Writing Meta ads that get clicked: copy, format and why CTR falls

Broad targeting and automated audience expansion moved most audience decisions inside Meta's system. What reaches the algorithm now is who responded, and who responded is decided by the creative. That makes the ad a targeting instrument rather than a message layered on top of one.

The practical consequence is that testing several genuinely different angles teaches you more than testing several audiences against one ad. Different angles attract different people, so the angles produce the audience segmentation you used to build by hand. Five variations of the same thumbnail test almost nothing, because the underlying promise never changed.

It also follows that a borrowed hook can mislead. If a hook works because of who it selects, the same words in front of a different product select a different group, and convincingly.

Low click-through usually means a weak hook, an unclear offer, or an audience whose intent does not match what you are saying. It is a message problem before it is a media problem, and adjusting bids or budgets will not move it.

Diagnose in this order, because the first item explains most cases:

1. Look at the first second of the creative and the first line of copy in isolation. If those two do not earn attention, nothing further in the ad is being read.
2. Check whether the offer is explicit. "Discover our range" is not an offer. A specific product, a specific outcome or a specific price is.
3. Compare click-through by creative, not by campaign. A campaign average hides a good ad being dragged down by three bad ones, and the fix is at ad level.

Judge the number against your own history rather than a published benchmark. Click-through varies enormously by objective, placement and industry, and a figure that means trouble for one account is normal for another. What matters is the direction of travel and the spread between your own ads.

A useful rule that avoids the benchmark trap: treat a creative as failing when its click-through sits below about 20 per cent of your own campaign benchmark, rather than below some published industry figure. That keeps the comparison inside your account, where objective, placement and audience are held constant.

If click-through is healthy but conversions are not following, the problem has moved past the ad. Clicks that do not convert covers that case. If click-through was fine and is now falling, see creative fatigue.

Ad copy does not sell. It creates enough relevance and trust to earn a click, and the selling happens after that. Most ads fail because they lead with the wrong element: the brand, the feature list, or the discount, none of which answer the only question a scrolling reader is asking, which is whether this is about them.

The structure that works consistently is simple:

• Hook. One line that names the reader's situation or the result they want. Treat it as the whole ad, because anyone who does not read past it has still seen this.
• Body. The mechanism. Why this works, briefly and concretely, so the promise in the hook becomes plausible rather than just appealing.
• Proof. Something external: a number, a named customer, a review in the customer's own words. Your own adjectives are not proof.
• Call to action. What happens next, stated plainly. Vagueness here loses people who were otherwise ready.

The most usable copy is often not written from scratch. It is lifted from what customers already say: review text, support tickets, the sentence someone uses when they recommend you to a friend. Mid-rating reviews are worth reading closely, because that is where people describe what nearly stopped them buying, which is the objection your copy has to clear.

For worked examples, 41 brand breakdowns take real ads apart frame by frame, and the creative tool turns an angle into a brief you can hand to whoever produces it.

Video outperforms static in published benchmarks. Static outperforms video in a lot of individual accounts. The discrepancy is execution, and understanding it tells you what to actually test.

One explanation worth considering: a benchmark comparing formats across many accounts is also comparing production effort, since video generally only gets made when someone has a reason to make it. In an account where video means a product photo with a slow zoom, the format label is not what differs. Treat effort as the variable rather than the format.

What each format is genuinely better at:

• Static carries an offer, a price, or a single sharp claim. It is cheap to produce, cheap to iterate, and it survives being seen for half a second, which is how long most of your audience looks.
• Video carries demonstration and sequence. If the product needs showing in use, or the value only makes sense once you understand the mechanism, video does what a still cannot.

Test them as competing angles rather than as a format contest. "Video versus static" is not a question with an answer; "does demonstrating this beat stating it" is, and it has a different answer per product.

One format is worth singling out. Running Reels as a standalone placement rather than only as part of a default buy has published gains behind it: around 38 per cent additional reach, 48 per cent more video plays and a 16 per cent lower CPM. If you are making vertical video already, not giving it its own placement leaves that on the table.

How many ads should I run in one ad set?

Three to six live ads per ad set, with genuinely different angles rather than minor variations. Meta concentrates spend on whichever ad it predicts will win, often within hours, so adding more does not divide budget evenly, it just means most of them never get tested. There is also a cost to leaving them running: conversion rates can fall by around 60 per cent after four exposures, which is why a rotation of different angles beats a larger pile of variations. Campaign structure covers how the concentration works.

Does heavy text on my image hurt performance?

Often, though not for a policy reason. A cluttered image reads as an advert instantly, and an advert is the thing people scroll past. The useful test is whether the image still communicates at thumbnail size with the text unread. If it only works because of the words on it, it is competing for attention it has not earned yet.

Should I use the same copy across Facebook and Instagram?

You can, and it is a reasonable starting point, but the placements reward different things. Instagram is more visual and less tolerant of long copy; Facebook feeds will carry a longer argument. If one placement is dragging your average down, split it out and write for it rather than accepting the blended number.

How do I know whether the hook or the offer is the problem?

Click-through separates them. A weak hook shows up as low click-through, because nobody engaged in the first place. A weak offer shows up as healthy click-through with poor conversion, because people looked and declined. The metric tells you which half to rewrite.

Do I need new creative constantly?

You need it more often than feels reasonable, and how often depends on how fast your audience saturates rather than on a schedule. Rising frequency alongside falling click-through is the signal to act on. Creative fatigue covers how to tell real fatigue from the things that imitate it.