Start From Proven Ad Messages, Not Blank-Page Brainstorms

Every brainstorm starts with the same problem: a room full of smart people guessing what strangers will care about. Some guesses are good. Most are not. And you only find out which after you have paid to test them.

There is a cheaper starting point. In almost every consumer niche, competitors have already spent real money testing messages. The ones that failed got switched off. The ones that worked are still running, sometimes for months. That record is public, and you can learn from it without copying a single word.

This article explains why proven messages make a better starting point, how to find them, and how to turn them into original ads that are yours.

Messages, not ads

First, a distinction that matters for both strategy and ethics.

An ad is a specific execution: the exact words, image, video, layout and brand. An ad belongs to the company that made it. Copying it is a bad idea for legal, policy and practical reasons.

A message is the underlying idea: the promise made, the proof offered, the offer presented, and the buyer it targets. "Hot sleepers want sheets that stay cool, and a long trial removes the fear of buying bedding online" is a message. No one owns it. Dozens of brands could express it in completely different ways.

When we say start from proven messages, we mean learn what ideas the market keeps paying for, then express them with your own product, your own proof, your own offer and your own creative.

Why blank-page brainstorms waste money

Brainstorms are biased toward what the team finds interesting. Founders love their origin story. Product teams love features. Designers love clever concepts. Buyers usually care about something simpler: a specific problem solved, with enough proof to believe it.

The cost shows up in testing. Say you run creative tests on Meta and each new concept needs about $300 of spend to get a read. Ten brainstormed concepts cost $3,000 to evaluate. If two of them work, you paid $1,500 per winner. If you start from ten messages that already have months of market evidence behind them, your hit rate should be higher, so the cost per winner falls even though the testing budget is the same.

That math is illustrative, not a benchmark. But the direction holds: the more of your testing budget goes to ideas with prior evidence, the less of it you spend learning what competitors already learned.

Where proven messages come from

The Meta Ad Library

The Meta Ad Library at facebook.com/ads/library shows every active ad running on Facebook, Instagram, Messenger and Audience Network. Each ad shows the date it started running. That date is the single most useful field in the library.

Advertisers rarely keep paying for ads that lose money. An ad that has been active for 60 days or more is a strong signal that it is profitable, or at least worth keeping. An ad that started yesterday tells you nothing yet. Our guide to why long-running ads are your best creative research explains the signal and its limits.

Your own account

If you have run ads before, your best historical performers are proven messages too. Sort by spend over the last 6 to 12 months and look at what kept getting budget.

Your customers

Reviews, support tickets and post-purchase surveys tell you which outcomes buyers cared about, in their own words. These are not proven ad messages yet, but they are the raw material. Our voice-of-customer research guide covers how to mine them.

A five-step process to extract proven messages

Step 1: Define the niche narrowly

"Skincare" is too broad. "Fragrance-free moisturizers for sensitive skin" is a niche. The narrower you go, the more relevant the ads you find. List three to five search terms and five to ten competitor brands.

Step 2: Collect long-running ads

Search the Ad Library for each term and brand. Filter to active ads in your country. Save every ad that has been running for 60 days or more. Aim for 20 to 40 ads. Note the start date, format and landing page for each.

Step 3: Break each ad into its message

For every ad, write one line for each element. The promise, proof, offer framework works well here:

The winning ad analyzer can speed this up: paste an ad's text and it breaks out the components for you.

Step 4: Cluster and count

Once you have 20 or more rows, group similar messages. You will usually find three to six clusters. For example, in a hypothetical sensitive-skin niche:

• "No fragrance, no irritation" (eight ads)
• "Dermatologist-tested for reactive skin" (six ads)
• "Fixes the tight feeling after cleansing" (five ads)
• "Simple routine, three products" (four ads)

The size of each cluster tells you how many advertisers have found that message worth paying for. The age of the oldest ad in each cluster tells you how durable it is.

Step 5: Map each message to your proof

Now ask, for each cluster: can we make this promise honestly, and what is our proof? If a cluster leans on dermatologist testing and you have not done that testing, skip it or find a different proof. If your product genuinely fits a message, write it down along with your specific evidence and offer.

What you end up with is a short list of messages that have market evidence behind them and that you can support. That is your creative brief.

Turning messages into original ads

This is the part where you stop looking at competitors.

Take one message and write it from scratch using your product details. Change the hook, the visual, the format and the voice. Use your own photos and your own customers. The goal is that someone looking at your ad next to the competitor's would see two brands making a similar point in completely different ways.

Here is a hypothetical example. The proven message is "people with sensitive skin are tired of products that sting, and they trust simple ingredient lists."

A ChatGPT ad built from that message for a hypothetical brand might read:

• Headline (36 chars): A moisturizer with seven ingredients
• Body (96 chars): Fragrance-free and made for skin that reacts. Read the full list, then try it with free returns.

A Meta primary text version could open: "If your moisturizer stings, start by reading the label. Ours has seven ingredients, all listed on the front."

Both carry the same message, both use the brand's own proof, and neither borrows a competitor's words.

What to avoid

• Copying copy or creative. It invites policy problems, trademark issues and an ad that does not fit your product.
• Using competitor names in ads. ChatGPT ad policies prohibit using other companies' trademarks, and comparative claims on any platform need careful substantiation.
• Borrowing proof you do not have. If competitors cite clinical testing and you have none, that message is not available to you yet.
• Treating longevity as certainty. Big brands sometimes run ads for awareness or out of inertia. Weight clusters of long-running ads over any single one.
• Ignoring fit. A message can be proven in your niche and still be wrong for your price point or buyer.

Proven messages on new channels

This approach is especially useful when you move to a new channel, because you have no history there. ChatGPT ads are the clearest current example. Self-serve ChatGPT Ads Manager opened to US advertisers in May 2026, so most brands have little or no performance data there yet.

Rather than guessing, you can carry over messages that have already survived months on Meta and rewrite them as answers to the questions buyers ask ChatGPT. The message stays the same; the execution changes to fit a short chat card shown under an answer. Our guide on turning Meta winners into ChatGPT ads covers that translation step by step.

How SecondWin does this for you

SecondWin automates the process above for ChatGPT ads. You enter your URL, and it reads your site to work out your product, buyer and niche. It then mines the Meta Ad Library for the longest-running ads in that niche, extracts the promise, proof and offer behind them, and identifies the questions buyers ask ChatGPT that each message answers. Finally it writes original ChatGPT ads for your brand and checks each against OpenAI's ad policies.

The preview is free and needs no account. Run a free analysis of your site to see which proven messages show up in your niche, or check the plans and pricing if you want SecondWin to launch and manage the campaigns.

FAQ

Is it legal to learn from competitors' ads?

Studying publicly available ads is legal and common practice; the Meta Ad Library exists so anyone can see what is running. What you should not do is copy another brand's copy, images, video or trademarks. Learn the underlying message, meaning the promise, proof, offer and target buyer, and then build original creative around your own product and evidence.

How long does an ad need to run to count as proven?

There is no official threshold, but 60 days or more of continuous running is a useful working rule. Most advertisers review performance weekly or monthly and switch off losers. An ad that survives two months has likely passed several of those reviews. Clusters of similar long-running ads from different brands are stronger evidence than a single ad.

What if my niche has very few ads in the library?

Widen the search to adjacent niches that share your buyer or problem. A new pet supplement brand might study pet food and pet grooming ads. Also look at your own customer reviews and support questions, which reveal the outcomes buyers care about. You can then test a smaller set of messages with modest budgets to build your own evidence.

Do proven Meta messages work on ChatGPT ads?

The message often transfers well, because the buyer and their problem are the same. The execution has to change. ChatGPT ads appear below an answer to a question, with a short headline and body, so the message should be written as a relevant answer to that question. Early results on any new channel vary, so test a few messages rather than assuming.