Let me start with something honest: the first time I used AI for content work, I hated what came out. It was technically correct, grammatically clean, and completely hollow. It sounded like a LinkedIn post written by a committee. Nothing about it felt like anything a real person would actually say.

So I did what any writer instinctively does when a tool doesn't work — I blamed the tool. Then I realised the problem wasn't the AI. It was how I was using it.

I was treating it like a ghostwriter. As if I could hand over my brief and get my article back. That's not what AI is good at. What AI is actually good at is the scaffolding — the structural, repetitive, analytical work that precedes real writing. The moment I understood that distinction, everything changed.

"I don't ask AI to write for me. I ask it to clear the path so I can write better."

The Real Division of Labour

There's a version of the AI conversation that goes: "AI will replace writers." Then there's the smarter version: "AI replaces the parts of writing that were never really writing to begin with."

Think about what SEO content work actually involves on a given day. Keyword research. Crawling competitor URLs. Pulling search intent from the top-10 results. Checking heading structures. Building semantic clusters. Writing meta descriptions for 30 pages. How much of that requires a human mind? Honestly — almost none of it. It requires pattern recognition, speed, and volume. That's AI territory.

The part that requires a human? The sentence that makes someone stop scrolling. The framing that makes a complicated idea suddenly feel simple. The instinct for what this particular reader, from this particular country, in this particular context, actually needs to hear. That's not in a training set. That's experience. That's taste.

AI handles
Keyword clustering & intent mapping
Semantic gap analysis vs. competitors
Content brief generation from SERPs
Meta description drafts for bulk pages
Heading structure suggestions
First-draft outlines from a title
Multilingual search volume queries
I handle
The angle — what makes this piece different
Tone calibration for audience & culture
Opening lines — the only thing that matters
All Arabic content, always
Cultural context and nuance
Editing for voice, rhythm, and feel
Final judgement call on every piece

My Actual Workflow, Step by Step

I'm going to be specific here, because vague "AI tips" are useless. Here's how a typical SEO content project moves through my process.

01
Research phase — AI as analyst
I start with the data. I use AI-assisted tools to pull keyword clusters, map search intent across the top 10 results, and identify semantic gaps — content angles competitors haven't covered. At this stage I'm not writing a word. I'm building a map.
AI-heavy
02
Brief generation — AI as structure
I prompt the AI to generate a content brief: proposed H2/H3 structure, key entities to include, word count range, related questions from PAA. Then I read it critically and rewrite half of it. The AI gives me a skeleton. I decide what the body looks like.
Collaborative
03
Opening — always mine
I never let AI write the first paragraph. This is non-negotiable. The opening is where voice is established. If I lose the reader here, the rest doesn't matter. I write the opening cold, before I've read back any AI output, so it comes from me and not from what I've just been primed to produce.
Human-only
04
Body draft — AI as first pass, me as editor
For body sections, I'll sometimes use AI for a first draft of a specific section — not the whole piece. Then I rewrite it, not edit it. There's a difference: editing means fixing what's there; rewriting means using it as a loose reference while writing my own version. The AI output is a prompt, not a product.
Collaborative
05
Multilingual work — almost entirely human
For French and Arabic content especially, I write from scratch. AI translation tools produce technically acceptable text but miss cultural register completely. Arabic content for a Saudi government audience requires a formality, a respect for rhetorical tradition, a precision of word choice that no current tool reliably delivers. This is where being trilingual is a competitive advantage that cannot be replicated.
Human-only
06
Final pass — reading it aloud
Every piece, without exception, I read aloud before publishing. This is the oldest editorial trick in the world and still the most effective. If my mouth trips over a sentence, the reader's brain will too. AI-generated text has a particular cadence — slightly too smooth, slightly too even — that becomes obvious the moment you hear it. Reading aloud catches everything.
Human-only

The Multilingual Question

I want to spend a moment on this because it's underexplored in most AI-and-SEO conversations, which tend to assume English as the default.

Working across French, English, and Arabic means working across three fundamentally different rhetorical traditions. Arabic, in particular, is a language where tone and register carry enormous cultural weight. An Arabic article written for a Rwandan government audience versus a Saudi audience versus a Moroccan SME audience requires three entirely different registers — even if the factual content is identical. No AI I have encountered reliably navigates this. The output tends to flatten toward a generic Modern Standard Arabic that serves no audience especially well.

My practical rule: AI can assist with English and French content at various stages of my workflow. For Arabic, I write from scratch, every time. The quality difference is immediately visible to any native reader.

What I've Learned About Staying Human

The writers I've seen lose their voice to AI are, almost without exception, the ones who started using it for the parts they actually enjoyed — the opening line, the unexpected metaphor, the section they were excited to write. They outsourced their best work and kept their most draining work for themselves.

The smarter move is the inverse: protect the parts that require taste, give away the parts that require time. AI is a time machine for the mechanical work. Use it that way and you arrive at the creative part faster, fresher, and with more of yourself left to bring to it.

I also think there's an underrated skill in prompting as thinking. Writing a good AI prompt is an act of editorial clarity. It forces you to articulate what you're actually trying to achieve — who the audience is, what the one argument is, why this piece needs to exist at all. If you can't write that prompt clearly, you haven't thought through the piece clearly either. In that sense, AI has made me a more rigorous thinker before I write a single word.

"The best SEO today lives at the intersection of human creativity and machine efficiency. Tipping too far either way costs you something real."

A Note on Authenticity and Rankings

The SEO dimension matters here too. Search engines are increasingly capable of detecting thin, templated, or undifferentiated content — not because they're looking for AI fingerprints per se, but because they're rewarding content that demonstrates genuine expertise, experience, and depth of perspective. These are things you earn by having done the work. They show up in how you cite evidence, how you acknowledge nuance, how you handle the edge cases a surface-level treatment would skip.

AI can help you produce more content, faster. It cannot help you produce content that ranks for the queries where real expertise is the differentiator. For those — and they tend to be the most valuable queries — the only path is the human one.

The writers who will do best in this environment are not the ones who refuse AI, nor the ones who fully surrender to it. They're the ones who know exactly where the line is — and who have enough sense of their own voice to know when they've crossed it.

Want to talk about this?
I'd love to hear how you're navigating AI in your own workflow — or help you build a content strategy that actually sounds like you.