Ad Creative Review: The Checklist Before You Publish

You made the creative. You looked at it. It looked fine.

Then you published it, and three days later you noticed the disclosure text was printing on top of your logo — on every image in the set. It had been there the whole time. You had looked at that file more than anyone alive and you did not see it, because by the fourth look you were no longer reading the frame, you were remembering it.

An ad creative review is the step that catches what you stop being able to see, and almost nobody running their own marketing has one.

Who has this problem

You are the person who wrote the copy, chose the template, picked the colour and hit render. There is no art director, no brand manager, and no second pair of eyes between you and the publish button.

If you have ever shipped something and spotted the mistake within an hour of it going live, you already know this feeling. The problem is not carelessness. It is that the author is the worst available reviewer of their own work, and you are always the author.

Why this problem exists

Three reasons, and they compound.

You are reviewing intent, not output. You know what the card is supposed to say, so you read what you meant. A stranger reads what is there. This is the same reason you cannot proofread your own writing after the second pass.

The defects are silent. Nothing errors when your text lands at 1.4:1 contrast, when a claim on the frame is one you cannot source, or when your call to action sits under the platform's own action rail. There is no red squiggle. The creative renders successfully and looks finished.

And a documented rule is not an enforced rule. This is the one that gets people. You can have a brand rule saying "never frame self-promotion as fake or pushy," have it loaded, have it printed in the plan file — and still have a persona's pain point, phrased in exactly that language, quoted verbatim onto a card. Journey data reaches the frame with nothing checking it on the way. A rule nobody enforces is documentation, not a rule.

That is not hypothetical. It happened, and it was caught by a person reading the output, not by any of the systems that had the rule in hand.

The ad creative review checklist

Work through these in order. The first five are mechanical — you can check them without taste. The last one outranks all of them.

1. Brand rules, applied to every word on the frame

Not just the copy you wrote. Every word, including the ones pulled in from your product record, your persona and your positioning. Those arrive on the card looking like data and they are exactly as likely to violate a rule as anything you typed.

Check your banned words, your banned claims, your emoji policy, your rules about contractions. Hard rules are absolute — one violation fails the creative regardless of how good the rest is.

2. Claims you cannot source

Every number, every count, every superlative on the frame. "Trusted by 4,200 operators" is a claim you now own publicly.

Bias this check toward over-reporting. It is better to flag five numbers and dismiss four than to miss the one fabricated figure, because a bare count has no unit to give it away. The exception worth carving out: small bare integers like "3 steps" are structural, not claims, and flagging them buries the real finding in noise.

3. Platform mechanics

Is the copy inside the safe area for the specific surface? A Story reserves the top progress bar and the bottom reply field. A Reel reserves the right action rail instead. These are not interchangeable, and using one set of insets for the other medium is the plausible mistake — it either wastes a seventh of your width or puts your call to action underneath a Follow button.

4. Intent — and this is where most reviews go wrong

Decide, out loud, what the creative is for before you grade it:

• Awareness — being seen and understood. A hard sell is a defect here. It gets scrolled past.
• Performance — a tracked click. Having no ask is a defect here.
• Mixed — earn attention, then ask at the end.

The same creative is a success under one and a failure under another. Grading without declaring intent produces confident nonsense: "add a call to action" is bad advice for a knowledge-sharing post. If you have never run an ad, know that these are genuinely different activities, not different intensities of one activity.

5. Look at actual frames, not the settings

A props file cannot show you that two elements are printing on top of each other, or that a card sits still for two seconds, or that your demonstration is a postage stamp in the corner of a vertical frame.

Do not "watch the video and check these six things" — that gets skimmed. Instead pick three specific frames and ask one question of each:

1. The hook frame — at roughly 0.7 seconds, is there anything legible on screen yet? A feed gives you about a second. An entrance that finishes at 0.8s is an entrance nobody saw.
2. The mid-hold of the longest card — is anything still moving? A card whose entrance has finished is a still image, however it arrived.
3. The closing frame — does it point somewhere real, and is that thing inside the safe area?

6. The stranger test

Then the one that outranks everything above.

After watching it once, could someone who has never heard of your company say what it does?

Answer it as the stranger. If the answer is only yes because you already know the product, the answer is no.

A false here is a blocker, and it beats every craft finding on the list. A viewer who cannot say what your product does cannot want it, however well the cut is put together. The usual cause is copy that names what you built rather than what it does for the person watching — which is the developer-founder's own failure mode, and why positioning has to come before creative.

7. Confirm the AI disclosure is actually on the file

If the creative was generated, it should carry a content credential. Attaching it and reading it back are different facts — an agent that ran an attach command will report success whether or not the signing step worked, and a failed signing leaves a file that looks finished. Read the credentials off the rendered file itself.

Why trust this checklist

Because it came from counting real defects rather than from first principles. A single working session produced nine of them, every one found by hand, and every one findable mechanically. The list above is what those nine had in common.

The review also runs as a separate step from generation, on purpose. A review folded into the generation response would be graded by the thing that wrote it — the one arrangement guaranteed to pass. It has to be able to run on work it did not just produce.

Where this will not help you: it grades whether a creative was made well. It cannot tell you whether the offer behind it is any good. A perfectly built creative for a proposition nobody wants is a well-made failure, and no checklist catches that — customer interviews do.

Frequently asked questions

How long should a review take? The mechanical checks are fast. The three frames take a minute. The stranger test takes as long as it takes to be honest.

Can I skip this for organic posts? The stakes are lower, but the stranger test matters more, not less — organic has no targeting to compensate for an unclear message.

What do I do with a "regenerate" verdict? Regenerate, and do not reuse the seed — reusing it reproduces the creative that just failed.

Is a clean review worth recording? Yes. A check that can only ever hurt is a check nobody runs.

Your next step

You now have the checklist. The Mark Platform runs the mechanical half for you — brand rules against every word on the frame, unsourceable claims, platform safe areas, intent — and hands back the three frames to look at yourself, with the exact commands to render them.

See how the creative tools fit together, or create your account and review your first creative →

Related reading: why your ads all look the same and shipping video ads without a video team.