What the words on a Meta ad do
The creative hub explains the mechanism. Meta documents that its auction weighs a bid together with estimated action rates and ad quality, and that estimated action rates draw on how people have responded to the ad so far. An ad whose words speak to one kind of person gets responses from that kind of person, delivery learns who they are, and the ad is shown to more of them. Copy is therefore not only persuasion aimed at whoever happens to see the ad; it shapes who sees it next.
That changes what good copy looks like. Copy written to appeal to everyone gets a thin response from everyone and teaches delivery little. Copy that names a situation, a price band or a commitment gets a stronger response from the people it names and a weaker one from the rest, and both responses are useful signals. The visual still does most of the work of stopping the scroll, which the hooks and user-generated content guide covers for video; the words decide what the person understands once they have stopped, and whether the click that follows is worth paying for.
The fields: primary text, headline, description, button
The formats and placements guide lists the text fields each format uses: primary text, a headline, a description and a call-to-action button for single images and videos, and a headline, description and link on each card of a carousel. It also notes that in several cases the text can be customised per placement, alongside the media. Each field has a different job, and copy works best when the jobs are kept apart.
| Field | Its job | Writing rule |
|---|---|---|
| Primary text | Opens the conversation and carries the argument | Put the line that matters first; everything after it may be hidden until the person taps to read more |
| Headline | States the offer or the next step, beside the button | Specific and concrete: the product, the price, the outcome or who it is for |
| Description | Adds one supporting fact where the placement shows it | Treat it as optional: a guarantee, delivery terms or a proof point, never the only place a key fact appears |
| Button | Names the action the click will lead to | Choose the button whose wording matches what the landing page asks the person to do |
The button is the ad's call to actionThe prompt that tells a reader the one action to take next, such as a button or link., and choosing it is about honesty more than flair: the options available are the ones Ads Manager shows for the ad, and a button that promises one thing while the page asks for another breaks Message MatchHow closely a landing page repeats the promise and wording of the ad or link that sent the visitor. at the moment of decision. The creative brief template is the place to agree each field's job before anyone writes a line.
Writing for truncation and sound-off feeds
Placements show different amounts of text. The formats guide's advice is to keep primary text short enough to survive truncation in the placements that show less of it, and to treat the current specification and the preview in Ads Manager as the reference rather than a remembered figure. In practice that means writing the primary text as if only its first line will be read, because in some placements it will be. A useful test is to cover everything after the first line and ask whether the ad still makes sense.
Silence is the second constraint. The formats guide notes that feed ads are watched largely without sound, so captions and on-screen text carry the message, while Reels are more often watched with sound but the first frame still has to hold without it. For copy this means the words cannot depend on a voice-over to make sense. If the video's argument lives in the narration, the primary text or the on-screen captions have to carry it as well.
Placement asset customisation helps with both. Where text can be customised per placement, a shorter version for the placements that cut most aggressively keeps the message intact without weakening the version feeds show in full. Check each one in the preview on a phone, which the formats guide calls the test that catches problems a desktop monitor hides.
Copy that selects the right people
The most useful thing copy can do on Meta is make the right people more likely to respond and the wrong people less likely. This is the pre-qualifying principle the guide to pre-qualifying search ad copy applies to Google: state what would make someone the wrong customer, and they stop clicking. On Meta the effect compounds, because delivery learns from the responses the copy attracts.
Three kinds of line do most of the selecting:
- Who it is for. Name the person or the situation: the kind of household, business or problem the offer suits. A reader who does not recognise themselves moves on.
- The price or price band. Stating it filters out people who would abandon at the checkout, and stops delivery learning from clicks that were never going to convert.
- The commitment. A subscription, a minimum term, a consultation, a long form: if it will surprise someone after the click, say it before.
The cost is visible and easy to misread. Clearer copy usually earns fewer clicks, and a team watching click-through rate will see the change as a failure. The diagnosis page for low-quality leads from Meta lead ads shows the opposite problem: copy and forms so frictionless that delivery fills the CRM with people who never meant to buy. Copy that pre-qualifies is one of the cheapest fixes for it.
Claims, proof and personal attributes
Copy is where most policy problems are written. The site's pages describe the Advertising Standards as prohibiting misleading claims and deceptive practices and requiring testimonials and endorsements to reflect genuine experience, which the hooks and UGC guide covers for creator content. The healthcare marketing guide adds the rule that catches copywriters most often: ads must not assert or imply personal attributes, and medical conditions are among them. Copy that addresses the reader as someone who has a condition risks breaking it; copy that describes the service and who it is for, and lets the reader decide whether it applies, does not. The same caution applies to other personal attributes the standards list, so "you" lines deserve a second read.
Proof belongs in copy, provided it is real. A specific, genuine customer quote, a guarantee the business honours or a fact about the product is stronger than an adjective, and the features, advantages and benefits framework is a quick way to turn a feature into the benefit a reader cares about without inflating it. Every claim should be one the business can substantiate on request. The guide to ad review and rejected ads sets out what happens when copy crosses the line, and the pre-launch check that catches it first.
Patterns by funnel stage
The reader's state decides what the copy has to do. Someone who has never heard of the business needs a reason to care; someone who visited last week needs a reason to come back. The patterns below are general practice, a starting point for writing candidates rather than formulas.
| Stage | Reader's state | What the copy does | Useful structure |
|---|---|---|---|
| Cold prospecting | Has not heard of you; scrolling for something else | Names the situation and the outcome; states who it is for | Problem, outcome, who it suits, then the offer |
| Consideration | Has seen or engaged with an ad or the profile | Answers the obvious objection; gives one proof | Objection, proof, how it works |
| Retargeting | Visited the site or added to basket | Removes the remaining doubt; makes the next step easy | Reminder, reassurance, the offer or guarantee |
| Existing customers | Bought before | Gives a reason specific to them: a refill, an upgrade, a related product | What is new for them, and why now |
The AIDA framework maps well onto cold prospecting copy, where attention and interest have to be earned before desire and action can follow; later stages can skip straight to desire and action because the earlier work has been done. Whatever the structure, one ad should carry one idea. Copy that tries to introduce the brand, answer three objections and announce a sale at once ends up doing none of them.
Testing copy without changing everything
Copy is cheap to change and easy to test badly. The creative testing guide sets the rule: one variable per test, decided before launch, with the question written down. For copy that means holding the visual constant and changing only the words, or the reverse, so that the result says something about copy rather than about the whole ad. The guide's most controlled method, Meta's A/B test tool, splits the audience into non-overlapping groups, which an A/B TestControlled comparison of two variants. of copy needs, because several ads in one ad set are a competition that delivery settles early rather than a fair comparison.
Generated copy makes the discipline harder. It is easy to produce dozens of lines that differ in everything at once, and the AI-assisted ad copy testing guide shows how to keep generated variants tied to one question and reviewed before they run.
Read the result on the metric the objective cares about. Click-through rate (CTRShare of impressions that result in a click.) is a diagnostic, and the diagnosis page for a low click-through rate explains why link click-through, which counts clicks to the destination, is the figure to use rather than all clicks, which include taps to expand the text. The CTR calculator puts a rate on each variant; cost per result and conversion from the click decide the winner.
A meal-box service rewrites its ads (illustrative)
Turnipwren Meal Boxes states the price and who the boxes suit
Turnipwren Meal Boxes, an invented meal-box delivery service, runs prospecting ads whose primary text reads "Delicious dinners delivered to your door. Try us today!" with the headline "Fresh meal boxes". First orders are expensive and many customers cancel after the first box, saying it was not what they expected. The team writes a second version with the same video: primary text "Three dinners for two from £39 a week. Recipes take about 30 minutes, and you can skip or pause any week." and the headline "For couples who cook on weeknights". Both run through the A/B test tool for three weeks at £1,500 each, judged on cost per first box.
| Version | Spend | Impressions | Link click-through | Link clicks | Click to first order | First orders | Cost per first box |
|---|---|---|---|---|---|---|---|
| Original copy | £1,500 | 125,000 | 1.6% | 2,000 | 2.0% | 40 | £37.50 |
| Price and audience stated | £1,500 | 125,000 | 1.4% | 1,750 | 3.4% | 60 | £25.00 |
Link click-through fell slightly, which the team had agreed in advance not to treat as the verdict. The new copy turned away people who wanted cheap single meals or did not cook, and the people who clicked were far more likely to order. The team keeps the visual and plans the next test on the headline alone.
Common mistakes
- Writing the important line last. Placements cut primary text; put the message in the first line.
- Copy that depends on sound. Most feed viewing is silent; the words must carry the argument without narration.
- The same line in every field. Primary text, headline and description each have a different job.
- Writing to everyone. Copy that names who the offer is for, the price and the commitment selects better customers.
- Addressing the reader as having a condition. Describe the service and who it is for instead.
- Testing copy and visual at once. Change one, hold the other, and write down the question.
- Calling the winner on click-through. It is a diagnostic; judge on cost per result and conversion from the click.
- A button that does not match the page. The click should lead to exactly the action the button names.
