Humanizer

AI-written text is often easy to spot even when there is nothing technically wrong with it.

The sentences are a little too polished. The transitions are predictable. Every paragraph lands neatly. Certain words keep showing up. Sometimes the writing says a lot without sounding like anyone in particular actually wrote it.

Humanizer is built to fix that.

Give it a draft and it looks for the patterns that tend to make AI writing feel like AI writing. Then it rewrites the parts that need work while keeping what you were actually trying to say.

What it catches

The skill checks for 34 common AI writing patterns.

Some are obvious, like filler phrases, excessive headings, repeated AI vocabulary, fake enthusiasm, or dramatic conclusions.

Others are easier to miss: sentences that all have the same rhythm, unnecessary em dashes, forced groups of three, vague claims like "experts say," overly tidy transitions, awkward metaphors, constant hedging, or that habit of turning every simple point into something "pivotal."

It also catches chatbot language that sometimes slips into finished copy, such as "Great question," "I hope this helps," or "Let me know if you'd like..."

The point is not to blindly ban a list of words. Context still matters. Sometimes an em dash is fine. Sometimes a technical term is exactly the right term. Humanizer looks at how the whole piece reads.

It keeps your meaning

Humanizing a piece of writing should not mean making everything casual.

A technical document can stay technical. A work email can stay professional. A research summary should not suddenly sound like a Reddit comment.

Humanizer keeps the purpose and tone of the original, then works on the parts that feel stiff, generic, or strangely polished.

That might mean shortening a sentence. It might mean using a simpler word. Sometimes it means leaving a sentence alone.

Make it sound like you

You can also give Humanizer a sample of your own writing.

It pays attention to how long your sentences tend to be, the words you actually use, how you punctuate things, how you start paragraphs, and how you move from one thought to the next.

If you tend to write short sentences, it will not turn them into polished mini-essays. If you say "stuff," it does not automatically upgrade it to "components."

Without a sample, it aims for writing that feels natural and varied rather than perfectly assembled.

Where it works well

Humanizer is useful for drafts that are already basically correct but still feel AI-written:

• Blog posts and articles
• Emails and memos
• Essays and reports
• Social posts
• Product and marketing copy
• Documentation
• PR descriptions
• Resume bullets

You can also use it as a final pass before publishing something you wrote with ChatGPT, Claude, Gemini, or another LLM.

One last pass

Humanizer does not stop after the first rewrite.

It reads the result again and asks a useful question: what still makes this obviously AI-generated?

If something still feels too tidy, generic, performative, or formulaic, it gets another pass.

Because removing a few suspicious words is easy. Making the whole thing sound like somebody actually sat down and wrote it is the harder part.

Source

Humanizer was created by Siqi Chen (@blader) and ported to Hermes Agent. The Hermes version includes the original Humanizer patterns along with additional checks added by the Hermes team. It is released under the MIT License.

Source:
https://github.com/NousResearch/hermes-agent/tree/main/skills/creative/humanizer