"How far can you go if you leave SEO to AI?"
I tested this on six sites over a period of 180 days. Here is the record of outsourcing SEO entirely to Claude Code.
In short: 3 sites succeeded, and 3 sites were devastated.
※ The PDF distribution campaign has ended, but for those who reposted by 3/8, I will still provide it. Please contact me again via DM.🙇♂️
An example of a successful site. Growing steadily from zero over 6 months without a single dip.
An example of a failed site. After peaking in the 3rd month, it plummeted due to an update.
Below is the main report.
What did I delegate to Claude Code?
Across all 6 sites, I entrusted the following tasks to the AI:
Keyword Research & Clustering Bulk extraction of hundreds to 2,000 keywords in the target genre, clustered by search intent. It categorized them into "comparison/consideration," "know-how," and "case studies," and even generated a prioritized posting schedule. A process that used to take 2–3 weeks was completed in half a day.
Topic Cluster Design Designing the structure of pillar pages and satellite articles and automatically generating an internal link matrix. The AI outputted the mapping of "which article should link to which," with a consultant performing the final adjustments.
Article Outlining & Drafting Analyzing the structure of top-ranking competitor articles to automatically generate headings, subheadings, and essential elements. The AI also created drafts, and in a workflow where humans added content and edited, the production time per article was reduced by an average of 60–70%.
Automatic Generation of Structured Data Bulk generation of JSON-LD schemas like LocalBusiness, Product, SoftwareApplication, FAQ, and Person depending on the site type. Work that would take 30 minutes per page manually was compressed into a few minutes.
Bulk Revision of Titles & Meta Descriptions Analyzing the title and search intent of all pages to generate revision plans for maximizing CTR. The AI applied principles such as "including specific numbers," "showing expertise," and "indicating freshness with the year" to each page.
Competitor Analysis Investigating the structure, conversion paths, and structured data implementation of 10–20 competitor sites in the same genre to create a checklist of "elements competitors have implemented that my site lacks." A process that usually takes 2 days was shortened to 3 hours.
In other words, AI handled almost all stages of SEO: "research," "design," "drafting," and "technical implementation." Humans were responsible for final decisions and adding first-hand information.
The 3 Successful Sites
Site A: 0 → 300/day, approx. 100 monthly CVs
Site A: The curve changed around December. Bounce Rate 47%, average session duration over 2 minutes.
Starting from zero, it reached 300 visits per day after 6 months. Monthly conversions reached about 100. Zero advertising costs. A hybrid model of SNS and search, where the update acted as a tailwind.
Site B: 0 → 300/day, approx. 300 monthly CVs
Site B: Bounce Rate 33%, average session duration over 4 minutes. Visitors view an average of 4.45 pages.
A B2B site. Starting from zero, it reached 300 visits per day and about 300 monthly conversions in 6 months. No SNS management, purely SEO-driven. Evaluation rose consecutively across two updates.
Site C: 0 → 100/day, approx. 60 monthly CVs
Site C: Not flashy, but hasn't dropped once in 6 months. This is the most reproducible pattern.
Starting from zero, it reached 100 visits per day and about 60 monthly conversions in 6 months. While not flashy, it grew steadily through consistent content accumulation—the most reproducible pattern.
The 3 Failed Sites
Site D: 0 → 200 → 10/day, approx. 10 monthly CVs
Site D: Fell off a cliff from its peak. Bounce Rate 70%, session duration 7 seconds. Readers leave immediately.
Also started from zero. 100 articles were generated in bulk by AI and posted with almost no editing. It grew to 200/day in the 3rd month but dropped 92.5% after an update. In the 6th month, it has 10 visits per day.
Site E: 0 → 400 → 50/day, approx. 10 monthly CVs
Site E: Bounce Rate 64%, session duration 29 seconds. Cumulative 16.9k unique visitors. The 2nd highest among the 6 sites. Yet, it was devastated.
Also started from zero. At its peak, it grew to 400 visits per day. However, it dropped 87.5% after an update. The quality of the articles themselves wasn't bad. The problem lay elsewhere.
Site F: 0 → 80 → 3/day, approx. 1 monthly CV
Site F: Devastated in stages by two updates. Session duration 9 seconds. Barely being read.
Also started from zero. The most catastrophic result. Hit by two updates, it dropped 96%. In the 6th month, it has 3 visits per day. It wasn't about how the AI was used, but a more fundamental problem.
Comparing the 6 Sites
Comparison table of the 6 sites.
Using the same AI tools, delegating the same SEO processes to the AI, and spending the same 180 days resulted in this difference.
As written above, AI handled keyword research, topic cluster design, article drafting, and structured data for all 6 sites. It wasn't a difference in tools.
The turning point between successful and failed sites came down to three factors. I can share one here:
Whether or not human first-hand information was added to the AI output.
The 3 successful sites treated the AI drafts as "raw material" and added field experience, real-life trials, and specialized knowledge. The 3 failed sites either published the AI output as-is or processed it in the wrong direction.
The remaining two turning points, along with specific actions taken and rejected for each site, have been published on Note.
https://note.com/seotigerjp/n/n952f1f727dc8
※ The PDF distribution campaign has ended, but for those who reposted by 3/8, I will still provide it. Please contact me again via DM.🙇♂️
