it with Firecrawl, then pulls out the offer, prices, guarantee, main call to action and proof points. Each proof point keeps the exact sentence it came from, so the strategist can see the AI didn't make it up.

Competitor ads. This is where most tools get vague. Meta's official Ad Library API is free, but for ordinary commercial ads it only returns ads that were delivered to people in the EU or UK. That's a consequence of the EU's Digital Services Act. A US-only competitor will usually come back empty.

The blueprint handles that head on. When the API returns zero ads, the app shows a plain-English explainer and jumps the strategist to a manual tab. Paste ad copy, or drag in screenshots from the public Ad Library website, and Claude's vision model reads each image into body text, headline, CTA, visible offer and a format guess. You also need to confirm your identity at facebook.com/ID before Meta grants API access, which takes a few days, so the build steps tell you to start that first.

Reviews. Paste them, or upload a CSV from Trustpilot Business, a Google Business Profile export or any spreadsheet. Tag each review as the client's own or a named competitor's. Competitor 1 to 3 star reviews are gold here: that's where you find the complaints no brand in the category is answering. On import, the app strips surnames, emails and phone numbers.

Audience notes and guardrails. Known segments, things the client refuses to say, compliance flags, platforms, campaign goal, and any past winning ads. Past winners become the control to beat in the test matrix.

The pipeline

Then it runs nine steps, each retryable on its own if something fails.

It mines the reviews for pains, desires, objections, trigger moments and outcomes, and every item carries at least one verbatim quote. It classifies each competitor ad into one of 14 angle families (problem agitation, social proof, mechanism, risk reversal, founder story and so on) and weights ads that have run 30+ days as likely winners.

Then the bit agencies care about most: the whitespace map. A heatmap grid shows angle families against competitors, so you can see at a glance that three brands are all shouting about price while the guarantee angle sits empty. Each angle gets a whitespace score, high when customers talk about it a lot and competitors rarely touch it.

From there it clusters customers into up to 4 personas (you can change that), named by situation and trigger. "The owner who just lost a weekend to admin" tells a copywriter far more than "women 35 to 44".

Hooks that pass a real strategist's sniff test

This is where most AI ad tools fall over. They write ten hooks that are all the same sentence wearing different hats, and every one of them could run on a competitor's page unchanged.

The creative brief generator blueprint bakes in four rules that came from running real Meta campaigns, and a second AI pass checks every hook against them.

1. Name the persona early. The first line has to name or clearly describe who the ad is for within the first 8 words. "Run a dog grooming van?" beats a vague pain statement every time. The right people self-select, and Meta's delivery system gets a clear text signal about who should see the ad. The app stores the word position so you can sort by it.

2. Stay inside Meta's personal attributes policy. Meta won't let an ad assert or imply that the viewer has a sensitive attribute: health, finances, weight, religion and so on. "Run a gym?" is fine. "Struggling with your weight?" gets rejected. The checker flags anything close and rewrites it around the situation or the product. A human still signs off, but you stop wasting review cycles.

3. Vary the structure. Hooks are spread across at least 6 of 15 structures (persona question, customer quote, objection flip, cost of doing nothing, POV scene, myth versus reality and more), with no more than 2 of any one. That gives the buyer a spread worth testing.

4. The competitor test. For every hook, the checker asks: could a direct competitor run this exact line? If yes, it gets rewritten around something structurally specific to this brand, like a guarantee term, a pricing structure or a phrase only this brand's customers use. Each hook shows a one-line note on what makes it ownable.

Every hook also links to the review quote or proof point it came from. If a number appears in a hook, it has to exist in the inputs. No invented stats, ever.

From hooks to a brief the team can shoot

For each persona, the app writes concepts in four formats: static, UGC, talking head and carousel. Statics get a layout description and alternate headline treatments. UGC gets a 20 to 40 second timestamped script. Talking heads get a speaker type, setting and B-roll cutaways. Carousels get 4 to 7 frames where frame 1 stops the scroll and frame 2 pays it off.

Each concept carries "why this should work" and "what would kill it". That second line is the one strategists end up loving, because it's the risk they'd raise in a creative review anyway.

Copy comes in three lengths. The short version stays under 125 characters so it shows without "See more" on most placements. You also get 5 headlines under 40 characters and 3 descriptions.

Then the UGC creator brief, written so a freelancer can shoot it without a call: creator profile, deliverables, a numbered shot list with duration and framing, dos and don'ts, and a reminder to agree paid usage rights in writing. Creators are always asked to film 3 alternative hook openings, so the buyer can test hooks without paying for new shoots. That matters when UGC averages somewhere around $150 to $212 per video according to Influee's 2026 rate guide, and experienced creators charge a lot more.

Finally, the testing matrix. One cell per persona and angle, each holding a concept and 3 to 5 hook variants, a one-line hypothesis, a primary metric, and kill and scale rules the buyer edits to suit the budget. Ad names follow a BRAND_PERSONA_ANGLE_FORMAT_HOOK_v1 convention, so when something wins in Ads Manager you can trace it straight back to the review quote that started it.

A worked example: Northfield Coffee Roasters

The blueprint seeds a demo brand so you can see the whole flow on first load. Here's how a brief for a subscription coffee brand like that plays out (illustrative, not a real client).

The strategist pastes the landing page. The app pulls out a monthly subscription, free shipping, a "skip or cancel any time" policy and a freshness promise that beans ship within 48 hours of roasting.

She adds three competitors. Two have run ads in the UK, so the API pulls around 60 ads between them. The third is US-only and comes back empty, so she drops in 5 screenshots from the Ad Library website and the vision model fills in the fields.

She pastes 180 reviews: 110 for Northfield, 70 for competitors. The confidence meter goes green.

The review mining surfaces something sharp. A cluster of competitor complaints says the same thing in different words: "stale by week two", "tastes flat after the first bag". And Northfield's own 5-star reviews keep mentioning the roast date printed on the bag.

The whitespace map confirms it. All three competitors are running price and convenience angles. Zero are running freshness as a mechanism. Whitespace score: high.

Personas come out as "The home espresso nerd who weighs their shots", "The office manager who restocks the kitchen" and "The gift buyer who wants it to feel special". Each one is a situation, with no age band in sight.

For the espresso persona, one hook reads: "Weigh your shots? Check the roast date on your beans." It names the persona in 3 words, it's built on a phrase lifted from the reviews, and a competitor can't run it without a printed roast date of their own. It passes the competitor test.

The UGC concept that comes with it has a home barista holding a bag to camera, tapping the roast date, then pulling a shot. The shot list runs 7 shots across 30 seconds, with 3 alternate openings to film. The test matrix puts it in cell P1-A3-UGC with a hypothesis that freshness beats the price angle the account has been running.

Total strategist time: under an hour to review, edit and approve, with a branded share link sent to the client for sign-off.

Who should build this, and the money

This is a build-and-sell blueprint. You build it once in Lovable, brand it for an agency, and charge for setup plus a retainer. Your buyers are paid social and performance agencies with 5 to 40 active client accounts, freelance media buyers who sell strategy and hate writing briefs, and in-house growth teams at brands where creative fatigue is a weekly fire.

The suggested pricing in the pro prompt looks like this:

• Setup: $2,500 to $4,500, covering the build, branding, account connections, a 2-week pilot on 2 client accounts and team training.
• Monthly: $397 to $697 for up to 10 active client brands, plus $150 a month per extra block of 5 brands.
• Performance bonus: $100 for each brief-sourced concept that beats the account's control ad in a live test, capped at $1,000 a month.
• White-label option: $1,500 setup plus $49 per client workspace if the agency wants to resell it under its own brand.

Run the numbers on one agency at the middle of each range. $3,500 setup and $547 a month is about $10,000 in year one before any bonus. Five agencies and you've got a small, sticky business that needs maybe a few hours a month per client to keep running.

Your running costs are low. Lovable Pro starts at $25 a month. Firecrawl's Hobby plan is $19 a month for 5,000 page reads. Claude Sonnet is about $3 per million input tokens and $15 per million output, so a brief with a few hundred reviews costs cents to low single-digit dollars. The Meta Ad Library API is free.

And for the agency, the maths is easy. Your monthly fee is a small slice of what they bill a single client for creative, and it gives every strategist more briefs per week.

Shipping it as a real app

The final build step wraps the tool as a native iOS and Android app (plus a PWA) through AppBuild.DIY. Two native features matter for this one.

Push notifications: "Brief ready", "Client approved 4 concepts", "New client comment". Strategists live on their phones between calls, and a push gets acted on in minutes.

Camera and photos: most competitor research happens while scrolling Instagram or TikTok. With the app on their phone, a strategist screenshots a competitor ad and drops it straight into the brief. The vision model reads it on the spot.

Why build it now

Creative volume is what Meta rewards right now, and agencies know it. They're all trying to ship more angles per month without hiring more strategists. A tool that turns their best strategist's process into a repeatable system, grounded in real customer quotes and checked against Meta's policies, is an easy yes for the right agency owner.

The searches are already happening, too. People are looking for an ad creative brief generator and AI prompts for ad briefs. This blueprint answers both with a complete tool you can build, brand and sell.

Build it this week: grab the Paid Social & Performance Agencies, AI Ad Creative Brief Generator blueprint on Prompt-King. You get the full master prompt, every setup step, the pricing model and a cold pitch ready to send to your first agency.