//Guide · Updated October 11, 2026

How to test native ads

Testing native ads is simple once you change one thing at a time and start from ideas that already work. Here is the method we use, step by step, for Taboola, Microsoft, Outbrain, Revcontent and MGID.

See what’s winning for your business, free →

Seven steps to a native ad test you can learn from

Write down what you are testingPick one question per round, like which kind of picture works best for your offer. Keep the audience, country, device, bid and landing page the same for every ad in the round, so the picture is the only thing that changes.
Start from winning formulasChoose three to five kinds of picture that are lasting in your space right now, and make your own version of each for your product. Starting from proven winning formulas means your first round tests good ideas against each other instead of guessing.
Pair each picture with two or three headlinesWrite headlines that explain the picture, with the hook in the first 40 characters. The network rotates the combinations and shows the stronger ones more often.
Give every ad a fair amount of spendDecide before you start how much each ad gets before you judge it. A common rule of thumb is a few times your target cost per conversion. Clicks alone are not enough: judge on the result you pay for, like a lead or a sale.
Cut the clear losers, keep the leadersAfter each ad has had its spend, pause the ones far behind. Keep the leaders running. Don’t pick a winner on a handful of conversions; differences that small are often luck.
Add new pictures every weekReplace each loser with a new picture built on a winning formula from your space, so every round starts stronger than the last. Most native ads are switched off within weeks, so a steady supply of new pictures matters more than any single ad.
Keep a simple logNote each picture’s winning formula, its headline, the spend and the result. After a few rounds you will see which kinds of picture work for your offer, which is worth more than any single winning ad.

What to change, in order

The kind of picture. A crew on a roof against a product shot against a happy customer. The biggest lever. Kinds of picture that keep winning.
The take on a winning kind. Once you know the kind that works, test different moments, people and angles within it.
The headline. Two or three per picture: different hooks, same promise. What the top headlines do.
The landing page. Only when the ads are already getting clicks that don’t convert.

Common testing mistakes

Changing several things at once. New picture, new audience and new bid in one go, and you can’t tell what worked.
Judging on clicks. A curious picture can win clicks and lose money. Judge on the result you pay for.
Calling it too early. A few conversions either way is noise.
Running out of new ads. Tests stall when there is nothing new to try. That is the part ConvertBetterAds takes off your plate: new pictures every week, built on what is winning in your space, sized for every network.
Ignoring the rules. A winning test is worthless if the ad gets rejected at scale. Why native ads get rejected.

See what’s winning for your business, free

Paste your website. We find the native ads winning in your exact space right now, free, with how long each has run. On a plan, we show you why they work and make new ads for your business from them every Monday.

Your website, or the page you’re advertising · Free look in about a minute · No card

Questions

How many native ads should I test at once?

Three to five pictures with two or three headlines each is a practical start. Fewer gives the network too little to choose from; many more spreads a small budget so thin that nothing gets enough spend to judge.

Should I test the picture or the headline first?

The picture. On native, the picture is what stops a reader in a feed of articles, and in our experience it moves results more than the words do. Find the kind of picture that works for your offer, then refine headlines on it.

How long should a native ad test run?

Until each ad has had the spend you set in advance, not a fixed number of days. In practice, most tests can be read within one to two weeks. Ads that haven’t shown promise by then rarely turn around.

Can I test native ads on a small budget?

Yes, by testing fewer ads at a time. Three pictures with real spend each teach you more than ten with a few clicks each. Start with the winning formulas in your space, so even a small test starts from strong ideas.

What’s winning this week, space by space → · Native Winners, free every Monday →

Figures from our research for the week of November 2, 2026: the most-bought native ads in the US from public native ad libraries. They count ads, advertisers, days running and placements, never anyone’s clicks or sales.