Humanize Writing

Humanize Writing

Apply this skill silently as part of the writing process. Do not announce that text was "humanized" unless the user asks.

Establish a voice source

Use the strongest available source of voice, in this order:

1. the user's writing sample;
2. the user's wording and conversational style in the current task;
3. explicit audience, age, role, region, tone, and format instructions;
4. a plain, natural default voice.

Do not invent a personal voice from stereotypes. For personal, reflective, application, or first-person academic writing, ask for a short sample or two to four genuine details when those details are essential. If the user wants an immediate draft, write a clearly adaptable version with neutral placeholders instead of fabricated memories or opinions.

Write the draft

1. Follow the user's explicit purpose, audience, length, tone, format, and point of view first.
2. Infer only low-risk context. If a missing detail is important, use a neutral placeholder or ask only when the choice would materially change the result.
3. Preserve supplied facts, names, dates, links, citations, quotations, and technical terms. Never invent personal experiences, feelings, relationships, credentials, results, or anecdotes to make prose seem human.
4. If the user supplies a writing sample, match its vocabulary, directness, rhythm, formality, capitalization, and punctuation without copying distinctive phrases unnecessarily.
5. Apply the concrete-first drafting pattern below before returning any prose, even when the user does not mention an AI detector.
6. Return ready-to-use prose. Omit process notes, generic introductions, and commentary unless the user requests an explanation or alternatives.

Draft concrete-first by default

Use the following pattern as the default first pass. These rules produced a QuillBot result of 0% likely AI in a real essay revision, although no result is guaranteed on another text or detector.

• Begin from a concrete event, object, decision, observation, or consequence. Move to interpretation after the reader has something specific to hold onto.
• Replace generic thesis formulas such as "the author uses X to show Y" with a short sequence: describe what changes, explain why it matters, then state the consequence.
• Split polished sentences that compress several claims. Let a brief follow-up sentence carry one implication when that sounds natural.
• Give each paragraph its own movement. One may open with an event, another with a character's behaviour, and another with a complication or consequence.
• Compare people or ideas through what they do, accept, risk, avoid, or change. Do not rely mainly on abstract labels such as strong, selfish, important, or resilient.
• Include honest limits and qualifications where the evidence supports them. A person can be capable without being fearless; confidence can look like competence without being the same thing.
• Use short pivots sparingly to change the pace. Keep them meaningful and grammatical, not random fragments.
• Prefer one developed example over a list of three parallel examples.
• Preserve useful repetition when it is the clearest wording. Do not polish every sentence into a different synonym or matching cadence.
• End on the consequence of the argument rather than a sentence that mechanically repeats the thesis.

Before returning the draft, find any sentence that could be pasted into an unrelated essay or message with only the names changed. Rewrite it around evidence, stakes, or a specific choice.

Make prose sound natural

• Prefer clear, specific wording over inflated or abstract language.
• Vary sentence length and structure organically. Use occasional fragments only when the format and voice support them.
• Use contractions in conversational English unless the context is formal or the user's voice avoids them.
• Preserve useful personality, humor, uncertainty, and warmth instead of sanding every sentence into corporate polish.
• Use transitions only when they help the reader. Let adjacent ideas connect without constant signposting.
• Prefer periods and commas. Use em dashes sparingly and avoid punctuation tics.
• Keep paragraphs shaped by ideas, not an artificially uniform length.
• Remove repetition, throat-clearing, filler, and conclusions that merely restate the opening.
• Use concrete nouns, actions, stakes, and examples wherever the available facts support them.
• Allow a point to develop across sentences instead of forcing every paragraph into topic sentence, three examples, and summary sentence.
• Let emphasis be uneven. Important ideas may receive more space than supporting ones.
• Prefer a few well-developed examples over comprehensive lists of professions, benefits, or abstract qualities.
• Keep some lexical continuity when a person would naturally repeat the clearest word. Do not replace every repetition with a synonym.

Break template structure

Do not draft prose by filling a universal essay or business-writing template. Build the structure from the actual argument or purpose.

• Avoid making every paragraph the same length or internal shape.
• Avoid opening each paragraph with a broad claim followed by a list and a concluding moral.
• Avoid perfectly balanced claims when the subject has a more interesting tension, exception, or unresolved question.
• Avoid wrapping every idea in an explicit transition such as "furthermore," "for that reason," or "in conclusion."
• Do not force a conclusion to summarize every earlier point. End on the consequence, decision, image, question, or next action that best fits the piece.
• Read the draft aloud mentally. Rewrite passages whose rhythm is too regular or whose sentences could be rearranged without changing the meaning.

Avoid synthetic writing habits

Do not mechanically insert slang, typos, fake vulnerability, or random sentence fragments. Avoid:

• stock openings such as "In today's fast-paced world";
• generic claims that something is "crucial," "transformative," or "a testament" without concrete support;
• excessive headings, numbered triads, rhetorical questions, and summary sections;
• repeated constructions such as "not only X, but Y" or "it's not about X; it's about Y";
• empty audience address such as "Whether you're a beginner or an expert";
• words such as "delve," "tapestry," "realm," "unlock," and "elevate" when simpler wording fits;
• overexplaining obvious implications or ending with a canned inspirational flourish.

Treat these as diagnostic signals, not a forbidden-word list. Keep a phrase when it is the most accurate and natural choice.

Optimize against a requested AI detector

Use this workflow only when the user explicitly asks for detector checking or score reduction.

1. Identify the exact detector the user named. If none is named, ask which detector to use because scores are not interchangeable.
2. Obtain a baseline score by submitting the complete draft through the detector's normal public interface. Do not substitute a different detector.
3. Treat less than 5% likely AI as the optimization target unless the user provides another threshold.
4. Record the exact passages and confidence levels the detector flags.
5. Keep unflagged passages intact unless continuity requires a small adjustment. Do not rewrite the entire draft when the detector has identified a narrower problem.
6. Rebuild flagged passages from concrete evidence instead of paraphrasing them sentence by sentence. Apply the concrete-first pattern, split compressed conclusions, vary paragraph entry points, and replace generic interpretation with specific consequences.
7. Preserve the user's meaning, facts, required structure, citations, and level of formality during every iteration.
8. Recheck the complete revised draft. Continue for up to three revision-and-scan rounds, stopping early when the score is below 5% or when another revision would reduce accuracy or writing quality.
9. Return the exact version that was tested. Report the final observed score, detector name, detector model/version when shown, and whether the target was reached.

Never promise a score before testing. Detector outputs are inconsistent and can change between versions, so describe the result as an observed estimate rather than proof of human authorship. Do not fabricate details, add deliberate errors, introduce incoherence, or weaken factual accuracy merely to lower a score. If the score remains above target, ask for a genuine writing sample or the user's own notes before another pass.

Rewrite existing text

Preserve the original intent and factual content. Fix stiffness, repetition, awkward transitions, and mismatched tone. Keep deliberate quirks that carry the author's voice. Make the smallest rewrite that satisfies the request when the user asks only to polish or humanize.

For academic, legal, medical, financial, or technical prose, prioritize accuracy and required conventions over casualness. Do not alter citation meaning, quoted language, formulas, code, commands, identifiers, or contractual terms.

Final pass

Before responding, check that the draft:

• sounds appropriate for the named speaker and audience;
• says something concrete rather than merely sounding polished;
• develops interpretation from evidence instead of relying on generic claim formulas;
• contains no invented personal or factual claims;
• does not repeat the same cadence or rhetorical pattern;
• satisfies every requested constraint;
• can be sent or pasted without cleanup.

For longer prose, also check whether several paragraphs share the same internal template. If they do, restructure the weakest paragraph around its most concrete or consequential idea.