Human or machine: Who are you writing for?

Somewhere in the last few years, writing for humans, not algorithms turned from good advice into something that sounds almost naive.

In 2025, I was making a case against FAQs – highlighting their impact on the user experience. But it felt different this time, like I was talking into a void. The prevailing AI rhetoric had already won: pad the word count, stuff in a few more variations of that keyword, make it search-friendly. Feed the crawler.

Being invisible seemed like the worst outcome. 

Fear of being invisible

First it was search crawlers and keyword indexers, now it's generative AI and language models that are the gatekeepers of digital content. The search engine decides whether your article gets seen, the parser decides whether your resume gets read, the summarizer decides what gets extracted and what gets ignored.

A brilliant article buried on page 10 of Google, or a resume auto-rejected in under a second, has an audience of zero – failing at the foundational layer of the Content Hierarchy of Needs.

Optimizing for AI no longer feels optional but a prerequisite for reaching an audience at all. But the real danger starts when optimization becomes the entire goal and we forget who we’re actually trying to reach.

When AI optimization becomes the goal

The failure isn't optimizing for AI, it's mistaking AI for the audience.

Generative search and AI overviews

For years, SEO writing meant padding articles with redundant subheadings and forced keyword phrases, hoping to get cited in a snippet or an AI overview. But when a human clicks through and lands on 3,000 words of filler, all you'll get is a bounce and broken trust. AI overviews are doing them a favor by extracting any meaning from content that's designed to waste the reader's time – or not designed for the reader at all. The fix is to write something genuinely useful instead of something hollow that has nothing worth extracting.

'SEO-optimized' recipe blogs

You just want to know how long to bake a potato, but instead, you're forced to rage-scroll through 1,500 words of a life story, the history of tubers and keyword-stuffed wellness tips. Why? Because engagement metrics reward time on page and scroll depth. The content was engineered to trap a crawler and make you scroll past as many ads as possible, rather than help a human with messy hands in a kitchen.

AI-assisted corporate communications

When you prompt an AI tool to sound professional you get a bunch of LinkedIn articles full of circling back, synergistic deliverables and leveraged bandwidth. That's not AI failing to reach a human, it's a writer failing to set parameters and user needs. The tool blindly fills the gap with the blandest average of corporate English. That doesn't mean you can't use AI to draft, but you still need to know who you're writing for or you'll end up writing for no one.

Reframing a false dichotomy

Does writing for a machine come at the cost of the user experience?

An algorithm uses headings to parse a document's hierarchy. A human does something surprisingly similar: they land on a page and scan the headings to work out whether it's worth their time. 

An LLM or search engine looks for semantic keywords to understand what a page is about. A human typing a search query is looking for the same signal – the fastest possible confirmation that they're in the right place, that this page speaks their language.

Structured data helps a machine index content cleanly. For a human (especially one on a phone, tired, trying to get through a task) the exact same structure is what keeps the cognitive load manageable.

Clear headings, relevant terminology, clean structure – the machine and the human turn out to want the same thing. This is also the argument of my piece on why content and code aren't all that different. The machine-readable layer of good content was never a separate skill bolted onto writing. It's the same content design discipline that's always existed.

Farewell to the FAQ playbook

For a prime example of how algorithmic incentives warped our pages, look no further than FAQs. FAQ sections have existed mainly for search, not to help users.

Google used to feature expandable FAQ dropdowns directly in search results – valuable real estate that let one page dominate a listing and boost click-through rates. FAQs were also a low-friction way to smuggle a dozen long-tail keyword variations onto a page without disrupting the main content.

But none of that has much to do with whether a human actually wants a wall of collapsed questions between them and the answer they came for.

Now the incentive is gone. Google phased out FAQ rich results and modern AI search doesn't need an isolated block labeled 'Frequently Asked Questions' to find an answer. AI overviews pull answers from anywhere on a well-structured page, whether that's a product description, a guide, or a single clear paragraph.

The pattern that was always better for users – answering a question in context, instead of dumping it into a separate FAQ silo – is now the pattern that performs best for search. In fact, Google wants you to focus on people-first content. Design it well for the human, and the extraction takes care of itself. You no longer have to choose.

Getting back to basics

AI advances haven't invented a new discipline that content designers have to learn. Ironically, they're dragging the industry back to fundamentals – clarity, structure and addressing real user needs – that were being abandoned in favor of tricks and hacks. When you design for the human first, you automatically give the machine what it needs.