The best call to action for a video ad is the one where a viewer could tell you, in their own words, what happens in the next three seconds if they act. Ranked by that test, "Get your first box for £9, ships Monday" beats "Shop now", which beats "Discover more", which is barely a CTA at all. Below are eighteen CTAs sorted into four tiers by that criterion, with what each one is good for and where it falls apart.
The test used for the ranking
Three questions, in order:
- Is there a verb naming a physical action? Tap, start, book, order, download, reply.
- Does the viewer know where they land? A page, a store listing, a form, a call.
- Do they know the terms? Price, duration, whether a card is needed.
A CTA answering all three is tier one. Two of three is tier two. One is tier three. None is tier four, and tier four is where most video ads end. The ranking is not about which words are punchiest, it is about how much guessing you leave the viewer to do while their thumb is already moving.
Eighteen CTAs, ranked
| Rank | CTA | Verb | Destination clear | Terms stated | Best used for |
|---|---|---|---|---|---|
| 1 | "Tap through, first box is £9 and ships Monday" | Yes | Yes | Yes | DTC first purchase |
| 2 | "Start with your two busiest ledgers, no card needed" | Yes | Yes | Yes | SaaS free tier |
| 3 | "Book a 20-minute call, we will send the recording" | Yes | Yes | Yes | B2B considered purchase |
| 4 | "Reply with your size and we will hold a pair" | Yes | Yes | Yes | DM-led selling |
| 5 | "Order before Thursday for weekend delivery" | Yes | Implied | Yes | Deadline offers |
| 6 | "Search Cadence in the App Store, free for a month" | Yes | Yes | Yes | App install |
| 7 | "Download the two-page checklist, no email required" | Yes | Yes | Yes | Top-of-funnel lead capture |
| 8 | "Claim the half-price box, this week only" | Yes | Implied | Partly | Promotional windows |
| 9 | "Get the size guide before you order" | Yes | Yes | No | Objection-led retargeting |
| 10 | "Try it free for 14 days" | Yes | No | Partly | Familiar category, warm audience |
| 11 | "Shop the shelf" | Yes | Yes | No | Single-product catalogues |
| 12 | "Link in bio" | Weak | Partly | No | Organic, not paid |
| 13 | "Shop now" | Yes | No | No | Retargeting a known brand |
| 14 | "Sign up today" | Yes | No | No | Anything, weakly |
| 15 | "See how it works" | Yes | No | No | Complex products, mid-funnel |
| 16 | "Join the community" | Vague | No | No | Brand campaigns, not response |
| 17 | "Discover more" | No real action | No | No | Nothing |
| 18 | "Follow for more tips" | Yes, wrong one | Yes | No | Growing an account you cannot bill |
Numbers 1 to 7 are worth writing for a specific ad. Numbers 11 to 15 are defaults, and a default CTA is what you use when nobody decided. Numbers 16 to 18 are the ones to cut.
Note number 18. It has a clear action and it still ranks last for a paid response campaign, because it asks for a follow instead of a sale. Clarity alone does not save a CTA that points at the wrong outcome.
Why "discover more" underperforms
Two reasons, and only one of them is obvious.
The obvious one: it names no action. Discovering is not something a thumb does. The viewer has to translate it into "tap the button", and translation costs you the fraction of people who were only half decided.
The less obvious one: it withholds the offer, which means it also fails as a filter. Every unqualified click is paid for twice, once on the platform and again in the conversion rate you then report. Most accounts see the same pattern: a vague CTA holds click-through near the platform median (our benchmark composite puts Meta feed at 1.62 percent and Meta Reels at 1.78 percent) and drops conversion rate, because the people clicking are curious rather than ready. The ad looks fine in the platform column and loses money.
The related mistake is treating the CTA as decoration on the end of the script. It is one of three separately measurable parts, and it is judged on conversion rate, not on watch time. How the hook, body and CTA get measured separately sets out which metric belongs to which slot, and it is worth being strict about, because a strong hook with a tier-four CTA produces exactly the numbers that make teams rewrite the hook.
On-screen and spoken CTAs do different jobs
Say it, show it, and do not make them identical.
| Spoken CTA | On-screen CTA | |
|---|---|---|
| Length | A full sentence, 6 to 10 words | 2 to 5 words, plus the offer on a second line |
| Job | Give the reason to act now | Give the mechanical instruction |
| Timing | Starts as the last shot begins | Appears with the voice, holds 4 to 6 seconds |
| Example | "Start with your two busiest ledgers, it is free while you connect them" | "CONNECT ONE LEDGER" / "FREE, NO CARD" |
| Fails when | It repeats the caption word for word | It arrives with 1.5 seconds of ad left |
The most common technical error is holding the end card too briefly. A URL or brand name read once at speed is not remembered, so hold it for at least four seconds and longer if it contains a word people will misspell. If your ad ends at 15.0 and the card appears at 14.2, you have no CTA, you have a credit.
Sound-off viewers only get the on-screen version, so it has to work alone. That is also why the offer belongs on the card and not only in the voiceover.
Repeat the offer inside the CTA
The viewer who reaches your final four seconds may have arrived at second nine. They did not hear the offer, so restating it is not repetition, it is the first time for a large share of the audience.
Repeat three things: the price or terms, the action, and the friction you removed. "£9, tap through, cancel any time" carries all three in six words. Compare that with "Shop now" following an identical 12 seconds of footage, and the difference in conversion rate is usually larger than anything the hook test produced that month.
Practical way to do it at volume: build the CTA as its own clip rather than an edit at the end of a timeline, so you can swap it independently of the rest of the script. Our notes on CTA clip design cover the frame layout, hold times and safe areas. In Genyad, our product, each variation is generated as a fresh script with its own CTA rather than a re-cut of one master, so a batch can test three CTAs against the same offer. Being clear about the limits: we do not publish to Meta or TikTok for you and we give you no predicted score for a CTA, so you export the files and the platform tells you which one worked. If you want the script and the CTA written from your offer and proof, the AI ad script generator does that part.
How many CTAs to test at once
Two, held against one offer, across a batch of 8 to 20 live variations. More than two and you cannot attribute the difference, since CTA effects are smaller than hook effects and get buried by noise.
Retest when the offer changes, not on a schedule. Hooks fatigue fast: our 2026 fatigue benchmark reports week three landing 45 to 70 percent below the launch baseline, and by week five a creative averaging 38 percent below its peak. A CTA that converts does not decay the same way, so once you find a tier-one CTA, keep it and spend your variation budget at the top of the script.
Frequently asked questions
What is the best call to action for a video ad?
The one that names the verb, the destination and the terms in a single line, such as "Tap through, first box is £9 and ships Monday". Anything that leaves the viewer guessing what happens next converts worse, even when click-through looks acceptable.
Why does "discover more" perform badly?
It names no physical action and hides the offer, so it attracts curious clicks rather than ready ones. Click-through can look normal while conversion rate falls, which is the most expensive way for a CTA to fail because the platform metrics do not flag it.
Should the call to action be spoken or on screen?
Both, worded differently. The spoken line gives the reason to act, the on-screen card gives the instruction and the offer in two to five words, held for at least four seconds so sound-off viewers get the whole ask.
Where should the CTA appear in a 15 second ad?
In the final three to five seconds, with the offer stated earlier as well, before the halfway mark. The end CTA is a second ask, not the first time the viewer hears what you sell.