In this article 7 sections
The current state of AI in SEO
After two years of using generative AI intensively for SEO, I have a pretty clear view: it’s an incredible tool that most people misuse.
The problem isn’t AI. It’s how we use it.
What works really well
1. Research and analysis
AI is brilliant for:
- Analyzing SERPs and extracting patterns
- Summarizing technical documentation
- Identifying entities and semantic relationships
- Generating keyword variations
A prompt I use all the time:
"Analyze the top 10 results for [keyword].
Identify: common structure, entities mentioned,
questions they answer, content gaps."2. Scaling technical content
For product pages, structured FAQs and data-driven content, AI saves hours. The key is having a solid template and clean input data.
3. Translation and localization
Better than Google Translate, especially if you give it industry context and the brand’s tone of voice.
What doesn’t work (or works badly)
Content built on real expertise
If your competitive advantage is deep knowledge and experience, AI can’t replicate it. It can sound convincing, but it lacks:
- Real anecdotes
- Opinions shaped by experience
- Nuances that only come from years of work
- The “instinct” for knowing what matters and what doesn’t
YMYL content
In Your Money or Your Life topics (health, finance, legal), AI is dangerous. Hallucinations are unacceptable when there are real consequences.
My current workflow
| Phase | Who does what | Approximate time |
|---|---|---|
| 1. Research | AI analyzes SERPs, clusters, gaps | 15 min |
| 2. Outline | A human defines the structure and key points | 20 min |
| 3. Draft | AI + human: technical parts vs. opinions | 1-2h |
| 4. Editing | A human reviews everything and adds experience | 30-45 min |
| 5. Optimization | AI generates title/meta variations | 10 min |
The result: content that sounds human because it is human where it matters (judgment, experience, opinion) but accelerated in the mechanical parts (research, variations, formatting).
The million-dollar question: does Google detect AI content?
Honestly, I think it’s the wrong question.
Google doesn’t penalize AI content. It penalizes bad content. And it turns out a lot of AI content is bad.
| Signs of bad AI content | Signs of AI used well |
|---|---|
| Generic, sounds like a template | Specific, with concrete data |
| No unique perspective | Opinion shaped by experience |
| Repeats what already exists | Brings a new angle |
| No E-E-A-T at all | Verifiable author with expertise |
| Made-up dates and data | Cited, verifiable sources |
| Uniform robotic tone | Recognizable personal voice |
If you use AI to create differentiated content with real expertise, there’s no problem. If you use it to churn out 1,000 filler articles, sooner or later you’ll pay for it.
Tools I use
| Tool | Best for | Weakness |
|---|---|---|
| Claude | Long analysis, reasoning, code | Slower on long answers |
| ChatGPT | Brainstorming, variations, creativity | Frequent hallucinations |
| Perplexity | Research with cited sources | Less depth in analysis |
| Gemini | Data analysis, Google integration | Inconsistent quality |
The choice depends on the task. For SERP analysis I use Perplexity (it cites sources). For writing briefs I use Claude (better structure). For title variations I use ChatGPT (more creative).
The real limit
AI is an amplifier. It amplifies good SEOs and exposes bad ones.
Google has made this clear in its guidance on AI-generated content: it doesn’t penalize AI, it penalizes low-quality content. The problem is that a lot of AI-generated content is generic, offers no unique perspective and lacks real E-E-A-T.
If your strategy is built on real expertise and differentiation, AI makes you more productive without sacrificing quality. If it’s built on volume without value, it just speeds up the road to irrelevance.
The limit is where your real knowledge ends.