Facebook ad copy vs creative is not a debate you settle by opinion, you settle it by looking at where the funnel breaks. A low hook rate means people are not stopping, which is a creative problem and no amount of rewriting the primary text will touch it. A healthy hook rate and decent completion with a weak click-through rate means people watched, understood, and did not act, which is a copy and offer problem. In video, creative gets fixed first roughly four times out of five, because the ad has to be seen before the words matter.
Which metric tells you whether copy or creative is the problem?
Read the funnel in order and stop at the first stage that is materially below where it should be. That stage is the thing to fix. Everything downstream of it is unreadable, because it is being measured on a self-selected sample.
| Hook rate | Completion | Link CTR | What is actually wrong | Fix this |
|---|---|---|---|---|
| Low | n/a | n/a | Nobody is stopping. Downstream numbers are noise from a tiny sample | The first two seconds: new opening shot, new spoken hook, new on-screen line |
| Good | Low, drops before 5s | Low | The hook overpromised and the middle does not pay it off | The body of the video: get the proof point earlier |
| Good | Good | Low | They watched, they understood, they were not moved to click | Copy and offer: primary text, headline, the CTA, the promise itself |
| Good | Good | Good, weak conversion rate | The ad sold something the landing page does not deliver | Message match and the page, not the ad |
| Good | Good | High, high cost per result | The ad is attracting the wrong people well | The qualifying line: say the price or the audience out loud |
| Falling week on week | Falling | Falling | Fatigue, not a creative or copy defect | Rotation: ship replacements |
Two definitions, because people use these words loosely. Hook rate is three-second views divided by impressions, and it is the cleanest read on whether the creative earns attention. Completion rate is the share of viewers reaching the end, and it tells you whether the middle holds the promise the opening made. Between them they separate "not seen" from "seen and not persuaded", which is the only distinction that matters for this decision.
Why creative comes first in video
In a video ad, the copy above the video does not get read until the video has already earned a pause. Watch anybody use Facebook or Instagram on a phone: the thumb moves, the video autoplays silently, and the primary text is truncated to a line and a half with a "more" link nobody taps. The sequence is visual, then audio if sound is on, then text if the first two won.
That ordering has a consequence for where your effort goes. If your hook rate is 12 percent, four-fifths of your audience never reached your copy, so a copy test on that ad is a test on the surviving fifth. The number you would be optimising is derived from a sample that already self-selected as interested.
For scale on what a normal hook rate looks like, our 2026 fatigue benchmark puts the Meta feed and Reels median at 28 percent hook rate, with median CTR of 1.62 percent on feed and 1.78 percent on Reels. Those are medians from a synthesis of published platform and agency figures, not targets we measured ourselves, and they are useful mainly as a floor: sitting at half the median hook rate is a creative diagnosis, not a copy one.
There is one caveat worth stating, because it cuts against us. On static image ads and on carousels, this ordering weakens a lot. The image is processed in a glance and then the reading starts almost immediately, so copy carries far more of the persuasion. Genyad only makes video, with no static banner formats at all, so if your account is mostly static then the ordering in this post applies less to you and copy deserves a bigger share of your testing budget.
When copy is genuinely the problem
Four patterns where the video is fine and the words are the constraint. We see all four regularly.
Good completion, weak CTR. People watched to the end and did not click. That is not indifference, it is a missing reason to act now. Usually the ad demonstrated the product and never said what happens next, or the offer is buried in the last two seconds of voiceover instead of stated in the headline.
Strong CTR, terrible conversion rate. The copy wrote a cheque the landing page cannot cash. Read your primary text and your page hero side by side. If the promise or the price differs, that gap is your leak, and it is a copy fix rather than a creative one.
High volume of cheap clicks, high cost per acquisition. The copy is not qualifying. Adding the price, the commitment, or a plain statement of who this is for ("for teams already running paid social") lowers CTR and lowers cost per acquisition at the same time. Many teams read the CTR drop as failure and revert. Watch cost per result instead.
The offer is the problem, not the words. Nobody wants to hear this, but the copy is where it becomes visible. If four honestly different creatives and three copy angles all land within noise of each other on a bad number, the thing being sold, or its price, is what needs changing. No hook fixes a €90 product that people value at €40.
How to test them separately
The temptation is to change the video and the copy in the same round. Then something improves and you cannot say why, and next month you cannot reproduce it.
The clean pattern:
- Creative round. One ad set, four to six genuinely different videos, identical primary text and headline on every ad. Read hook rate at 48 hours and cost per result at day five to seven. You now have a winning video.
- Copy round. Hold that video constant. Three to four primary text variants that differ by angle, not by wording: a pain-led open, an offer-led open, a proof-led open, a plain-product open. Same headline, or vary the headline in a separate round.
- Recombine, then keep going. Winning video plus winning copy becomes the control. Both fatigue, so both come back around.
Two rules that keep this honest. Never fewer than four days per round, because day-of-week variation is bigger than most copy effects. And do not use the winning copy from twelve months ago as a permanent control, because our benchmark report finds most creative effectively dead within three weeks and week three running 45 to 70 percent below the launch baseline. Copy decays more slowly than video, but it decays.
Our creative testing workflow sets out the round structure with the reporting side attached. Genyad, our product, sits on the creative half of this: it turns footage you already own into new variations, each with its own script, shot selection, voiceover and captions, which is what makes a four-to-six-video round practical when a manual edit takes half a day. It writes the on-screen and spoken copy inside the video. It does not write your primary text or headline, it does not publish to Meta, and it gives no predicted performance score, so the diagnosis above stays your read of your own numbers.
Frequently asked questions
What matters more on Facebook ads, copy or creative?
For video ads, creative, because the video has to stop the scroll before anyone reads a word. For static images and carousels the balance is much closer, since the image is processed instantly and the copy does most of the persuading. Let your hook rate decide: below about half the platform median, the problem is creative regardless of how good the copy is.
How do I know if my low CTR is a copy problem?
Look at hook rate and completion first. If both are healthy and CTR is weak, the audience watched and understood and still did not act, which points at the copy, the CTA and the offer. If hook rate is low, your CTR is being calculated on the few people who stopped, and fixing copy will not move it.
Should I test ad copy and creative at the same time?
No, run them as separate rounds. Test four to six videos with identical copy first, then hold the winning video and test three or four copy angles against it. Changing both at once gives you a result you cannot attribute or repeat next month.
How many copy variants should I test per video?
Three or four, differing by angle rather than by phrasing: pain-led, offer-led, proof-led, plain-product. Twelve rewordings of the same angle is one test pretending to be twelve, and it will spread your event volume so thin that nothing reaches a readable difference.