Short-form video earns the watch; the CTA decides what happens next. Most CTAs fail for three reasons: they ask for too much, they sound generic, or they land after the viewer has already scrolled. These 15 prompts cover every short-form CTA scenario — follow, save, send/share, DM, click-out, comment, email opt-in — and pair the spoken CTA with on-screen text, the combination that consistently out-converts either alone.
One thing changed the math in 2026: the engagement signals platforms reward shifted. On Instagram, DM sends (“Sends Per Reach”) are now the single strongest signal for reaching new audiences — weighted roughly 3-5x higher than likes, with anything above a 2% sends-per-reach ratio reading as a viral signal (as of June 2026). On TikTok, saves and shares have been weighted above likes since 2025. So a “send this to…” or “save this for…” CTA is not just polite — it directly feeds reach. The prompts below are built to ask for the action your platform actually rewards.
TL;DR
- Pick one action per video. Splitting the ask (“follow + save + share”) kills all three.
- Match the CTA to the algorithm: sends/DMs on Reels, saves+shares on TikTok, subscribe + pinned-comment link on Shorts, collects + comments on Xiaohongshu (June 2026).
- Land the CTA before the visual payoff ends (last 5-10 seconds), not at the very last frame.
- Pair a spoken CTA with a complementary on-screen text CTA — they should say different things (prompt 7).
- These prompts are model-agnostic. They run well on GPT-5.5, Claude Sonnet 4.6, or Gemini 3.1 Pro; for batch-rewriting a CTA bank, the cheaper Gemini 3.1 Pro or Sonnet 4.6 is fine.
Who this is for
Creators using Reels / TikTok / Shorts / Xiaohongshu video as a lead source, founders running personal-account funnels, brand teams testing CTA hypotheses, and KOL agencies pricing CTA-rate cards.
When not to use these
Skip CTAs entirely for pure entertainment or brand-awareness posts where any ask breaks the spell. Skip them too for autoplay-heavy paid placements, where the CTA belongs on the ad button, not in the video itself.
Match the CTA to the platform signal
Each platform ranks a different action highest in 2026. Ask for the one that compounds into reach.
| Platform | Strongest 2026 reach signal | Best default CTA | Avoid |
|---|---|---|---|
| Instagram Reels | DM sends (Sends Per Reach; ~3-5x likes) | “Send this to the friend who…" | "Tag a friend” |
| TikTok | Saves + shares (above likes since 2025) | “Save this before you…” / “Send to…" | "Smash that like” |
| YouTube Shorts | Watch time + subscribe + pinned link | ”Full version is pinned below” | Bare “subscribe” |
| Xiaohongshu | Collects, follows, comments (interaction ≈ 40% of rank) | “收藏起来，[场景] 再翻出来" | "点个赞” |
YouTube note: Shorts loop, so the final 5 seconds may be seen more than once — that is the strongest CTA slot. Pinning a playlist link (not a single video) in the top comment is worth it: viewers who enter via a playlist are roughly 2-3x more likely to become long-term subscribers. Xiaohongshu note: as of 2026 the recommendation algorithm counts interaction at about 40% of the score, so an interaction hook every ~30 seconds plus a collect CTA matters more than completion rate.
Prompt anatomy
A short-video CTA prompt should carry six elements:
- Role: who the AI plays (Xiaohongshu KOC / TikTok script writer / personal-brand strategist / community manager).
- Context: platform, niche, audience persona, account size, voice — anything that shifts what lands.
- Goal: one concrete deliverable — a hook, a caption, a 60-second script, 10 reply variants, a bio.
- Constraints: length, banned phrases, native idiom, the target signal (send vs. save vs. comment), hashtag count.
- Output format: numbered options, A/B variants, paste-ready blocks, or labeled sections.
- Examples / signal: 1-2 reference posts you like, or anti-examples (“not this generic creator voice”).
Fill the [bracketed] placeholders before sending.
15 copy-ready prompt templates
1. Follow CTA (10 spoken variants)
Default — earn the follow without “follow me”.
You are a short-video CTA writer. My niche: [niche]. My video payoff: [payoff].
Write 10 spoken CTA lines, 8 words or fewer each, that earn a follow WITHOUT
using "follow me". Each must do one of: name the next post coming, name an
ongoing pattern the viewer would miss, or hint at a specific future moment.
Banned: "follow for more", "subscribe", "smash that like".Swap: niche, payoff
If outputs feel pushy, add: “Each CTA must feel like an offer, not a request — the viewer should want the next post, not feel asked to give something.”
2. Save CTA (TikTok’s above-likes signal)
My video covers [actionable topic]. Write 6 save-CTA lines, 10 words or fewer
each, that earn the save WITHOUT saying "save this for later". Name a specific
future moment the viewer will want this back, or a specific person they could
send it to.3. Send / share CTA (Reels’ #1 reach signal)
My video about [topic] would resonate with one specific person type
([friend type]). Write 6 "send this to..." CTA lines that nudge the viewer to
DM the video to that specific person. 10 words or fewer each. Avoid "tag a friend"
— sends, not tags, drive Reels reach.4. DM-specific CTA
My goal: get qualified DMs about [offer]. Write 5 DM-CTA lines that screen for
serious askers: state who this is for, who it is not for, and the exact word to
DM. Spoken-ready, 15 words or fewer each.5. Click-out CTA (link in bio / blog)
My video is a teaser for [longer asset: blog, course, newsletter]. Write 5
click-out CTAs that frame the link as a continuation, not a sales pitch. Avoid
"link in bio" — instead name exactly what the viewer gets at the link
(e.g., "the free setup checklist", "the exact bundle"). Each 12 words or fewer.6. Comment-driven CTA
My video about [topic] ends with a question for viewers. Write 5
comment-prompting lines, 10 words or fewer. Each must invite a specific 5-word
answer, not "thoughts?".7. On-screen vs. spoken CTA pairing
My video has a spoken CTA: "[paste]". Write 3 on-screen text CTAs that pair with
it WITHOUT duplicating it. The two should complement: spoken sets up the why,
on-screen names the exact action. Keep on-screen text under 6 words (readable
on mute).8. Series-continuation CTA
My video is part [N] of [total] in a series on [topic]. Write 5
series-continuation CTAs that earn the viewer back for the next part. Tease the
next part's payoff without spoiling it. 12 words or fewer each.9. Live-stream CTA
I have a livestream on [date / topic]. Write 6 short-video CTAs that drive
viewers to RSVP or set a reminder. Each must give a specific reason the live is
worth the time. 15 words or fewer.10. Newsletter opt-in CTA
My newsletter delivers [value prop] weekly. Write 5 newsletter-CTA lines for a
short-form ending. Each must signal what the newsletter actually delivers
(not "join my newsletter"), why this viewer specifically, and where to opt in.11. Soft CTA (no ask)
My niche audience dislikes pushy CTAs. Write 5 soft CTAs that signal where to
go next without asking outright. Each 10 words or fewer, feels like a casual
mention.12. Multi-platform CTA adaptation
Below is my short-video CTA: "[paste]". Adapt it for four platforms, matched to
each one's 2026 reach signal:
- TikTok: casual, lean on a SAVE or SHARE ask
- Reels: lean on a "send this to..." DM ask (sends drive reach)
- YouTube Shorts: point to the pinned playlist/full version
- Xiaohongshu: warmer, emoji-led, a 收藏 (collect) + comment ask
Output 4 labeled versions.13. CTA A/B test design
I want to A/B test 2 CTAs on the same video concept. Pick 2 different CTA
archetypes (e.g., save-bait vs. send-bait). Output: 2 full CTA lines + a
hypothesis per variant + the single metric I should track to call a winner
(e.g., saves-per-view or sends-per-reach).14. Affiliate / sponsored CTA
My video is sponsored by [brand] or contains an affiliate link to [product].
Write 5 CTAs that disclose honestly, keep my voice, and convert. Each 15 words
or fewer. Use FTC-compliant phrasing and put the disclosure FIRST.15. CTA diagnosis from data
After-post review — diagnose why a CTA underperformed.
My video about [topic] got [views] views but only [count] [action: follows,
saves, sends, DMs]. My CTA: "[paste]". Diagnose in 2 sentences why the CTA
likely failed, then propose 3 rewritten variants engineered for that specific
action.Common mistakes
- Asking for too much (“follow + save + share + DM”) — split attention kills every action.
- Generic CTA (“follow for more”) — every creator says it, so the signal is zero.
- CTA after the visual payoff — viewers scroll the moment the value lands.
- Spoken and on-screen CTA saying the same thing — wasted bandwidth; pair them instead (prompt 7).
- Asking for the wrong action — pushing likes on Reels when sends are weighted 3-5x higher.
- No CTA at all on a video meant to convert — view counts without action.
- Same energy across niches — sophisticated audiences need a softer ask.
How to push results further
- Pick one action per video; the more you ask, the less you get.
- Match the ask to the platform signal: sends on Reels, saves/shares on TikTok, subscribe + pinned playlist on Shorts, collects + comments on Xiaohongshu.
- Place the CTA before the visual payoff ends. On Shorts, use the final 5 seconds (the loop replays it).
- Pair a spoken CTA with complementary on-screen text (prompt 7).
- A/B test 2 CTA archetypes on the same concept (prompt 13) across 4-6 posts to find your niche pattern.
- For sponsored content, lead with the disclosure, then your voice, then the CTA — never bury the disclosure.
- Refresh your CTA bank quarterly; the same lines lose effectiveness as your audience habituates.
FAQ
- Should every video have a CTA? No. Pure brand-awareness or entertainment posts can skip it. But any video meant to convert needs one specific CTA.
- When in the video should the CTA land? Before the visual payoff ends — typically the last 5-10 seconds, not the very last frame. On YouTube Shorts, the loop means the final 5 seconds may be seen more than once, so it is the strongest slot.
- Which converts better, spoken or on-screen CTA? Both together beat either alone. Many people watch on mute, so the on-screen text carries the action while the spoken line carries the why. Use prompt 7 to pair them.
- What CTA should I ask for on each platform? As of June 2026: DM sends on Instagram Reels (the strongest reach signal, ~3-5x likes), saves and shares on TikTok (weighted above likes), subscribe plus a pinned-playlist link on YouTube Shorts, and collects plus comments on Xiaohongshu (interaction is ~40% of the rank).
- Is “link in bio” still effective? Less so in 2026 — platforms suppress vague link-pushing. Instead, name exactly what the viewer gets at the link (prompt 5), or point to a pinned comment on Shorts.
- Which AI model should I use to generate these? Any current frontier model works. GPT-5.5, Claude Sonnet 4.6, and Gemini 3.1 Pro all handle short CTA copy well; for bulk-rewriting a large CTA bank, Gemini 3.1 Pro or Sonnet 4.6 keeps cost down.
Related
- Reel hook prompts
- TikTok caption prompts
- YouTube Shorts script prompts
- Viral opening line prompts
- Trend Adaptation Prompts for Riding Trends Without Cringe
- AI Short-Video Ideation: A Month of Ideas From Your Top Performers
- Short-form Storytelling Prompts for 30-90 Second Arcs
- Social Media Prompts hub