GuidesSocial

The Claude social growth kit: 11 content builds

Built from your own posts, comments and analytics.

Eleven systems for the content work you repeat: repurposing, hook testing, gap analysis, comment mining, recycling, brand deals and performance analysis. Copy-paste prompts, what to feed each one, and the four that work together as a single operation.

Free. No email required, nothing to download.

What you get
  • Eleven builds, each with a copy-paste prompt and what to feed it
  • The four that chain together into one content operation
  • Where to keep a build so it improves instead of resetting every time
  • A level-up for each one, for when the basic version is working
  • The honest limits, including what no AI can tell you about a hook

Six steps, and the one everybody skips

You do not need eleven new tools. You need your own posts, your comments, your analytics and your brand information in one place that Claude can read.

Every build below follows the same six steps, and step four is the one that decides whether any of this lasts.

  1. 01
    Start with the prompt

    Copy the one from the build you picked. It works before you customise anything.

  2. 02
    Add your context

    Your posts, your analytics, your audience, your voice, your past results. This is the step that separates useful output from generic output.

  3. 03
    Correct it

    Tell it plainly: this part is useful, this part is not, here is what I would have written. Your corrections are the most valuable context you will ever give it.

  4. 04
    Save it as a system

    Put the prompt in a Claude Project as the project instructions and add your files as project knowledge. Skip this and you are starting from zero every time, which is why most people try one of these once.

  5. 05
    Add data

    A good prompt helps. Your own numbers change what it can actually tell you. The performance and recycling builds are nothing without them.

  6. 06
    Automate last

    Only once the workflow has proved itself. Do not automate a bad workflow, you will just get more of it.

Three of these builds keep growing: the viral database, the recycling tracker and the performance dashboard. Those need somewhere real to live, like a Google Sheet or a Notion database you connect, not a chat you scroll back through. Connectors for Drive, Notion and Sheets are in the directory at claude.ai/directory.

Start with one. Get it working on real content this week. Then improve it.

Content repurposing machine

Takes one piece of content and rebuilds it for each platform, rather than trimming the same words to fit.

The difference matters. A shortened LinkedIn post is not a Threads post, and everyone can tell.

Feed it:

  • An original transcript, usually from a short-form video
  • Five to ten posts you wrote yourself, so it has something to match
  • The platforms you actually publish on
  • Your usual CTA style, and the phrases you hate
  • Your content pillars
01

Repurpose one piece of content

Give it the transcript and your own examples. Without the examples it will write competent content that sounds like nobody.

The prompt
Help me build a reusable content repurposing system.

I will give you one original piece of content, usually a short-form video
transcript. Turn it into platform-native versions for:
- TikTok or Reels caption
- YouTube Shorts title
- LinkedIn post
- Threads post
- X post

Do not copy or simply shorten the same content. Adapt the opening, structure,
length and CTA to how people actually consume content on each platform.

Preserve the original idea and my tone.

Before generating anything, identify:
1. The core idea
2. The strongest insight
3. The most interesting hook
4. The main takeaway for the audience

Then write the platform versions.

Avoid generic AI phrasing, unnecessary emojis and exaggerated claims. Do not add
statistics, results or examples that were not in my original content.
Result

Five versions that read as though they were written for each platform, from one transcript.

Level it up. Give it modes so it knows what kind of content it is handling before it adapts anything: educational, opinion, news, promotional, tutorial. An opinion post and a tutorial should not come out with the same structure.

Viral video database

Turns the videos you save and never look at again into research that tells you why they worked.

This is the build with the highest ceiling in the guide, and it is worthless for the first two weeks. It needs volume before it says anything useful.

Feed it:

  • Creator, platform and topic
  • The hook, word for word
  • The transcript
  • Format and length
  • Views, and likes or comments where you can see them
  • The CTA, and what actually made you stop scrolling
02

Analyse and log a video

Run this on every video you save. The value is in the pattern across thirty of them, not in any single analysis.

The prompt
Help me create a Viral Video Research Database.

Every time I give you a successful social post or transcript, analyse it and
record:
- Topic
- Target audience
- Hook, quoted exactly
- Hook type
- Format
- Main promise
- Structure
- What creates the curiosity
- Proof used
- CTA
- Why someone would keep watching
- Elements worth adapting

Do not tell me a video worked because it had a strong hook. Explain specifically
what the hook or the structure is doing.

Keep the categories consistent between entries so patterns become visible over
time.

Do not copy another creator's wording. Extract the underlying format or principle
that I could apply to an original idea of my own.
Result

A swipe file that explains mechanisms instead of a folder of links you never reopen.

Level it up. Once you have thirty or more entries, ask it to rank the recurring hook structures, formats, topics and CTAs by how often they appear, then propose original experiments using the same principles. That is the point it turns from a collection into research.

Content gap finder

Compares what you publish against what your audience keeps asking about, what performs, and what your competitors cover.

Most content plans repeat the creator's favourite three topics. This is how you notice.

Feed it:

  • Your last thirty to a hundred posts
  • Your content pillars
  • Competitor posts or accounts in your niche
  • Who your audience actually is
  • Comments and questions you get repeatedly
  • Performance data if you have it
03

Find the gaps

The instruction not to recommend a topic just because a competitor posted it is doing real work here. Otherwise you get their content calendar.

The prompt
Act as my content gap analyst.

I will give you:
1. My recent content
2. My target audience
3. My main content pillars
4. Competitor content or examples from my niche

Analyse all of it and identify:
- Topics I cover heavily
- Topics I have neglected
- Audience problems I am not addressing
- Questions competitors answer that I do not
- Topics competitors are ignoring
- Successful topics worth going deeper on
- Opportunities for a recurring series

Do not recommend a topic simply because a competitor posted about it. Prioritise
gaps that overlap with my audience's needs and my positioning.

Finish with twenty specific content opportunities, split into quick wins,
evergreen ideas, experiments and potential recurring series.

Where you are inferring an audience problem rather than seeing it in what I gave
you, say so.
Result

Twenty opportunities sorted by type, and a clear view of the topic you have quietly been avoiding.

Level it up. Run it monthly and compare four columns: what you planned, what you published, what performed, what is still missing. The gap between the first two is usually the real problem.

Audience comment miner

Uses your own audience to tell you what to make next, in their words rather than yours.

The phrasing matters as much as the topic. People tell you their exact search terms in the comments and most creators scroll past them.

Feed it:

  • Comments from TikTok, Instagram, YouTube and LinkedIn
  • Reddit or community threads in your niche
  • DMs, where it is appropriate to use them
  • The comments on your worst performing posts, which are often the most honest
04

Mine the comments

Paste comments in bulk. Hundreds at a time is better than a curated handful, because you are looking for repetition.

The prompt
Analyse these audience comments as customer and content research.

Categorise them into:
- Questions
- Problems
- Confusion
- Objections
- Requests
- Strong opinions
- Success stories
- Content suggestions

Look for repeated themes, and note the exact language people use.

Then give me:
1. The ten most important audience insights
2. Questions worth answering with a video each
3. Potential hooks using the audience's own words
4. Follow-up opportunities from my existing content
5. Topics people seem especially interested in or confused about

Do not invent audience problems that are not supported by the comments. If a theme
appears only once, say so rather than presenting it as a pattern.
Result

Content ideas in your audience's own language, which is usually better than anything you would have phrased yourself.

Level it up. Have it maintain a chain: comment, insight, content idea, video made, how it performed. After a few months you can see which kinds of audience signal actually turn into good content and which ones you imagined.

Brand voice checker

Not a write-in-my-voice prompt. This one checks work you have already written and tells you where it stopped sounding like you.

Build the profile once, then use it as an editor on everything.

Feed it:

  • Ten to twenty things you have written, not your best, your most typical
  • Video transcripts, which capture how you actually talk
  • Captions, emails and posts
  • Examples you love from other people
  • Examples of your own that read as too AI-generated
05

Build the voice profile, then check drafts against it

Run the first half once to create the profile. Keep it, then paste drafts in against it.

The prompt
Help me build a Brand Voice Profile from the writing samples I provide.

Analyse the recurring characteristics, including:
- Sentence length
- Vocabulary
- Tone and pacing
- Formatting
- Humour
- Level of formality
- How I explain complicated ideas
- Hook style
- CTA style
- Phrases I use often
- Things I consistently avoid

Turn that into clear writing rules, each one based on evidence you can point to in
my samples.

Do not invent personality traits that are not demonstrated in the examples. Where
my samples are inconsistent, tell me rather than picking a rule.

Once the profile exists, I will give you new drafts. For each one:
1. Identify what does not match my voice
2. Explain why
3. Suggest specific changes
4. Rewrite only the parts that need it

Preserve the sections that already work instead of rewriting everything.
Result

A profile built from evidence, and edits that leave your good sentences alone.

Level it up. Use a plain checklist rather than a score: sounds like me, too polished, too generic, too long, too much jargon, weak opening, CTA does not fit. Resist asking for a percentage. A confident 87 percent brand voice score is invented precision and tells you nothing you can act on.

30 day content planner

Builds a calendar from strategy instead of producing thirty unrelated ideas you will not make.

Feed it:

  • Your audience and their main problems
  • Content pillars
  • Products, services or launches coming up
  • Your recent posts and your previous best performers
  • How often you can realistically post
  • The platforms you are planning for
06

Plan the month

The instruction not to generate the calendar immediately is the whole trick. Ask for thirty ideas and you get thirty ideas, unconnected to anything.

The prompt
Help me build a 30 day social content plan.

Do not generate the calendar immediately. First analyse:
- My target audience
- Their main problems
- My content pillars
- My recent content
- My previous high performers
- Topics I have overused
- Topics I have neglected
- Any upcoming launches or events

Then create a balanced 30 day plan. For every idea include:
- Topic
- Content pillar
- Format
- Working hook
- The audience pain point or desire it targets
- The goal: reach, engagement, authority or conversion
- A CTA where one fits

Do not repeat the same idea with different wording. Make roughly one in five ideas
genuinely experimental rather than a variation on something that already worked.

Do not promise performance for any idea. Flag which ones are proven formats and
which are untested.
Result

A month of content where every slot has a reason, a format and a goal attached.

Level it up. Weight the mix deliberately: about 70 percent proven formats, 20 percent adjacent experiments, 10 percent completely new tests. Those numbers are not a law of nature, just a useful frame that stops a calendar becoming either repetitive or random.

The custom version

Guides like this are the generic version.

Everyone reading this page gets the same system. We build content systems that learn your voice from what you have actually published, know who you are writing for, and make every piece better because of how the last one performed. Then we either hand it to your team or run it for you.

Our systems drive results
267K+
Followers built
8M
Views in 7 months
4
Channels in parallel

Hook testing lab

Creates genuinely different hooks for one idea, using different angles rather than reworded versions of the same sentence.

Worth being clear about what this cannot do. Nothing can tell you which hook will win before people see it. What it can do is make sure your three tests are actually testing different things.

Feed it:

  • The video idea, in a sentence
  • Who it is for
  • The real payoff for watching
  • Any proof or specific result you can use
  • Hooks of your own that have worked before
07

Generate hooks worth testing

Then go and test them. The recommendation at the end is about clarity and fit, not a prediction.

The prompt
Act as my Hook Testing Lab.

I will give you a social video idea. First identify:
- The audience
- The core payoff
- The main pain or problem
- The most surprising element
- The strongest proof available

Then write hook variations using distinctly different approaches:
- Direct payoff
- Curiosity
- Pain or problem
- Contrarian
- Demonstration
- Specific result
- Mistake or warning
- Story

Keep each one usable in the first one to three seconds of a short-form video.

Do not produce minor wording variations of the same hook. For each one, explain in
a sentence what makes the angle different.

Finally, recommend the three most worth testing based on clarity and relevance to
the audience. Do not claim to know which will perform best, because you cannot.
Result

Eight angles that are actually different, and three worth putting in front of people.

Level it up. Log the results: hook, video, views, retention, shares, saves. Feed those back in later and it can analyse your own hook performance instead of reasoning from theory. Your account's data beats general principles every time.

Content recycling tracker

Finds the old posts worth reposting, updating, rescripting or rebuilding with a new hook.

Most creators sit on a back catalogue of proven ideas and keep inventing new ones from scratch.

Feed it:

  • Your historical posts with dates, topics and hooks
  • Views, saves, shares and comments
  • Which platform each one went on
  • Whether the information is still accurate
08

Sort the back catalogue

Needs your actual numbers to be useful. Without them it is guessing at which posts deserve another run.

The prompt
Help me build a Content Recycling Tracker.

Analyse my historical content and classify each post as one of:
- Leave alone
- Repost
- Update
- Rescript
- New hook test
- Create a part two
- Repurpose to another platform

Consider previous performance, how old it is, whether the information is still
current, how the audience responded, and whether the topic could carry another
angle.

Explain the reason for every recommendation.

Prioritise ideas where there is already evidence of audience interest, rather than
recycling something purely because it is old. If the data is too thin to judge a
post, say so instead of guessing.
Result

A ranked queue of content worth bringing back, with a reason attached to each one.

Level it up. Run it monthly and keep a standing content worth bringing back queue. This gets dramatically more useful once you have a few hundred posts, which is exactly the point at which you can no longer remember them.

Brand deal rate calculator

Organises the factors that decide what you should quote, using your own baseline rather than a made up market rate.

Built carefully, because the wrong version of this confidently invents industry standard pricing that does not exist.

Feed it:

  • Your average views and engagement, per platform
  • Your own previous rates, which are the actual baseline
  • Deliverables and how much production each takes
  • Usage rights, duration and whether they want paid amplification
  • Exclusivity, revisions and turnaround
09

Price a deal

It starts from your numbers. If you have never quoted before, your first few quotes are estimates and this makes them consistent, not correct.

The prompt
Help me build a Brand Deal Pricing Calculator for my creator business.

Do not pretend there is one objectively correct market rate, because there is not.
Instead, help me build a quote from my own baseline rates plus the additional work
and rights involved.

Ask me for:
- Platform
- Deliverables
- Production complexity
- Expected revisions
- Usage rights and duration
- Paid advertising or whitelisting
- Exclusivity
- Turnaround
- Cross-posting
- Any raw assets required

Start from the baseline prices I give you. Then show clearly:
- Base fee
- Add-ons
- Rights and licensing costs
- Total proposed quote
- Every assumption you made

Never invent industry standard rates. If I have not given you a benchmark, say you
do not have one rather than producing a number that sounds authoritative.
Result

A quote you can explain line by line, built from your rates rather than someone else's blog post.

Level it up. Log every deal: quote, negotiated price, accepted or rejected, how much work it turned out to be, how it performed. After a dozen you are pricing from your own evidence, which is the only pricing data that actually describes your audience.

Sponsorship brief analyser

Pulls what you are actually agreeing to out of a brief, including the parts buried on page four.

Usage rights and exclusivity are where creators lose money, and they are never in the headline.

Feed it:

  • The brief or agreement, uploaded
  • Your own standard terms, once you have them
  • What you normally charge and normally allow
10

Break down a sponsorship brief

Upload the document and run it before you reply to the email, not after.

The prompt
Act as my sponsorship brief analyser.

I will give you a creator campaign brief or agreement. Extract and organise:

Deliverables
Videos or posts required, platforms, length and format requirements.

Creative requirements
Mandatory talking points, required visuals, claims, CTAs, hashtags and tags.

Timeline
Draft deadline, feedback period, publishing date.

Commercial terms
Payment, usage rights, usage duration, paid advertising or whitelisting,
exclusivity.

Approval
Number of revisions and the approval process.

Questions and ambiguities
Anything unclear, missing, or worth clarifying before work starts.

Do not give me legal conclusions. Clearly separate what the document actually says
from questions or concerns you are raising. Quote the exact wording for anything
about rights, exclusivity or payment.
Result

Everything you are committing to, on one page, with the ambiguities listed before you sign.

Level it up. Feed it your own standard terms: normal revisions, standard usage period, typical turnaround, baseline rates. Then it can compare every incoming brief against what you usually accept and tell you what is unusual. For anything high value or unfamiliar, get a person who does contracts for a living to read it. This is a brief analyser, not a lawyer.

Content performance dashboard

Answers what is actually working, rather than telling you your average view count.

The strongest build here, and the one that needs the most real data. It is also the only one that can correct the other ten.

Feed it:

  • Per post: date, platform, topic, hook, format and length
  • Views, watch time and completion rate where you can get it
  • Likes, comments, shares and saves
  • Followers gained
  • The CTA you used
  • Exports from TikTok, Instagram Insights or YouTube Analytics
11

Analyse what worked

Start with whatever metrics you can actually export. Incomplete data analysed honestly beats complete data you never gather.

The prompt
Act as my social content performance analyst.

I will give you data from my recent posts. Do not simply summarise the numbers.

Analyse the relationships between topics, hooks, formats, video length, posting
patterns, views, watch behaviour, saves, shares, comments and followers gained.

Then identify:

Winners
Content that meaningfully outperformed my normal results.

Underperformers
Content that significantly underperformed.

Patterns
Recurring characteristics associated with stronger or weaker performance.

Uncertainty
Where there is not enough data to identify a pattern with any confidence. Be
explicit about this rather than filling it in.

Experiments
Ten specific things worth testing next.

Do not treat correlation as causation. Compare performance against my own baseline
rather than industry benchmarks, and tell me when a difference is small enough to
be noise.
Result

Winners, misses, real patterns, an honest list of what the data cannot tell you yet, and ten things to test.

Level it up. Run it weekly in a fixed shape: three winners, three misses, three patterns, three experiments, ten ideas for next week. A fixed shape is what makes it comparable week to week, which is where the value actually comes from.

The four that work as one system

If you only build four, build these, because they feed each other. Each one answers a question the next one needs.

  1. 01
    Viral video database

    Collects what is working elsewhere, and why, as mechanisms rather than clips to copy.

  2. 02
    Audience comment miner

    Tells you what your own audience is asking for, in the words they use.

  3. 03
    Content gap finder

    Turns both of those into what you should actually make next.

  4. 04
    Content performance dashboard

    Tells you what happened, which corrects the other three instead of letting them run on theory.

Then the results go back into the database and the gap finder, and the loop tightens. That is a content operation rather than eleven prompts in a folder.

The order matters. Most people start with the planner because planning feels productive, then plan from instinct because nothing upstream is feeding it.

Common questions

  • Do I need Claude Pro for these? No. All eleven work on the free plan, and Projects are available there too. A paid plan buys you more usage and larger uploads, which matters most for the performance dashboard and the viral database, where you are feeding in a lot of data at once.
  • Can Claude tell me which hook will go viral? No, and any tool claiming otherwise is guessing. Nothing can know how an audience will react before they react. What the hook build does is make sure your three tests are genuinely different angles rather than three rewordings, so the test tells you something. Then your own results are the only evidence that counts.
  • Do I need analytics exports to start? Not for most of them. Repurposing, hooks, comment mining, voice and briefs need no numbers at all. The performance dashboard, the recycling tracker and the gap finder get much better with real data, so start with what you can export today and add to it.
  • Will this just copy other creators? Only if you let it. The viral database prompt specifically tells it to extract the underlying format or principle and not to reuse anyone else’s wording. That distinction is the whole difference between research and plagiarism, so keep that line when you adapt the prompt.
  • Does the rate calculator know standard creator rates? No, deliberately. There is no single market rate, and a confident invented number is worse than no number. It works from the baseline rates you give it and shows how deliverables, usage rights, exclusivity and revisions change the quote, so you can explain every line of it.
  • Which build should I start with? Whatever is costing you the most time right now. Repurposing if you publish the same idea across platforms by hand. Comment mining if you are short of ideas. The performance dashboard if you are posting steadily and cannot tell what is working. Do not start with the 30 day planner, because it is only as good as what feeds it.
  • How is this different from just asking Claude for content ideas? Context and persistence. A cold request gets you generic ideas because it knows nothing about your audience, your voice or what already worked for you. These are saved as Projects with your posts, examples and numbers attached, so the eleventh time you use one it is working from evidence instead of averages.

Want this built around your business?

We build content systems for small businesses that are not AI native. Tell us your content problem and we will come back within 48 hours with what yours would include.

← All guides