Keeping Content Human in the Age of AI 

After ten years working in the content discipline, I’ve watched the field get mislabeled, relabeled, and transformed. Content marketing to content strategy to content design. Brand storytelling to SEO. Keyword-stuffed to data-driven.

But nothing has been quite as chaotic and consequential as the rise of consumer-facing AI. While I was first to the plate to give it a swing (with plenty of hits and misses), I’ve watched a growing fear take hold of content professionals in the industry: that we’re losing the human part of the process entirely. 

LLMs have changed how content gets made and who it gets made for. And while the industry once again debates what to call it (AEO, SEO, GEO), we’re forgetting what makes content successful in the first place. 

The main response has been to fill the web with AI-generated, copy-paste content. In January 2020, 97.84% of internet content was written by humans, with 2.16% written by AI. By April 2023, 5 months after the release of ChatGPT, the amount of AI-written content jumped to 25.51%. As of May 2025, that number is 51.72% (Graphite).

But even as bots become content consumers, humans are still the ultimate audience. Brand affinity and loyalty are built on relationships with real people. And you can’t automate a relationship.

---

The problem isn’t that AI exists. It’s that so many are treating it as a replacement for our audience and our teams, rather than as a tool to help them.

---

We now need to design content for both audiences: structured enough for algorithms, substantial enough for people. 

What we learned from reviewing hundreds of SERP and LLM results

E-E-A-T gave us a strong foundation. Expertise, authority, and trust still matter.

But after reviewing hundreds of search results, AI overviews, and LLM outputs, it’s clear that content requirements are evolving. AI surfaces content differently than traditional search, and designing for both audiences requires an additional layer of consideration.

AI algorithms want unbiased clarity 

It’s our instinct as marketers to put our brands front, center, and differentiated every chance we get. But when a consumer is searching for a generic question, the answer is no longer the time or place to plug your brand (unless, of course, it’s a branded query). 

So, how do we design content for humans and algorithms?

The E-E-A-T framework is a great starting point, but it has some gaps. It doesn’t account for the unbiased clarity that both LLMs and real audiences reward, and it says nothing about whether the content you’re creating is actually relevant to a human need. 

So when creating content for clients, we layer in two more dimensions: Clarity and Relevance, shifting the framework to C-R-E-A-T-E.

C- Clarity

Is the content instantly understandable to both humans and machines? 

LLMs surface content that answers questions directly, without brand noise or filler. Humans want the same thing. They’re fatigued by content that buries the answer under brand messaging and CTAs. 

The same directness that earns an AI citation is what keeps a human reader from bouncing.

Clarity looks like: 

• Plain language and a strong information hierarchy with clear signals in headlines about what the content covers
• Structured content (headings, bullets, schema) that helps engines parse and readers skim
• Direct question-answer pairings, summaries, and definitions that make concepts easy to understand and information easy to extract

R – Relevance

Does the content align with the audience’s needs and true search intent?

Relevance is the filter both audiences apply first. Humans ask “is this for me?” LLMs ask “does this answer the query?” An article that uses the right keyword but doesn’t deliver what a reader expects around it fails both checks before it gets a chance to prove itself.

The first question in content planning shouldn’t be “what do we want to say?” It should be “what does our audience actually need to know?”

Relevance looks like: 

• Topics are mapped to specific intents (informational, commercial, or transactional) so the content matches what the reader is actually trying to do
• Format matches how the audience wants to consume in that channel
• The language and framing mirror how the audience thinks and talks about the topic

E- Expertise

Does this demonstrate depth of knowledge, not just surface familiarity?

We’re trying to break free of copy-paste advice. Both audiences want topics fully explored, not summarized. Humans weigh pros, cons, and edge cases when making decisions. For LLMs, that same depth produces a more useful answer for the person asking.

Expertise looks like: 

• Answers that go beyond the basics, anticipating and addressing follow-up questions and objections
• Comfort with nuance: tradeoffs, edge cases, and “it depends” situations explained with enough depth to be genuinely useful
• Original frameworks, positions you can defend, and answers to questions most content overlooks 

A – Authoritative

Does this source get cited, linked, or mentioned by others organically?

Humans trust what other trusted sources have already vetted. LLMs operationalize the same logic: backlinks are citations, third-party mentions are endorsements. Both are looking for validation from outside the source itself.

Authority looks like: 

• Other reputable sites link to your content as a reference
• The brand or author is cited, quoted, or mentioned in industry publications or in audience communities 
• The site has established topical depth with a lot of credible content around a specific subject area

T- Trustworthy

Can a human and an LLM safely rely on this content?

Trust makes you the default source of answers. Humans and LLMs look for signals that your content is well-sourced, up to date, and transparent when deciding whether it’s reliable. 

In sensitive or high-stakes categories especially, visible human oversight and documented accuracy set you apart from content that gets deprioritized.

Trustworthiness looks like: 

• Transparency about methodology, limitations, and assumptions
• Up to date, with a clear sense of when it was last validated or revised
• Privacy-safe, non-exploitative, and aligned with brand + audience values 
• Honest sourcing, citations, and links to original data — no hallucinated references, no overconfident claims

E – Experience

Is this content grounded in real-world experience?

Expertise tells people what you know. Experience shows how you know it.

Sharing real-life scenarios—what you’ve tried, what failed, what you learned—separates content worth reading from content that adds to the noise. For humans, that specificity builds credibility. For LLMs, it’s what makes content distinctive in the retrieval pool with details that can’t be synthesized from other sources. 

Experience looks like: 

• Case studies, experiments, and “what we tried” narratives that only someone who did the work could write
• First-person accounts with named outcomes or observed details that only exist if you were there
• Honest accounts of tradeoffs, failures, and edge cases

Why generic AI-generated content won’t work

AI can generate thousands of words in seconds. But remember what we’re designing for: content structured enough for algorithms, substantial enough for people. On both counts, most machine-generated content falls short.

AI aims for the average (and often falls below average) 

LLMs generate text one word at a time, predicting what’s most likely to come next based on patterns across billions of web pages. The output reflects the most statistically probable version of whatever you asked for with familiar phrasing, well-circulated angles, and advice that could live on any other website in your category. 

But the prediction model also results in odd sentence fragments, data hallucinations, and logical leaps galore. We see those issues get published without a single glance, from blogs to potato chip bags to leading publications. 

AI fails the C-R-E-A-T-E framework

When you review AI’s first pass at writing for a topic, you’ll clearly see where it falls short of C-R-E-A-T-E standards.

Clarity suffers from overstuffed answers padded with filler rather than precise, straightforward copy. And you won’t find any lived experience stories to back up your advice (unless hallucinated, of course). The output is devoid of case studies, “what we tried”, and expert-level details that only someone who lives the subject could write. 

AI will flatten your brand voice

Prompt engineers will tell you they can shape an LLM to match your brand. And technically, they’re right; with exhaustive documentation, detailed style guides, and rigorous examples, you can get AI close. 

“Close” being the operative word.

Without that infrastructure, AI defaults to one of two modes: heavy marketing speak or corporate-safe, personality-free language. And even with the infrastructure in place, the output still needs a human editor who knows your brand well enough to catch what the prompt missed.

AI may erode brand trust

It’s easy for a writer to pinpoint AI writing, and for the general audience, it’s not as hard as one would think. Everyone is catching on.

One Bynder study asked 2,000 consumers to correctly identify AI-generated copy. 50% could easily spot it, and 52% said they’d feel less engaged reading content they suspected was machine-written. As AI tools become more accessible and more people use them daily, detection will only get easier.

Audiences will become exhausted by rinse-and-repeat “AI slop” and keep scrolling when they see you show up in their feed. 

AI content poses serious risks in high-stakes categories

In complex, regulated spaces like healthcare, finance, and legal, trust is paramount.

Machine-generated content often introduces the risk of subtle inaccuracies that slip past a quick skim, out-of-date guidance that appears current, and misinterpretations of research or regulations. For audiences, that’s a big credibility problem. For companies, it’s a legal problem.