Best Claude Prompts for Social Media

Why does AI-generated social media content so often sound exactly the same no matter which brand it's supposedly written for? Scroll through a few "AI-written" LinkedIn posts and you'll notice the same rhythm: a punchy one-line hook, a string of short paragraphs, an emoji-bulleted list, and a call-to-action question at the end. It's not that the AI is incapable of variation — it's that most prompts never specify what should make this brand's voice different from any other brand's voice.

Quick-Start (Copy This Right Now)

Write a LinkedIn post about [topic] for [brand/person].

Our voice: [be specific — e.g., "skeptical of hype, uses concrete numbers instead of adjectives, short sentences, no emojis"]
Audience: [specific — e.g., "B2B SaaS founders who've heard every growth tactic already"]
Hook style: [pick one — contrarian opener, surprising stat, specific story, direct question]

Avoid: emoji bullet lists, "Here's the thing:", rhetorical questions as the hook, ending with "Thoughts?"

What this does: The explicit "avoid" list targets the exact tics that make AI-generated LinkedIn content instantly recognizable as AI-generated — naming them directly works far better than just asking for content that "sounds human."

Understanding the Variables

The voice description is the variable that matters most, and it needs to be specific enough to actually constrain word choice — "professional but friendly" is too vague to change anything, while "uses concrete numbers instead of adjectives" gives Claude something concrete to optimize against. If you're not sure how to describe your own voice, paste 2-3 of your best-performing past posts and ask Claude to describe the pattern back to you first.

⚠️ Common mistake: Asking for "a viral LinkedIn post." Virality isn't a writing instruction Claude can execute on — it depends on timing, algorithm behavior, and luck. Ask instead for a post built around a specific hook style and a clear, single idea; that's the part actually within your control.

Step-by-Step: Building a Twitter/X Thread

Turn this idea into a Twitter/X thread: [core idea or paste a longer piece of content to extract from]

Structure:
- Tweet 1: hook, no context needed, must work as a standalone tweet
- Tweets 2-6: one idea per tweet, building logically
- Final tweet: a clear takeaway, not a generic "thanks for reading" or engagement-bait question

Each tweet under 240 characters. Don't use numbered list format (1/, 2/) — let the ideas flow naturally instead.

What this does: Requiring tweet 1 to work as a standalone hook (rather than depending on the rest of the thread) follows how threads actually get distributed — most people see the first tweet in their feed before deciding to click "show this thread," so it has to do real work alone.

⚡ Pro tip: After Claude drafts a thread, ask it to identify which single tweet is the weakest link and could cause people to drop off, then revise just that one. This catches the "thread fatigue" problem (people losing interest by tweet 4) without rewriting the whole thread from scratch.

Pro-Level Variations

Repurposing one piece of content across platforms:

Here's a blog post: [paste or summarize]

Turn it into: (1) a 3-tweet thread teaser linking to the full post, (2) a LinkedIn post that stands alone (doesn't require clicking through), (3) an Instagram caption under 150 characters for a carousel post covering the same idea.

Each version should feel native to its platform, not like the same text copy-pasted with length cut down.

What this does: Specifying that the LinkedIn version should "stand alone" while the Twitter version is explicitly a "teaser" reflects a real platform behavior difference — LinkedIn algorithmically penalizes posts that send people away from the platform, while Twitter threads are built around the click-through. A single repurposing prompt without this distinction tends to produce one generic version forced into three formats.

Three Real Scenarios

A B2B founder posting solo on LinkedIn used the voice-anchored hook prompt to break out of a pattern he hadn't noticed himself falling into — every post he'd written for months opened with a rhetorical question, which had quietly become predictable to his regular readers even though each individual post seemed fine in isolation. Forcing a specific, varied hook style per post (sometimes contrarian, sometimes a direct story) brought back engagement that had been slowly declining.

A small agency managing five client social accounts used the explicit voice-description requirement to solve a real operational problem: junior team members posting on behalf of clients kept defaulting to a generic "agency voice" regardless of which client's account they were working on. Requiring a specific written voice description (with example posts) per client, fed into every prompt, made the output distinguishable between accounts again.

A podcast host repurposing long-form episodes into social content used the platform-native repurposing prompt to stop a pattern where the same Twitter-style summary was getting copy-pasted into LinkedIn and Instagram with minimal adjustment, performing poorly on both because it wasn't actually native to either format. Treating each platform's version as a distinct creative brief, rather than one summary force-fit into three lengths, noticeably improved engagement on the LinkedIn version specifically.

Batching a Week of Posts Without Them All Sounding the Same

I need 5 posts for [platform] this week, all loosely about [theme], for [brand voice — reuse your voice description from earlier].

Give each post a genuinely different angle: one personal story, one contrarian take, one practical tip, one question to the audience, one behind-the-scenes detail. Don't let them blur into variations of the same idea — if two angles end up too similar, flag it yourself and suggest a third angle instead.

What this does: Explicitly naming five distinct angle types up front, rather than asking for "5 posts about X," is what prevents batch-generated content from reading as repetitive — and asking Claude to self-flag overlap catches drift that's easy to miss when reviewing five posts in sequence rather than side by side.

When a Hashtag or Trend Reference Might Be Outdated

One specific risk worth flagging for social content: trends, slang, and platform features move fast, and a model's sense of "what's currently trending" can lag behind what's actually current by the time you're posting. If a prompt result references a specific trend, format, or hashtag you're not personally familiar with as currently active, it's worth a quick check before publishing rather than assuming it's still relevant — platform trends can go stale within weeks, and posting something that reads as out of date undercuts the credibility a trend reference was supposed to add.

The safer pattern is to treat AI-generated trend or format references as a starting hypothesis rather than a confirmed fact, then verify against the actual platform yourself before incorporating them — a thirty-second scroll through the relevant hashtag or format on the platform itself tells you immediately whether something is still active, which is faster and more reliable than asking any tool to confirm current trend status.

This same caution extends to platform features and best practices generally — character limits, algorithm preferences, and what formats a given platform currently favors all shift more often than most people expect, so it's worth a quick check against the platform's own current guidance before locking in a content strategy built around a specific feature or format constraint. A content calendar built around a feature that quietly got deprecated last quarter wastes more time than the few minutes it takes to confirm the feature is still live before planning around it.

Troubleshooting Common Issues

If output still sounds generic after specifying voice, the most likely fix is providing actual examples rather than more description. Telling Claude "we're witty" produces different results than pasting three captions that demonstrate your specific brand of wit — concrete examples anchor tone far more reliably than adjectives describing tone.

⚠️ Common mistake: Running the same prompt repeatedly expecting different angles. If you need 5 post ideas on one topic, ask explicitly for 5 different angles (contrarian take, personal story, data point, common mistake, framework) rather than just asking for "5 posts" — otherwise you'll get five versions of the same idea.

Your Turn

Once you've found a voice description and hook-style combination that actually produces content matching your brand, it's worth locking that prompt down rather than rewriting your voice guidelines from memory every time you need a post. A shared library — something like PromptABCD — keeps your team's best claude prompts for social media consistent across whoever's actually posting that week.