Brands that advertise on Meta need a steady supply of ads that are genuinely different from each other, and most small teams cannot make them fast enough. This guide covers what that work sells for, how Meta's system decides what counts as a new ad, what you may and may not take from a competitor, and how to land the first client.
Published prices, checked 2026-11-02.
Worked out per finished ad, the studios above charge roughly $25 to $200. The spread is wide because the buyers are: a $500 a month subscription sells volume, a strategist sells judgement and accountability. Brands could also use a tool instead: Meta's own AI ad tools are free [9], and creative research tools such as Foreplay ($49 to $459 a month), Atria ($129 to $959 a month) and Motion (from $999 a month) are priced on their own pricing pages as of this guide's date. None of them decides which angle to try next or answers for the result, and that is what you sell. With no case studies yet, expect to start near the bottom of the studio range.
Meta's ad system now does much of the targeting itself. Its retrieval stage, Andromeda, narrows tens of millions of candidate ads down to a few thousand for each person, and Meta built it expecting the number of ads to grow sharply as advertisers use generative AI [9]. Its advice to advertisers is to keep a variety of creative in play, and its Advantage+ sales campaigns are built to test up to 150 creative combinations [11].
When people keep seeing the same image, Meta tells the advertiser. An ad is "creative limited" when its cost per result rises above the account's past ads, and "creative fatigue" when it doubles. The fix Meta recommends is a new ad with a materially different image or video [10]. Agencies told Marketing Brew in September 2026 that Meta treats ads sharing a hook or visual style as the same creative [12].
That is the business in one line: brands need a steady flow of ads that are genuinely different, not ten recolours of one idea. So count what you deliver in concepts. A concept is a distinct combination of these four things, and one concept exported in three sizes is still one ad.
Name the unit in your offer. "Five distinct concepts, each in square, portrait and story sizes" is clearer than "15 ads", and it is the unit Meta's system rewards.
The Meta Ad Library is free and public. For any brand it shows every active ad, when it started and where it runs. For normal commercial ads it does not show spend or results: that data appears only for political and issue ads, plus reach figures for ads delivered in the EU and UK [13]. An ad that has run for months is a reasonable sign it works, because brands rarely keep paying for a loser, but it is an inference. Say so in the report.
A teardown takes the three to five longest-running ads from a client's competitors and explains each one: the angle, the hook, the layout, the offer, the proof, the colour and type. This is where Prompt Reverse earns its place. Run each ad through Image to Prompt and you get its visual recipe in words (composition, lighting, palette, type treatment), which becomes the brief for your client's own version. Then name the gaps: the angles nobody in the category uses, and the formats everyone overuses.
End with five ranked concept briefs for the client's real product and real proof. Standalone reports like this are listed from about $500 [7]. While you have no case studies, a short version given away free and credited against the first paid pack is the most reliable way to start a conversation.
Do this step: Image to Prompt, Bulk Image Reverse, Competitor ad teardown kit.
Copyright protects expression, not ideas. The US Copyright Office states that ideas, methods and systems are not protected [14]. So the angle ("compare the price with a salon treatment"), the structure (big number top left, product in the centre, offer at the bottom), the format and the type of offer are all fair to learn from. Their photos, their words, their illustrations and their exact layout are not.
Trade marks and the overall look of a product go further. In the US, the Lanham Act covers designs likely to confuse buyers about who made something [15]. In the UK, the CAP Code forbids ads that cause confusion with a competitor, take unfair advantage of its trade mark or present a product as an imitation [19], and the ASA has found that a look-alike can mislead even when the two ads are seen at different times [20]. Meta removes ads after a rights holder reports them [22].
Comparisons are allowed in both countries when they are true and fair. The FTC encourages naming competitors if the comparison is clear and accurate [16], and the CAP Code requires like-for-like comparisons [19]. Every claim needs evidence before the ad runs: a reasonable basis for the FTC [17], documentary evidence under the CAP Code [19]. The ASA has said plainly that using AI does not remove the advertiser's responsibility [21].
General information, not legal advice. When a client wants a direct comparison ("half the price of X") or a strong claim ("clinically proven"), ask for the evidence in writing before you design it, and suggest they check it with whoever handles their legal questions.
Start from the client's real product photos. Build scenes and backgrounds around the real product rather than asking a model to invent it, because image models still struggle with exact text, small print and consistent brand elements; OpenAI's own image documentation lists precise text placement and brand consistency as current limits [25].
The flow for one concept: reverse the reference ad for its recipe, rewrite that recipe around the client's product and angle, generate the scene, then set the headline and offer yourself so every word is deliberate. Ad Creative and Creative Variations do the generating. Each concept then needs square (1:1), portrait (4:5) and story (9:16) versions.
Should the ads look made by AI? The evidence is mixed. NIQ's 2024 study found viewers spotted most AI ads and found them more annoying and less memorable [26], while Meta reports more conversions from its own generated images [9]. The safe position is to use AI for the scene and the speed, keep the product real, and avoid a glossy, too-perfect look unless that is the brand.
Meta labels some AI ads. It adds "AI info" under About this ad when it detects third-party AI, and shows a label next to Sponsored when an ad contains a photorealistic AI-generated person [23]. If a brand's appeal depends on feeling hand-made, tell the client before you use generated people.
Do this step: AI Ad Creative, Creative Variations, Social Carousel.
Your prospect list is public. Search your niche in the Ad Library and note brands running several ads that all use the same idea: that sameness is your opening, and running ads at all tells you they have a budget [13]. Check that they sell something you can show convincingly with their own product photos.
Small agencies are the second route. Many manage ad accounts for a monthly fee and bill creative separately, so extra production they can resell under their own name is something they buy. Offer white-label packs at a per-concept price.
Marketplaces can bring first reviews, but prices there start very low [1]. Do not build the business on them.
Follow the rules when you write. In the UK you can email a company or LLP without prior consent if you say who you are and give a simple way to opt out; sole traders count as individuals and need consent first [27]. In the US, CAN-SPAM applies to business email too: an honest subject line, your postal address and a working opt-out [28]. Keep a list of everyone who says no and never write to them again.
There is no reliable data on how long a first client takes, and anyone who gives you a number is guessing. Plan for many conversations before the first yes, and treat the free mini-teardown as the cost of having them.
Ask for read access to the ad account, or a monthly export of results per ad, as part of any paid work. Without it you cannot say which concepts worked, and you cannot defend your price at the next review.
Report monthly in the client's own numbers: which concepts went live, which beat the account's previous best on cost per result, which Meta flagged as tiring, and what you will try next. Recommend switching off your own losers. Nothing builds trust faster.
Keep a concept log with the angle, hook, format and result of every ad. After three months it is the most valuable thing you own: it is what makes your next concepts better than a stranger's, and it is what lets you raise your prices.
Your opening offer. Credit it against the first pack.. About 25 AI Credits to make, 180 minutes hands-on, turnaround 2 to 3 working days.
The core product, for a brand testing every month.. About 105 AI Credits to make, 240 minutes hands-on, turnaround 3 to 5 working days.
Once a pack has run and the brand wants a steady supply.. About 210 AI Credits to make, 480 minutes hands-on, turnaround Monthly, on an agreed date.
To a brand you found in the Ad Library, running several ads on one idea
Subject: [Brand]'s ads, and one angle worth testing Hi [Name], I looked through [Brand]'s current ads in Meta's Ad Library. Most of them use the same approach: [the product on a plain background with a discount]. Meta tends to treat ads that similar as one creative, so they tire out quickly. A few of your competitors are running angles you aren't, like [the before and after that Competitor has run since June]. I've put together a one-page breakdown of three of them and what a [Brand] version could look like. Happy to send it over, no charge. [Your name] [Business name and address] If you'd rather not hear from me, reply "no" and I won't write again.
To an agency that runs Meta ads for ecommerce clients
Hi [Name], I make static ad concepts for ecommerce brands, priced per concept and delivered white-label in square, portrait and story sizes within [3] working days. If any of your clients need more creative than your team can make in a month, I'd be glad to do a first pack for one of them at a reduced rate so you can judge the work. Three examples: [link]
Three or four days after they received it
Hi [Name], did the breakdown land? The concept I'd test first is [the price comparison with salon treatments], because nobody in your category is running it. I can have it ready in all three sizes by [day], as part of a pack of five concepts for [price]. Shall I go ahead?
Turn a winning ad into your own, Diff your best creative against your worst, Reverse a competitor's whole funnel, step by step, One offer, every channel.
No, but you need to understand how ads are judged. Read Meta's help pages on creative fatigue and Advantage+ campaigns, and learn the numbers your client watches: cost per result, click-through rate and return on ad spend. Running the ads stays with the client or their agency. You supply the creative and the thinking behind it.
Learning from them is; copying them is not. Copyright protects expression, not ideas, so angles, structures, offers and formats are fair to study. Their images, words, logos, packaging look and confusingly similar layouts are protected by copyright, trade mark and advertising rules in both the US and the UK. This is general information, not legal advice.
Sometimes. Meta labels ads made or substantially edited with its own AI tools, shows a label next to Sponsored when an ad contains a photorealistic AI-generated person, and adds AI info under About this ad when it detects third-party AI. Tell clients before you use generated people.
Meta gives no single number for ecommerce. Its Advantage+ sales campaigns can test up to 150 creative combinations, and its fatigue guidance says to replace an ad with something materially different once results slip. In practice, count concepts rather than files and plan a monthly refresh.
Be careful. The US Copyright Office says purely AI-generated material cannot be registered and that prompts alone do not make you the author. Do not promise exclusive copyright. Tell the client what they are paying for: the finished files, the concepts and your judgement.
Published prices work out at roughly $25 to $200 per finished ad from productized studios, and around $500 for a standalone competitor report. With no case studies yet, most people start near the low end, often with a free mini-teardown, and raise prices once they can point to a result in a client's account.
Reverse a competitor's ad, or browse every business guide. Pay as you go, no subscription, credit packs never expire.