The VoiceMoat blog
Growth systems, the craft of writing, and how the Twitter and LinkedIn algorithms actually rank you. No virality hacks.
You can get ChatGPT, Claude, or Gemini to write in your voice, but only for one session and only to about 70 to 80 percent, because chat models forget your style between chats and never learn from your edits. This is the honest cross-model guide: the analyze-then-write prompts that work, the settings that make them stick, the negative-constraint list that kills the AI tells, and the persistent alternative for anyone who posts regularly.
A LinkedIn-first, model-agnostic system for making ChatGPT, Claude, Copilot, and Gemini sound like you across posts and comments: a copy-paste custom instructions block, a banned-word list, comment-prompt packs, and an honest look at why the voice keeps resetting.
Use Claude Projects, Custom Styles, and Skills to write LinkedIn posts and tweets in your voice, understand why Claude forgets your style between sessions, and see how to make that voice persist across accounts and platforms.
AI tweets sound generic because models write the statistical average of every tweet ever posted. Here are the exact tells, the manual fixes, and the faster way to tweet in your real voice.
AI LinkedIn posts sound the same because general models return the statistical average: safe, agreeable, template-shaped English with no memory of how you write. Here is the template, the tells by model, and the durable fix.
Kill the em dash and the dozen quieter AI tells. A practical de-AI playbook for tweets and LinkedIn posts, plus why symptom-scrubbing never fully persists across a real posting cadence.
A reproducible head-to-head that runs the same prompts through Claude and ChatGPT on LinkedIn and Twitter, scored by platform, plus the variable both tools miss: whose voice they write in.
Yes, Claude can draft LinkedIn posts and tweets well, but it has no memory of how you write and cannot publish to either platform. Here is what it does, what it does not, and how to sound like you.
There is no single best AI for writing tweets; there is a best one per job. Grok wins on live Twitter data and banter, ChatGPT is the safest all-rounder, Claude writes the cleanest long threads, and DeepSeek is the budget pick. This is the honest 2026 head-to-head: tone, real-time X access, pricing, free tiers, and a best-use-case table, plus the one thing all four quietly miss, which is that every draft comes out in a capable stranger's voice, not yours.
We ran ChatGPT, Gemini, Copilot, Perplexity, and DeepSeek against the same LinkedIn post brief and scored them on hook, structure, tone, and originality. ChatGPT wins the general-assistant race, but every model failed the same dimension: sounding like the actual author.
A practical prompt-pack and workflow for turning Gemini, Grok, Copilot, Perplexity, and Claude output into social posts that read human and sound like you, plus the voice layer every guide leaves out.
A copy-paste pack of the best ChatGPT prompts for Twitter tweets, threads, and growth, plus the honest reason they still sound generic and the faster fix.