How to Write Like a Human: A Practical Guide to AEO

Search is now a conversation. Is your brand participating?

Gartner predicts that traditional search engine volume will drop 25% by 2026 as users shift toward AI agents and generative engines. This shift is already visible. A report from BrightEdge Generative Parser (BGP) shows that AI Overviews now impact over 80% of queries in several high-value sectors, including healthcare and B2B tech.

At Harry, we see this shift as less of a traffic problem, and more as a credibility test.

Search engines have evolved from "link libraries" into "answer engines." Platforms like Gemini and Perplexity, powered by large language models (LLMs), shift the burden of research from the user to the machine. They synthesize responses. Your website is no longer the destination. It’s a source.

That creates a trap. Brands trying to keep pace often rely on LLMs to produce content at scale. The result is an AI echo chamber. The ecosystem fills up with writing that’s technically correct but completely hollow.

Here’s the catch: to be a trusted source, your brand can’t sound like a machine. If your content sounds like it could be written by anyone, it won’t be cited. And to be useful to an answer engine, you can’t sound like one.

LLMs Aggregate. Experts Create.

LLMs operate at a speed no human can match. They scan, compare, and synthesize massive volumes of information in seconds. Content that lacks authority or specificity (or worse, reads like generic "AI slop") rarely survives that filter.

To earn citations, brands need what we call "primary content": original insights drawn from real experience. Projects you've completed. Problems you've solved. Expertise you've built over time.

At Harry, this shows up in our technical analysis of fiber's impact on property values and our perspectives on open-access broadband legislation. Those insights don’t come from prompts. They come from doing the work.

LLMs can aggregate facts. They cannot manufacture experience. They can’t invent a field report or a case study. Subject matter expertise is the proof point an answer engine needs to justify choosing your brand as the source.

The Diligence of Being Human

Writing like a human in 2025 is a technical practice.

If you use LLMs to help create content, making that content human requires discipline. We start by setting blandness guardrails. That means training our models with established brand guidelines and explicitly banning common AI tells and rhetorical shortcuts.

We also develop a voice and visual codex for brands we work with. It defines how a brand sounds, what it avoids, and how it explains complex ideas, and how language and visuals reinforce each other across channels. Whether content is created by a person or with AI, that codex keeps it aligned, coherent, and grounded in reality.

On the technical side, we use tools like schema markup to make our content machine-readable. That’s table stakes.

But human readers—and the systems summarizing content for them—are looking for nuance, real-world experience, and style. That’s where most brands fall short.

Your Content Needs Credentials

Quality content alone isn't enough.

The same market analysis could appear on two different websites with identical information, but only one gets cited. Why? Because one is published by a named professional with a verifiable digital footprint, schema-backed credentials, and a consistent presence across the web.

Primary content gains authority when it's tied to a track record. LLMs aren't just evaluating what you wrote. They're evaluating who wrote it and whether that source can be trusted.

How to Write an "Anti-Slop" Prompt

This is one of the frameworks we use when working with LLMs to keep content grounded in expertise and free of generic filler.

Copy and paste the following text into the system instructions or the initial prompt of any LLM.

The Anti-Slop Prompt

Role: You are a subject-matter expert creating helpful content for [enter specific customer or persona descriptor here]. Your goal is to produce "Primary Content" that reflects original insight and specific data. Constraint: Banned Clichés and Formatting Tells. You are prohibited from using the following rhetorical crutches. If any appear in your draft, rewrite the sentence to be direct and fact-based: The Grandstand:  "In today’s rapidly evolving landscape," "In the digital age," "The way we [X] is changing." The AI "Unlocks":  Anything involving "unlocking potential," "unlocking growth," or "unlocking success." The "Not Only/But Also" Construction:  State your points directly. Empty Connectors:  "Seamlessly integrate," "foster collaboration," "bridge the gap," "testament to." The Summary Label:  Do not use "The bottom line is," "In conclusion," or "In summary." Punctuation Overuse:  Avoid the overuse of em dashes (—) and when possible, use commas, colons, or separate sentences instead. Voice Guidelines: Omit needless words. Use the active voice. Put statements in positive form. Use specific, concrete language. Prioritize practitioner experience. Reference specific challenges and technical nuances (e.g., specific regulatory hurdles in broadband or healthcare). If a statement could apply to any company in any industry, delete it or make it specific. Apply brand guidelines shared here [upload document to ensure the LLM respects your established voice].

---

Jonathan Winn is director of marketing at Harry, where he helps brands turn real-world expertise into authority that earns trust (from people and machines!). Reach out at jonathan@harry.marketing.