Avoiding AI Slop: Keeping Outreach Quality High in the Age of Generated Text

Every inbox now carries the same tell: an opening line that compliments a "recent post," a paragraph of vague enthusiasm, a call-to-action that could belong to any product in any category. Buyers have been trained by volume. They've seen enough AI-assisted outreach to recognize the cadence before they finish the first sentence, and recognition kills the message before it's read. The tools that draft this copy aren't the problem. The problem is treating a draft as a finished send.

Why generated text gets caught

AI-written copy tends to fail in three predictable ways: it's generically flattering rather than specific, it mirrors the same sentence rhythm across every message, and it references a detail (a job title, a company milestone) without connecting that detail to a reason the recipient should care. None of these are hard to fix. They're hard to notice, because the sentence reads fine in isolation. It's only against a full inbox of similar messages that the pattern becomes obvious, and buyers see that full inbox, not the isolated sentence.

This is the core answer to the question every rep asks eventually: how do I make AI-written emails sound human? The honest answer is that the AI draft was never going to sound human on its own, the editing pass is where humanity actually gets added. Three checks do most of the work.

1. The editing pass

Treat the first draft as a skeleton, not a send-ready email. Read it out loud. If a sentence wouldn't survive being said across a table to the actual person you're emailing, cut it. Generated drafts often over-explain, restating the recipient's own job function back to them, or summarizing the product in a way no seller would say in person. Strip that out. What's left should sound like notes a colleague jotted down before a call, not marketing copy.

2. The specificity test

Generic personalization, "I saw you work in SaaS", reads as automation regardless of whether a human or a model wrote it. Real specificity ties a detail to a consequence: not that a company raised a funding round, but what that round likely means for their hiring plan or tooling budget in the next two quarters. If the specific detail could be swapped for any other company's detail without changing the sentence, it isn't specific, it's a template with a variable dropped in. This is where 360°-style prospect research earns its keep: tools like Humanlinker pull together public signals, role, company context, recent activity, into a single view specifically so a rep can find the detail that actually connects to a reason to reply, rather than the first fact a scraper returned.

3. Personality grounding

The same accurate information can land as sharp or as tone-deaf depending on how it's delivered. A buyer who wants a numbers-first business case reads a warm, story-driven email as padding. A relationship-oriented buyer reads a terse, data-heavy pitch as cold. This is the piece most AI drafting tools skip entirely, because it requires modeling how someone communicates, not just what they do. Humanlinker's approach, analyzing a prospect's communication style through the DISC framework and using that to shape tone and structure, is built around this exact gap. It doesn't replace the specificity from step two; it decides how that specific detail gets framed once you've found it. The output still needs a human pass, but it starts from a message shaped for the actual reader instead of a generic register.

Where the process actually breaks

Slop isn't usually a failure of the AI tool, it's a failure of process. Teams under quota pressure send the first draft because editing takes time they don't feel they have. But an unedited AI email is not faster outreach; it's outreach that gets deleted faster, which means the time was spent for nothing. The fix isn't slowing down every send, it's building the editing pass into the workflow the same way a subject-line check or a CRM update already is: a fixed, non-skippable step before anything goes out. Meeting-prep briefings, which Humanlinker and similar platforms generate ahead of calls, work on the same logic, they exist to make sure the human, not the model, walks in with the sharpest version of the context.

It's also worth being deliberate about where prospect data comes from, especially for teams selling into Europe. Enrichment tools differ in what they collect and how, and GDPR compliance depends on the specific data source and processing basis, this isn't a detail to guess at. Check a vendor's data-handling documentation rather than assuming all enrichment tools are interchangeable on this point.

None of this requires abandoning AI-assisted drafting, the category, spanning tools from broad enrichment platforms like Apollo.io or Lusha to writing-focused ones like Lavender and structured workflow tools like Clay and Cognism, exists because manual research and drafting at outbound scale isn't realistic for most teams. The output just needs a human editor who treats the draft as raw material, not a finished product.

FAQ

How do I make AI-written emails sound human? Run every draft through three checks before sending: an editing pass that removes anything you wouldn't say out loud, a specificity test that ties the personalized detail to a real consequence rather than a generic observation, and a tone match to how the recipient actually communicates. The draft is a starting point; the human pass is what makes it sound like one person wrote it to another.

Does using an AI drafting tool automatically make outreach worse? No, the tool isn't what buyers detect. What they detect is an unedited draft sent at volume. The same tool, used with a consistent editing and personalization step, produces a different result than a raw output pushed straight to send.