The Anti-AI-Slop Playbook: How to Sound Human in a World Full of ChatGPT Emails

The Baseline Has Shifted

Five years ago, the challenge was writing an email that was clear, concise, and grammatically correct. In 2026, an LLM will produce that on demand in three seconds. The new challenge is writing an email that a recipient — who has already read ten AI-drafted messages that morning — recognizes as *human*.

This playbook is what practicing writers, sales leaders, and founders are doing to keep their email distinctive in a saturated environment. It is deliberately opinionated and specific, because generic writing advice is the exact thing that produces slop.

Rule 1: Start Somewhere Non-Obvious

The average AI-drafted email opens with a ceremonial line. The average high-response email opens with a specific detail the recipient can immediately place. Some variants:

• Reference a moment. "Your Q&A after the keynote — the point about pricing being a positioning tool, not a revenue lever — I have been thinking about it for three days."
• Ask a specific question. "Did you land on Postgres or DuckDB for the OLAP layer?"
• Confess a small thing. "I have been meaning to write for six weeks. Kept getting to the second paragraph and losing the thread."

None of these can be produced by a prompt like "write a professional cold outreach." They require the writer to have engaged with a specific real thing.

Rule 2: One Sentence Under Five Words

Every email you send should contain at least one sentence with fewer than five words. This is a mechanical trick that breaks the metronomic rhythm of LLM prose.

Examples: "Here is the pitch." "That was my fault." "The rest is detail." "I disagree."

The short sentence acts as a texture change. It signals a human voice through the specific choice of *what to cut short*.

Rule 3: Use One Piece of Non-Standard Punctuation

Real human writing accumulates weird punctuation habits — a dash, a semicolon in an odd place, a colon-then-fragment, parentheses inside parentheses. AI writing is grammatically flawless in a way that reads as sterile.

Pick your habit: em dashes, ellipses, mid-sentence colons, whatever. Use it once per email. It marks the writing as yours.

Rule 4: Name Your Own Stance

LLMs are trained to be balanced and non-committal. Real high-signal writing takes positions. In every email of substance, take one clear stance — an opinion, a preference, a small disagreement, a bet.

• "I think the Q3 target is aggressive but achievable if we cut scope on the reporting rewrite."
• "I would not do the integration. The maintenance tail is too long for the deal size."
• "Your product wedge is the mobile UX. Everything else is a distraction."

Recipients remember stances. They forget summaries.

Rule 5: Cite a Number

Numbers are where AI writing fails most visibly. LLMs hallucinate figures or fall back on vague qualifiers. Human writers who care about credibility include a specific number in almost every claim.

• Not "we saw significant improvement" — "we cut p95 latency from 480ms to 190ms."
• Not "our team is growing quickly" — "we hit 34 engineers, up from 18 six months ago."
• Not "many customers requested this" — "17 customers requested this in Q2 alone."

A single specific number per email flags the writer as informed and the message as concrete.

Rule 6: Reveal One Small Failure

Nothing marks writing as human faster than admitting something that AI would never surface. A missed detail, a wrong assumption, a false start. This is not self-deprecation; it is calibrated honesty.

• "I tried the CSV export first, misread the spec, and ended up rebuilding it in JSON on Thursday."
• "The first draft of this pitch was 12 slides and terrible."
• "I was wrong about the pricing tier. The data comes in below."

The admission signals that a real person navigated the situation.

Rule 7: Kill One Filler Section

Look at your draft. Find the paragraph that could be deleted without losing meaning. Delete it. Do this every time.

Common candidates: the paragraph restating what the recipient already knows. The paragraph explaining why you are reaching out (usually obvious from context). The wrap-up paragraph summarizing what you just wrote.

Shorter is more human. AI writing tends to over-explain because the underlying model is rewarded for helpfulness. Human writing trusts the reader.

Rule 8: Use One Contraction Per 100 Words

Contractions are a strong human signal. Set a floor: at least one per 100 words, more in informal contexts. If you write "do not" three times in a paragraph, one of them should have been "don't."

Rule 9: Signoff Like a Person

Ceremonial signoffs are dead. The following are all better than "Kind regards":

• Your first name only.
• "Thanks,"
• "Cheers,"
• "More soon,"
• "—[first name]"

Save the formal signoffs for genuinely formal contexts (legal disclosures, external counsel, board communications). Everywhere else, be a person.

A Daily Routine for Keeping Your Voice

The playbook above is a checklist. The routine is what makes it habitual.

Morning: Read your own writing. Pick two emails you sent last week that got a strong response. Read them slowly. Notice what makes them yours — the sentence rhythms, the openers, the closers. This is your voice; you are re-loading it into memory.

Compose: Draft twice. For any email that matters, draft once (fast, unfiltered, possibly AI-assisted for the outline) and rewrite once (slower, applying the checklist above). The two-pass structure is where voice reasserts itself.

Post-send: Save the winners. When an email gets a great response, save it to a "hits" folder. Every quarter, review the folder. The patterns become obvious. Amplify them.

The Tool Layer

For most people, executing the checklist consistently on 40+ emails a day is unrealistic without help. This is where a voice-aware tool like Presend matters. It scores every draft against your historical writing and flags the ones drifting toward slop. When it flags a drift, it offers a "More Me" rewrite grounded in your own sent folder, not in a generic "make this sound human" prompt.

The tool does not replace the routine — it makes the routine cheap enough to actually run.

Anti-Anti-Slop: The Failure Modes

A few ways the anti-slop instinct itself goes wrong:

• Performative weirdness. Adding an em dash on purpose because "humans use dashes" is more slop, just from the other direction. Real voice is not a costume.
• Being interesting instead of being useful. The point of email is to communicate. If your recipient needs the ETA and you send them a 300-word letter with a well-placed contraction, you have lost the plot.
• Assuming voice is enough. Voice is table stakes. Substance still has to be there.

The right frame: voice is the delivery mechanism for real information. Both matter.

Bottom Line

The correction to the AI-slop era is not less AI. It is more attention to what makes writing feel human. Specific detail. A stance. A number. A short sentence. A contraction. A signoff that sounds like you.

None of these are hard. They just require the writer to still be present in the sending. Presence is the scarce resource. In an inbox where everyone is auto-drafting, the writer who is actually there wins the reply.

Add Presend to Chrome — Free →