AI SEO content writing is everywhere, but most of it ranks nowhere. Here's what actually works, and where Sage Writers draws the line.
Since the start of 2024, thousands of sites have been pumping out 20+ AI-generated posts a week. Google noticed. The March 2024 core and spam updates — and every follow-up since — have been quietly dismantling sites that treated SEO as a copy-paste prompt. Here’s a teardown of where AI SEO content writing actually earns its place in a workflow, where it crashes the site, and how to tell the difference before Googlebot does.
The AI content gold rush (and why Google noticed)
The pattern is easy to spot. A SaaS company publishes three blog posts a day for six months. Traffic climbs steadily. Then — one Tuesday morning — it drops 80% overnight. Nobody at the company changed anything. Google changed its mind.
The March 2024 spam policy update was explicit about “scaled content abuse”. Follow-up updates through 2024 and 2025 kept pulling the same lever. The real problem isn’t AI. It’s AI used without a human brain in the loop — no editorial judgement, no first-hand knowledge, no point of view. Just a prompt and a publish button.
So what does AI SEO content writing look like when it genuinely works? Not as a replacement for a writer. As a research assistant sitting next to one.
What AI gets right in SEO workflows
We use AI tools daily at Sage Writers. We’d be lying if we said otherwise. Where they genuinely save hours:
- SERP research. Pulling the top-10 outlines for a target keyword in 30 seconds instead of 30 minutes of manual tabs.
- Keyword clustering. Grouping 400 related phrases into a dozen topical buckets without a spreadsheet nightmare.
- PAA mining. Surfacing the “People Also Ask” questions worth answering in a sub-section.
- Repetitive formats. Draft meta descriptions at scale, product spec bullets, FAQ scaffolds.
- Schema drafting. A JSON-LD skeleton you can hand-correct instead of hand-build.
Here’s the caveat though. A “usable skeleton” is not a “publishable article”. The output of any of the above is a starting point — not a finished piece. Treat it as the first 10% of the work, not the last 90%.
The skeleton test
Read the AI draft aloud. If you can’t picture a specific person it’s written for, or a specific experience it’s drawing on, it’s a skeleton. Not an article.
Where AI SEO content writing falls apart
The failure modes are consistent enough that we can list them:
- Fabricated stats. The model confidently cites “a 2023 HubSpot study” that doesn’t exist. One of these in a published piece is enough to tank your E-E-A-T signal with reviewers.
- Hedge-everything prose. Every sentence qualifies itself. Nothing is claimed. Nothing is defended. It reads like every other site because, well, it was trained on every other site.
- No point of view. Google’s helpful content guidance has quietly been pushing opinion and lived experience up the stack since late 2023. An AI draft has neither.
- Zero customer knowledge. The model doesn’t know your pricing, your proof, your three best case studies, or the specific objection your sales team hears every week.
AI will happily write 2,000 words about your business without knowing a single true thing about it.
That’s the whole problem in one line.
The hybrid workflow that actually ranks
Here’s the four-step process we use on every piece — including this one.
- Brief. A human writes the brief. One target keyword, one target reader, one thing we want them to do after reading.
- AI research pass. The AI pulls SERP outlines, related terms, PAA questions, competitor coverage gaps. Output goes into a research doc, not the draft.
- Human draft. A senior writer writes the opening, the examples, the thesis and the voice. AI research sits beside it as reference, not template.
- Human edit for specifics. Every paragraph gets a specificity check — is there a named example, a number, a customer scenario? If not, rewrite.
The split matters. AI handles the parts of SEO that are genuinely mechanical. Humans handle the parts that make Google (and your reader) stay.
Quick tip — the specificity test. Every paragraph should contain at least one concrete noun that couldn’t appear on a competitor’s site unchanged. A brand name, a price, a client scenario, a specific number. If a paragraph passes the “could be anyone’s blog” test, it fails.
Can AI write a whole article on its own and rank?
Short-term, sometimes. Long-term, almost never. We’ve seen AI-only articles rank for low-competition long-tail queries for 60–90 days before quietly dropping. The pattern is consistent — the initial crawl indexes the content, then quality signals (bounce rate, time on page, lack of citations, lack of backlinks) pull the ranking down. If you’re building for the next quarter, hybrid wins. If you’re building for next week, you’re playing a losing game.
Three real examples from our desk
Three anonymised rewrites from the last few months. Industries changed, structure identical.
Example 1 — A B2B SaaS blog
An onboarding tool published 40 AI-drafted articles over six months. Combined traffic peaked at 1,200 clicks a month, then halved after the August 2024 update. We rewrote the top 10 by traffic as human-first pieces with real product screenshots and specific user quotes. Three months later: 2,800 clicks a month from those 10 pages alone. The other 30 we recommended de-indexing. They were pulling the domain down.
Example 2 — A local plumbing service page
The AI draft said “we provide reliable plumbing services across the Midlands”. Technically true. Said nothing. The human rewrite named three towns, listed the fixed call-out fee, and included one paragraph about a boiler job the owner had done the week before. Phone enquiries from the page doubled in a month. Same URL, same meta, different proof.
Example 3 — An ecommerce category page
The one case where AI genuinely pulled its weight. The client needed 400 product descriptions for a homeware range. We built a template, fed in the spec sheet per SKU, generated drafts, then had a human spot-check every tenth one for voice and accuracy. Shipped in a fortnight — would’ve been three months by hand. The category hero copy and the brand story? Still written by a human. The repetitive, structured, short-form product bullets? Perfect job for AI with a human editor.
How Sage Writers approaches AI in the brief
We’re a senior freelance SEO copywriter based in Birmingham — one writer, from brief to delivery. AI is a research assistant on our desk. It is never the author. Every article is read, rewritten and signed off by Bijal. The writer you brief is the writer who writes it. That’s non-negotiable.
Fixed price per project, first draft in five working days, two revision rounds included. No AI-mill outputs shipped as “ours”. If you want to understand how we think about search-led writing more broadly, our SEO content writing guide walks through the full approach, and our website content writing page covers how we handle longer-form site copy.
How much AI is too much AI?
Our rule — AI touches outlines, research and structured bulk content. AI does not touch the opening, the thesis, the examples, or the closing. Rough rule of thumb: if AI contributed more than 30% of the finished words you ship, you’re relying on it, not using it. The quality of the human 70% is what Google will remember.
What to ship this week
A 30-minute audit any founder can run today:
- Open your last 10 blog posts.
- For each one, ask — does any single paragraph name a specific customer, number, or scenario?
- Count how many posts pass. If it’s fewer than six out of ten, you have an AI-filler problem.
- Either delete the thin posts or rewrite them with real proof baked in.
That’s the lift. Not a full content overhaul. Just a clear-eyed look at what’s actually publishable and what’s placeholder.
If you’d like a second opinion on the last ten posts you shipped — or someone to take the next ten off your plate properly — get a free quote and we’ll have one back in your inbox within a working day.