The bet that looked smarter
The pitch was simple. A language model can write a passable article in a minute, so a business could publish ten times as much for a fraction of the cost. For SEO, that sounded like more pages, more keywords and more traffic. Plenty of companies tried it, from big publishers to small sites buying "hundreds of articles a month" packages.
I use AI tools every day, so this is not an anti-AI post. Automation is good at producing text. It is not good at producing something nobody else has, which is exactly what search engines and AI assistants now look for.
What happened when companies leaned on it
SEO and AI search · Dubai
Five public turning points
CNET corrects its AI articles
CNET issued corrections on 41 of the 77 stories it had published using an AI tool, after an internal review found errors. Its owner told staff it was pausing AI-generated content across its sites.
Sports Illustrated and the fake authors
Futurism reported product reviews on Sports Illustrated under author profiles that did not appear to be real people, with headshots listed for sale on an AI headshot site. The publisher blamed a third-party supplier, removed the content and ended the partnership.
Google names scaled content abuse
Google added a spam policy for pages generated at scale to manipulate rankings, "no matter how it's created". Mass-produced, unoriginal content became an explicit policy problem.
Klarna hires humans again
Klarna's CEO told Bloomberg the company was recruiting human customer service staff so that "there will always be a human if you want", after saying its AI-first approach had led to lower quality.
Semrush publishes its own numbers
Semrush, which sells an AI article generator, published a study of 42,000 ranking blog pages and described a human-led process for its own blog.
Klarna is customer service, not content, but the pattern is the same: automation cut costs, quality slipped, and a human had to stay in the loop.
The Semrush example, stated precisely
Semrush is the example people quote most, so it is worth getting right. Its AI writing product, ContentShake AI, has not been shut down: the old address now redirects to Semrush's Content Toolkit, which still includes an AI Article Generator.
The interesting part is how Semrush runs its own blog. In its April 2026 study, Ana Camarena, Head of Organic Content Strategy at Semrush, explains that the team uses AI for ideation, outlining and drafting, with a human reviewing every step. Every article involves at least three people (a strategist, a writer and an editor), often up to five, and all of them are subject matter experts or work closely with them. On editing, she says: "That step is still fully human-led."
The study took the top 10 Google results for 20,000 keywords, kept 42,000 blog pages and ran them through an AI detector. Pages classified as human-written held position one 80% of the time, against 9% for purely AI-generated pages, which Semrush summarises as position one results being 8x more likely to be human-written. One caveat: the classification relies on an AI detector (GPTZero), and detectors are not perfect. Semrush itself frames the finding as "search rewards human originality", not that AI is inherently bad.
The real lesson from Semrush
A company that sells an AI writer still puts strategists, writers and editors in charge of its own content. The tool speeds the work up. People decide what is worth saying.
Why automated content loses
Automated content has one weakness: anyone can make the same thing. If a tool can produce your article from a keyword, it can produce your competitor's too. When everything looks the same, the only things left to compete on are price and volume, and a small business cannot win that race.
"The real thing is having a personality. You can automate things, but then you can be beaten down on price, and there will never be a price low enough for you to win." Lohith Lalesh
By personality I don't mean jokes or a quirky tone. I mean what only you can say: how you do the work, what you have seen go wrong, what your customers ask. That is not a commodity, so nobody can undercut it.
Side by side
Same topic, two approaches
Automated, interchangeable
- Written from what already ranks
- No named person accountable for it
- Says what fifty other pages say
- Competes on volume and price
- Nothing for an AI answer to cite that it could not get elsewhere
Original, with a person behind it
- Starts from your own experience and data
- A real author who stands behind every claim
- Adds a method, an example or a number nobody else has
- Competes on trust and specificity
- Gives AI answers a reason to cite you
What AI readiness in marketing actually means
"AI readiness" usually gets sold as buying the right tools. I think it is closer to the opposite. A business is ready for AI search when it has something worth citing and a person who can produce more of it. Tools then make that person faster.
SEO and AI search · Dubai
Where AI helps, and where a person has to own it
| Task | AI is useful for | A person has to own |
|---|---|---|
| Research | Summarising sources, finding angles | Deciding which sources are trustworthy |
| Outlining | Suggesting structure and missing questions | Choosing what the page is really for |
| Drafting | A first pass to edit | The experience, examples and opinions |
| Facts | Pointing to where to check | Checking against primary sources |
| Editing | Spotting typos and repetition | Judgement, tone and what to cut |
| Original data | Tidying and charting it | Collecting it honestly |
How this connects to GEO
GEO (generative engine optimisation) means being cited or recommended in AI answers. The basics of being findable are covered in my guide on how to show up in AI answers. The part that tools cannot do for you is giving those systems a reason to pick your page.
Google says this plainly. In its May 2025 guidance on AI search, it tells site owners to "focus on making unique, non-commodity content". AI Mode and AI Overviews break a question into many related searches and pull from pages that answer each part. A page that repeats the consensus adds nothing to that answer. A page with a real test, a local price range from your own quotes, or a clear expert opinion does. I go deeper on this in query fan-out and information gain.
Google's policy is not "AI is banned"; it judges purpose and value. The detail is in what Google's policies say about AI content.
What I would do as a small Dubai business
SEO and AI search · Dubai
A human-led content process you can run
Pick one real question
Take a question customers actually ask you on WhatsApp or the phone, not a keyword on its own.
Write down what only you know
Your process, the mistakes you see, your pricing approach, photos of real work, numbers from your own enquiries.
Use AI for the scaffolding
Outline, first draft, a check for gaps. Treat the output as a rough draft from a junior.
Put a name on it
A real author with a short bio. If nobody is willing to sign it, do not publish it.
Check every fact
Against official or primary sources, listed on the page.
Publish less, update more
One strong page beats ten thin ones. Revisit it when prices, rules or your experience change.
If you would rather hand this to someone, this is how I write pages for SEO clients in Dubai: AI to speed up research and drafts, and a person (me) responsible for every word and every fact.