How-To
May 202610 min read

How to Build a LinkedIn Hook Library With AI

A LinkedIn hook library is the highest-leverage content asset a solopreneur can build. 50+ hooks in your voice, organised by type, pulled from every time you batch content. Two-hour build. Saves 30+ minutes per week of "staring at blank screen" forever.

A hook library is 50+ pre-written LinkedIn first-lines in your voice, organised by 7 hook types and 5-7 content pillars. Built in 2 hours using Custom GPT or Claude Project with your voice prompt loaded. Used every time you batch content. Eliminates the slowest part of writing — the blank-screen first-line decision — and produces consistently stronger posts because the hooks were generated when you had time to choose carefully, not under writing-time pressure.

Why hook libraries beat hook generators

Most solopreneurs try AI hook generators (Hookline, AuthoredUp, ChatGPT with a hook prompt) and get mixed results. The output is fine — sometimes good, sometimes generic — but it's produced under writing-time pressure when you're already in batch-content mode and don't want to wait for iterations.

A hook library inverts this. You generate hooks in a separate session, when you have time to filter aggressively. Then during batching, you pull from the library — every hook is pre-vetted, voice-matched, and ready. The selection time drops from 5-10 minutes per post to 30 seconds.

The other reason hook libraries beat generators: filtering. When you generate hooks live, you accept the second-best because you don't want to spend another minute. When you generate 70 hooks in advance, you can keep only the top 30 — variance moves up because you've got the luxury to be picky.

The 7 hook types every library needs

Type 1: Number-led

Open with a specific number, dollar amount, or count. The number signals specificity, which the algorithm reads as quality.

"47 hours of marketing this week. Three drafts. One published post."
"£17,000 sat in the wrong account for 11 weeks before anyone noticed."

Type 2: Question-led

A specific, contrarian, or counterintuitive question. Avoid generic "what do you think" questions — those signal low engagement.

"What if your content strategy isn't broken — your discovery process is?"
"Why are you paying a ghostwriter £4,000 a month to do what AI can do for £20?"

Type 3: Contrarian-statement

State a position that runs counter to common advice in your niche. Earns comments because people respond — agreement and disagreement both count.

"LinkedIn doesn't punish AI content. It punishes generic content. Most users can't tell the difference."
"Stop asking ChatGPT to 'write better.' Ask it to write more like a specific person."

Type 4: Named-example

Open with a specific person, business, or scenario. The named detail anchors the post in concrete reality before any abstraction.

"Jacob Olenick reverse-engineered his voice into one prompt. Three weeks later, prospects went from 'don't pitch me' to inbound leads."
"A coach client of mine spent £18,000 on ghostwriting last year. We built her voice system in 3 days for £497."

Type 5: Before-after

Set up a clear shift. Old way / new way. Used to think X / now think Y. Effective because the structure mirrors how readers actually update beliefs.

"I used to spend 14 hours a week on LinkedIn content. Now it's 90 minutes. Same output. Different system."
"Three months ago I thought voice prompts were marketing fluff. After 30 builds, I think they're the highest-leverage content infrastructure in 2026."

Type 6: Observation

A specific thing you noticed that others would miss. Reads as authentic because it requires you to have actually been there.

"Every consultant who tells me their content sounds generic also tells me they 'tried AI but it didn't work.' Same problem. Different framing."
"Three of the last five clients we built voice systems for had ghostwriters they were quietly trying to replace. Pattern: hourly value over £200, asset ownership matters."

Type 7: Framework-led

Open by naming a framework, system, or model. Most useful for consultants and B2B founders whose audiences expect frameworks.

"The Voice Build methodology has 5 sections. Most solopreneurs only get 2 of them right."
"There are 4 categories of done-for-you content service in 2026. Three of them are wrong for most solopreneurs."

The 2-hour build workflow

Step 1: Generate hooks (30 minutes). Open your Custom GPT or Claude Project (with voice prompt loaded). Run this prompt 7 times, once per hook type:

Give me 10 LinkedIn hooks of the [TYPE] format in my voice. Topic area: [pillar] Hook type: [number-led / question-led / contrarian-statement / named-example / before-after / observation / framework-led] Each hook should be 1-2 sentences max. Apply the LinkedIn voice rules from my voice prompt. Make each hook distinct from the others. Include at least one specific named detail per hook where relevant.

7 types × 10 hooks = 70 hooks in 30 minutes.

Step 2: Filter aggressively (30-45 minutes). Read all 70. Cut any that:

Typically you'll keep 30-45 of the 70 hooks. That's the right ratio. If you keep more than 50, you're not being picky enough.

Step 3: Organise (15-20 minutes). Open Notion, Airtable, or a Google Sheet. Columns:

Tag each hook. The tagging is what makes retrieval fast — when you batch content next week, you can filter by pillar and type in 5 seconds.

Step 4: Test against actual posts (20-30 minutes). Pick 3 hooks from the library. Build a full post around each one using your voice prompt. Read each post. If the hook didn't lead naturally to a strong body, the hook was weaker than it looked — kill it from the library.

Step 5: Document and commit to using it (5 minutes). Add the library link to your content batching workflow. Commit to pulling from it every batch. Set a quarterly reminder to refresh.

Using the library every batch

The discipline that makes this work: when you sit down to write content, the library is the first thing you open. Not your Notion ideas list. Not ChatGPT. The library.

Workflow: pick a content pillar for the batch. Filter the library to that pillar. Browse all hooks for that pillar. Pick 4-8. Build posts around them.

The output: posts where the hook is already strong before you write the body. Body quality is what you have headspace for, because you didn't burn 10 minutes per post deciding the first line.

Quarterly refresh

Every 3 months, run the build workflow again with adjustments:

30 minutes per quarter. The library compounds because the filter keeps tightening.

Common mistakes

Mistake 1: Generating without a voice prompt. Default AI hooks are generic. The voice prompt is non-negotiable.

Mistake 2: Keeping too many. 70 hooks all kept = 70 mediocre hooks. 30 hooks ruthlessly filtered = 30 strong hooks. Variance matters more than volume.

Mistake 3: Skipping the test step. Some hooks look strong but don't lead to strong posts. The test step catches these.

Mistake 4: Storing without tagging. An untagged 50-hook list is unusable in practice. Tagging is the leverage.

Mistake 5: Building once and never refreshing. Your voice evolves. Topics shift. The library has to evolve with you.

Related reading

Get the hook library built for you

DFY Voice System ships a 50+ hook library calibrated to your voice as part of the £497 build. Plus the voice prompt + Custom GPT + Claude Project + content batching workflow. Delivered in 2-3 working days.

See The Voice Build

Frequently Asked Questions

What is a LinkedIn hook library?

50+ pre-written first-line hooks calibrated to your voice and content pillars. Built once, used every batch.

How do you generate hooks with AI?

Custom GPT or Claude Project with voice prompt loaded. Generate 7-10 hooks across 7 types. Filter to the strongest 30-40.

How many hooks should be in a library?

50-60 hooks across 5-7 pillars and 7 types. Below 30: dries up fast. Above 100: too much browsing.

How often should I refresh?

Quarterly. 30 minutes per refresh.

Where should I store the library?

Notion, Airtable, Google Sheets — anywhere searchable with type + pillar tagging.