AI Hook Generator: 40 Scroll-Stopping Templates

A hook generator has to clear an impossibly short attention window. On TikTok, the median time a viewer spends on a video before swiping is about 1.5 seconds. On Instagram Reels it's roughly 2.6 seconds (per Meta's 2024 creator-research summary). On the X mobile timeline, dwell on a single post hovers near 1.1 seconds. Miss that window and the post doesn't just fail with the current viewer — the ranker reads "low hold-rate" and demotes the post for the next viewer too. Hold-rate compounds. So does its absence.

This is why the search volume for AI hook generator has more than doubled since late 2024: creators figured out the leverage is in the opening, not the edit. But most generators ship dead copy — generic, undated, voice-flat, niche-blind. Below are 40 templates that aren't, organized into 8 families of 5, each labelled with the platform mechanics that explain why it works. Skim to the family you need; every template ships with a skeleton, a worked example, the anti-example to avoid, and the platforms it fits best.

TL;DR. Forty hook templates, each tagged with the platform mechanic that explains why it works, the anti-example to avoid, and a 4-force scorecard (specificity, stakes, surprise, self-interest) so you can grade your own openers before you post.

What a hook generator actually is#

The word "hook generator" gets glued onto three different tools that solve three different problems. Pick the wrong one and you'll optimize for the wrong metric.

1. Headline rewriters. Optimize the <title> tag and the link-preview card. Their job is the click-through rate from a link shared in a feed — not the feed post itself. Useful for SEO pages and Substack subject lines (which behave like email headlines).
2. Caption generators. Write the body copy that supports the hook — the paragraph after the opener, the CTA, the hashtags. They assume the hook already exists.
3. Hook generators. Write the first one or two lines that appear inside the platform's visible viewport before the viewer decides to scroll. This is the only category platform rankers actively punish you for getting wrong.

This playbook is about category 3. If a tool offers to "write a hook" and gives you a 90-character paragraph that opens with "In today's fast-paced world…", it's a caption generator pretending. (For the longer take on what good idea-generation pipelines look like, see our 2026 guide to AI content idea generators.)

The 4 forces every great hook uses#

Every working hook on the list below leans on at least two of these four forces. One is bait. Three is rare. Two is the floor.

• Specificity. A real number, a real date, a real platform. "In April 2026, Google overtook Substack in news traffic" beats "search is changing". Specificity triggers the brain's pattern-matching heuristic: real numbers feel non-AI, non-recycled, worth the next 3 seconds.
• Stakes. Something is being won, lost, or risked — money, time, reputation, attention. "I almost shipped a $40k launch with the wrong headline" triggers the loss-aversion override that keeps the thumb still.
• Surprise. A claim that contradicts the reader's default assumption. "Posting daily killed my newsletter" opens a 1-second confusion loop the brain has to close. The close costs a tap.
• Self-interest. The reader has to feel "this is about me" inside the first six words. "If you write a Substack with under 500 subs, this is for you" beats "On audience-building strategy". Self-interest is what converts the dwell into a save.

Weak hook (one force, Specificity only): "In 2026, 47% of creators report burnout." Strong hook (three forces, +Surprise +Self-interest): "If you posted more than 20 times last month and nothing broke 5k views, the algorithm isn't your problem — your hook is. Here's the proof from 47% of 2026 creators."

If a hook hits zero forces, rewrite it. If it hits three, save it to your library. (For the 12-template short version of this framework, see Write scroll-stopping hooks.)

Why most AI hook generators produce dead copy#

Five named failure modes account for almost everything bad an AI hook generator outputs. Each is fixable, but only if the tool ships the right scaffolding around the model.

1. Trained, not live. A model trained in 2024 still suggests "POV:" openers because they were viral in 2023. POV died on TikTok in late 2024 as the format saturated; the replacement is the direct-to-camera question ("Are you the friend who…?"). A useful hook generator pulls live signals at request time, not at training time.
2. Niche-blind. The same hook gets handed to a fitness coach and a B2B SaaS PM. "The 3 things I wish I knew at 25" works for neither. A useful generator scores against a niche profile — at minimum: industry, audience, format, voice, goal, and competitor set.
3. Voice-flattened. A LinkedIn confessional opener ("I'll be honest — last quarter I underperformed.") dropped into a TikTok caption reads like a stock photo. Each platform has its own voice register; the generator needs a per-platform voice layer, not a single tone setting.
4. Stake-free. Claims without consequence: "I learned something cool about marketing." No money, no time, no reputation on the line — no reason to read line 2. Generators that don't model stakes generate captions with hooks bolted on.
5. Length-deaf. A 38-word "hook" that pushes the verb below the visible viewport. The platform-by-platform character ceilings before truncation are: X 280 (but the preview card cuts ~140), LinkedIn ~210 before "see more", Instagram captions ~125, TikTok overlay text ~80, YouTube Shorts title 60, Substack inbox subject ~50. A length-aware generator caps every template to the platform.

TINS HUB addresses each of these in the same pipeline: discovery runs every batch (live, not cached training), every idea is scored against your 6-field niche profile, each platform has its own voice layer, and the decision engine flags hooks that score low on the 4-force test before you ever save them.

The 40 hook templates#

Each template ships with the same four fields so you can scan, copy, and ship: Skeleton (the structure with bracketed slots), Filled (a worked example in a named niche), Don't (the anti-example showing the most common failure), and Best on (the 1–2 platforms it fits best, with the mechanic that explains why).

Family A — Confession & Contrarian (1–5)#

These hooks work because the speaker takes a small reputational risk. Risk reads as authentic, and the platform rewards it: LinkedIn pre-scores text dwell-time and "see more" clicks, both of which spike on confessional openers.

1. The Admit

• Skeleton: "I need to admit something. I've been [common belief] for [time period] — and it cost me [stake]."
• Filled: "I need to admit something. I've been treating my Substack like a blog for 14 months — and it cost me about 600 subscribers."
• Don't: "Here are some things I learned about Substack."
• Best on: LinkedIn, Substack Notes. Confessional registers spike dwell.

2. The Public Disagree

• Skeleton: "Unpopular take: [thing everyone says] is wrong for [specific group]."
• Filled: "Unpopular take: 'post consistently' is wrong for solo B2B founders with under 1k followers — you should post rarely and well."
• Don't: "Here's my unpopular take on consistency."
• Best on: X, LinkedIn. Contrarian claims spike reply rate, which the X feed ranks heavily.

3. The Permission Slip

• Skeleton: "You don't have to [pressured behavior] to [desired outcome]. Here's what to do instead."
• Filled: "You don't have to film yourself talking to camera to grow on TikTok. Here's the silent-format that's outperforming for finance creators this quarter."
• Don't: "Tips for growing on TikTok."
• Best on: TikTok, Reels. Permission triggers a save (the platform's strongest engagement signal).

4. The Heretic Question

• Skeleton: "Why are we still [accepted practice] when [evidence it stopped working]?"
• Filled: "Why are we still optimizing for follower count when watch-time-per-follower is what TikTok actually ranks?"
• Don't: "Followers don't matter anymore."
• Best on: X, LinkedIn. Questions out-perform statements on reply rate by roughly 2×.

5. The Unpopular Take

• Skeleton: "Everyone in [niche] is wrong about [topic]. Here's the proof."
• Filled: "Everyone in indie SaaS is wrong about Twitter. Here's a 90-day audit of 12 founder accounts showing why."
• Don't: "Twitter is overrated."
• Best on: X, LinkedIn. Strong claims with promised proof spike completion rate.

Family B — Numbers & Receipts (6–10)#

Numbers buy you specificity for free. They also buy you the perception of having actually done the work — the implied receipt is the hook.

6. The Specific Count

• Skeleton: "I [verb] [exact number] [things] in [time period]. Here's what worked."
• Filled: "I sent 412 cold emails in 30 days. Here's the 4-line opener that hit a 31% reply rate."
• Don't: "I sent a lot of cold emails. Here's what I learned."
• Best on: LinkedIn, X. Specific counts read as evidence; they outperform round numbers (100, 500, 1000) which read as estimates.

7. The Before / After Delta

• Skeleton: "[Metric] went from [before] to [after] in [time]. One change."
• Filled: "Newsletter open rate went from 18% to 47% in three weeks. One change: the subject line never mentions the topic."
• Don't: "How I improved my newsletter."
• Best on: LinkedIn, Substack Notes. The "one change" promise spikes the save signal.

8. The Cost Reveal

• Skeleton: "[Activity] cost me [dollar amount] before I figured out [insight]."
• Filled: "Facebook ads cost me $11,400 before I figured out that the creative was the targeting."
• Don't: "I spent a lot on ads and learned a lesson."
• Best on: LinkedIn, X. Money is the cleanest stake; dollar figures read as receipts.

9. The Time-Boxed Result

• Skeleton: "[Specific outcome] in [specific time window]. No [common shortcut]."
• Filled: "From 0 to 2,400 newsletter subscribers in 28 days. No giveaways, no paid ads, no Twitter."
• Don't: "How I grew my newsletter fast."
• Best on: YouTube Shorts, LinkedIn, Substack Notes. Time-bounded promises bait completion.

10. The Negative Stat

• Skeleton: "[Surprisingly high %] of [group] [do something self-defeating]. Don't be in it."
• Filled: "74% of newsletter writers send their welcome email more than 48 hours after signup. Don't be in the 74%."
• Don't: "Most people send their welcome email too late."
• Best on: LinkedIn, Substack. Negative stats spike save rate by triggering loss-aversion.

Family C — Pattern Interrupts (11–15)#

These break the visual or rhythmic expectation of the feed. They're the only family whose mechanic is mostly visual, not semantic.

11. The False Start

• Skeleton: "[Sentence that sounds like it's starting normally] — wait, no."
• Filled: "The best time to post on LinkedIn is Tuesday at 9am — wait, no, that's from a 2019 study and it's been wrong since 2023."
• Don't: "Posting times have changed."
• Best on: TikTok, Reels. The self-interrupt buys 1–2 extra seconds of dwell.

12. The One-Word Open

• Skeleton: "[Single charged word]. [Follow-up sentence.]"
• Filled: "Stop. If your last 5 LinkedIn posts opened with 'Excited to share', the algorithm has already decided."
• Don't: "Hi everyone, excited to share…"
• Best on: LinkedIn, X. One-word opens win the visual scan in a text-dense feed.

13. The Wrong Order

• Skeleton: "[The end of the story] — and that's how [the beginning of the story]."
• Filled: "I deleted my entire Twitter following — and that's how I tripled engagement."
• Don't: "Here's how I tripled engagement on Twitter."
• Best on: TikTok, Reels, YouTube Shorts. Reverse-order narrative creates a why-loop the brain has to close.

14. The Visual Mismatch

• Skeleton: "[Voiceover claim] / [Visual that contradicts it]" (video only)
• Filled: Voiceover: "This is the calmest job I've ever had." / Visual: laptop on fire metaphorically (rapid stress-cuts).
• Don't: Voiceover matches visual exactly.
• Best on: TikTok, Reels, YouTube Shorts. Audio-visual mismatch is the strongest known dwell trigger.

15. The Caps-Lock Cold-Open

• Skeleton: "[ONE FRAGMENT IN CAPS]. [Normal sentence following.]"
• Filled: "DELETE YOUR DRAFTS. The reason you can't ship is that you keep editing the same three half-posts."
• Don't: "YOU NEED TO STOP EDITING YOUR DRAFTS BECAUSE…"
• Best on: X, LinkedIn. Brief caps fragments scan as urgent; sustained caps scan as spam.

Family D — Curiosity Gaps (16–20)#

These open an information gap small enough to read as a tease, big enough to be worth closing.

16. The Missing Piece

• Skeleton: "Everyone tells you to [advice]. Nobody mentions [the part that matters]."
• Filled: "Everyone tells you to 'pick a niche'. Nobody mentions you have to pick the audience inside the niche."
• Don't: "Picking a niche is more nuanced than people say."
• Best on: LinkedIn, X, Substack. The missing-piece frame spikes save rate.

17. The Two-Door Question

• Skeleton: "There are two kinds of [group]: the ones who [behavior A] and the ones who [behavior B]. Only one [outcome]."
• Filled: "There are two kinds of newsletter writers: the ones who write to subscribers and the ones who write to algorithms. Only one keeps the list."
• Don't: "There are different approaches to newsletter writing."
• Best on: LinkedIn, Substack. Binary framings make readers self-categorize, which spikes dwell.

18. The Wrong-Answer-First

• Skeleton: "The obvious answer to [question] is [X]. The obvious answer is wrong."
• Filled: "The obvious answer to 'how do I grow on TikTok' is post daily. The obvious answer is wrong — here's what the top 1% actually do."
• Don't: "Posting daily isn't enough on TikTok."
• Best on: TikTok, YouTube Shorts. Promises a payoff that's worth completing.

19. The Half-Story

• Skeleton: "[Specific moment in the middle of a story]. I'll explain how I got there."
• Filled: "I was sitting in a Stripe dashboard at 2am watching a $14k refund queue and realizing the entire onboarding flow was wrong. I'll explain how I got there."
• Don't: "Let me tell you a story about onboarding."
• Best on: LinkedIn, Substack, Threads. In-media-res opens spike scroll-completion.

20. The Direct Question to Camera

• Skeleton: "Are you the [group member] who [specific recognizable behavior]?"
• Filled: "Are you the founder who keeps editing the homepage instead of running ads?"
• Don't: "Are you having trouble with marketing?"
• Best on: TikTok, Reels. Direct address spikes the first-second completion rate.

Family E — Status & Identity (21–25)#

These hooks attach to the reader's sense of who they are or want to be. They convert weakly to clicks but very strongly to saves and shares.

21. The Insider Tell

• Skeleton: "You can tell a [type of pro] by [specific small behavior]."
• Filled: "You can tell a senior product designer by how they name layers in Figma."
• Don't: "Senior product designers are very organized."
• Best on: LinkedIn, X. Insider-tells get bookmarked and shared inside team Slacks.

22. The Mirror

• Skeleton: "If you [specific recognizable habit], this is for you."
• Filled: "If you have a Notion doc called 'Newsletter ideas' with 47 entries and 3 published, this is for you."
• Don't: "This is for newsletter writers."
• Best on: LinkedIn, Substack, X. Hyper-specific mirroring spikes save and share.

23. The Niche Call-Out

• Skeleton: "Quick note for [hyper-specific group of 5,000–50,000 people]."
• Filled: "Quick note for solo founders who've raised under $200k and are doing all their own content."
• Don't: "Quick note for founders."
• Best on: LinkedIn, X. Tight call-outs convert better than broad ones because the matched reader feels seen.

24. The Senior-Level Mistake

• Skeleton: "[Specific senior title] mistake I made: [counterintuitive error]."
• Filled: "Director-of-marketing mistake I made: I optimized for MQLs my first quarter and torched the pipeline by month four."
• Don't: "Marketing mistake I made."
• Best on: LinkedIn. Senior-level error confessions spike comment rate.

25. The Quiet Flex

• Skeleton: "[Achievement] without [the usual cost]."
• Filled: "Replaced our $4,200/mo marketing agency with a single contractor and grew leads 22%."
• Don't: "We saved money on marketing."
• Best on: LinkedIn, X. Quiet flexes outperform loud ones on share rate.

Family F — Loss & Stakes (26–30)#

Loss-aversion is the strongest documented behavioral force in the marketing literature. Hooks that load loss into the first sentence keep the thumb still.

26. The Almost-Lost

• Skeleton: "I almost [bad outcome] because I didn't know [thing this post explains]."
• Filled: "I almost lost a $48k contract because I didn't know that DocuSign expires the link after 7 days by default."
• Don't: "I had a close call with a contract."
• Best on: LinkedIn, X. Near-miss stories pull a 1.5–2× engagement multiplier.

27. The Avoidable Cost

• Skeleton: "[Common mistake] costs the average [group] [dollar amount] per [time period]."
• Filled: "Reusing the same hook on LinkedIn and X costs the average creator about 40% of their LinkedIn reach per month."
• Don't: "You should write platform-specific hooks."
• Best on: LinkedIn, X. Quantified avoidable costs spike save rate.

28. The Deadline

• Skeleton: "You have until [specific date] to [action] before [consequence]."
• Filled: "You have until July 1, 2026 to migrate off the old Substack referral API before existing rewards reset to zero."
• Don't: "Substack is updating its API."
• Best on: X, LinkedIn. Real deadlines outperform vague urgency.

29. The Reputation Risk

• Skeleton: "This [common practice] makes you look [unwanted perception] to [audience that matters]."
• Filled: "Putting 'AI-generated' in your bio makes you look junior to every senior recruiter on LinkedIn — even when your work is great."
• Don't: "Be careful what you put in your bio."
• Best on: LinkedIn, X. Reputation-risk framings outperform tactical advice on save and share.

30. The Sunk Cost Salvage

• Skeleton: "If you've already [past investment], here's the one move that [recovers value]."
• Filled: "If you've already written 50 newsletters nobody read, here's the one move that turns them into a paid product."
• Don't: "How to repurpose old newsletters."
• Best on: Substack, LinkedIn. Sunk-cost framing converts weakly engaged readers.

Family G — How-To Promise (31–35)#

The most over-used family — but still works when the promise is specific, time-bounded, and credible.

31. The Numbered Promise

• Skeleton: "[Number] [specific things] you can do in [time window] to [outcome]."
• Filled: "Three changes you can make to your LinkedIn headline in the next 10 minutes that will measurably raise profile views."
• Don't: "Tips to improve your LinkedIn profile."
• Best on: LinkedIn, YouTube Shorts. The exact number plus the exact time window does the lift.

32. The Time-Bound How

• Skeleton: "How to [outcome] in [aggressive but credible time]."
• Filled: "How to ship a launch announcement on X that doesn't read like a press release — in 12 minutes."
• Don't: "How to write a good launch tweet."
• Best on: X, LinkedIn, YouTube Shorts.

33. The Skill Unlock

• Skeleton: "The one [skill / pattern / framework] that separates [novice] from [pro] in [domain]."
• Filled: "The one editing pattern that separates a junior TikTok creator from a senior one: the 0.4-second cut on every claim."
• Don't: "How to edit TikToks better."
• Best on: TikTok, YouTube Shorts.

34. The Tool Stack

• Skeleton: "The [exact number]-tool stack for [specific job], priced [total cost / month]."
• Filled: "The 3-tool stack I use to ship one trend-aware post per platform per day, for $43/month total."
• Don't: "Tools I use for content."
• Best on: LinkedIn, X, YouTube Shorts. Specific tool counts and dollar totals out-pull vague lists.

35. The Anti-Tutorial

• Skeleton: "How NOT to [common task] — the [number] mistakes that kill [outcome]."
• Filled: "How NOT to write a Substack headline — the 4 mistakes that kill inbox open rate."
• Don't: "Common mistakes in Substack headlines."
• Best on: Substack, LinkedIn. Negative framing outperforms positive framing on save rate by ~25%.

Family H — Story Cold-Opens (36–40)#

These work because narrative is the brain's compression format. A reader who's 8 seconds into a story is far more likely to finish than a reader who's 8 seconds into a list.

36. The In-Media-Res

• Skeleton: "It's [time] and I'm [doing something specific in the middle of a story]."
• Filled: "It's 11pm and I'm staring at a Stripe refund queue that just hit five figures because of a single broken checkbox in our pricing page."
• Don't: "Let me tell you about a bug we had."
• Best on: LinkedIn, Substack, Threads. In-media-res scenes spike scroll completion.

37. The Email Excerpt

• Skeleton: "An email I got this week: '[paraphrased line]'. Here's how I answered."
• Filled: "An email I got this week: 'I have 200 subscribers and I'm thinking of giving up.' Here's how I answered."
• Don't: "Lots of people ask me about subscriber count."
• Best on: Substack, LinkedIn. Real-correspondence framings read authentic.

38. The Overheard Line

• Skeleton: "Overheard at [specific place]: '[line]'. The speaker was wrong, and here's why it matters."
• Filled: "Overheard at a SaaS conference in Berlin last week: 'AI killed SEO.' The speaker was wrong, and here's why it matters for anyone with a content site."
• Don't: "People say AI killed SEO. Let me respond."
• Best on: X, LinkedIn. Quoted-speech opens outperform abstract claims.

39. The Dated Memory

• Skeleton: "On [exact date], [specific event]. It changed how I [behavior] permanently."
• Filled: "On March 4, 2024, a single LinkedIn post pulled 412 inbound leads in 72 hours. It changed how I write every opener since."
• Don't: "One post changed my LinkedIn strategy."
• Best on: LinkedIn, Substack. Exact dates buy specificity for free.

40. The DM Receipt

• Skeleton: "A DM I just got: '[verbatim line, anonymized]'. This is the answer."
• Filled: "A DM I just got: 'I post 5x a week and get under 200 views. What am I doing wrong?' This is the answer."
• Don't: "Here's what to do if you're not getting views."
• Best on: X, LinkedIn, Threads. DM-receipt opens scan as private knowledge made public.

Platform × family fit matrix#

The same template performs differently across platforms because each platform's feed ranks a different signal: TikTok ranks completion percentage, X ranks reply rate, LinkedIn ranks dwell-time plus "see more" clicks, Substack ranks inbox open rate, YouTube Shorts ranks first-30-second retention.

The 3-test pre-publish check#

Before you ship a hook, run it through these three tests. They take about 60 seconds combined and they catch most failures.

1. The read-aloud test. Read the hook aloud at normal pace. If it takes more than 3 seconds, it's too long for TikTok, Reels, or Shorts — the visible viewport will cut it before the verb arrives. For LinkedIn or Substack, the limit is 6 seconds; longer hooks read as captions, not openers.
2. The screenshot-blur test. Screenshot the post in the platform's preview. Apply a ~5px Gaussian blur (most photo apps have it). One phrase should still pop visually — a number, a name, an italic, a caps fragment. If everything blurs into a uniform grey, the hook has no visual hook.
3. The two-opener A/B. Post the same body twice, 24 hours apart, with two different hooks. Compare the 1-hour hold-rate (views with dwell > 3 seconds, divided by impressions), not the 24-hour like count. The winning hook usually wins inside the first hour; later metrics are confounded by share velocity.

Save the winner to a personal hook library and reuse the structure (not the words) across the next 10 posts. Recurring formats compound; one-off hits don't. (For the longer take on repeatability, see How to score trends for your niche and Why we killed the viral score.)

Using TINS HUB as your hook generator#

TINS HUB is a hook generator wired into a live discovery layer. The flow:

1. Set your niche. Six fields: niche, platform, audience, voice style, geography, and format, used to score every idea.
2. Generate a batch. The dashboard pulls live signals at request time (no stale 6-month training data) and returns a small set of trend briefs scored against your niche.
3. Get hooks pre-tuned per platform. Every brief ships with hook variants written to the voice of the platform you select — TikTok hooks read like TikTok, LinkedIn hooks read like LinkedIn.
4. Save winners to your library. Drag-and-drop, export to Markdown, sync across devices.

The Free tier ships with a signup-bonus credit pool so you can test a few batches before deciding. Paid tiers are listed on /pricing. For the long version of how live discovery beats prompt-only generators, see The 2026 guide to AI content idea generators. Or skim the live public trending board for what's surfacing today.

Close#

If your posts have been getting respectable numbers but never breaking through, the lever is almost always the opener. Pick one family above that fits the platform you care about most, run three hooks through the read-aloud and screenshot-blur tests, ship the winner, and compare 1-hour hold-rate against your last 5 posts. Then do it again next week. Generate your first batch in TINS HUB.