How AI-Generated Content Performs: Experiment Results

Businesses have been using AI to automate different tasks, including content creation, for a while. But how effective is it exactly?

From an ethical standpoint, Google says that content should be high-quality and genuinely benefit users, no matter how it’s created. However, using AI to manipulate rankings goes against Google’s spam policies.

We conducted two experiments to see how AI-generated articles perform in different environments. First, we published six AI-assisted articles between June and September 2024 on the SE Ranking blog to see how they’d perform on our domain. We took it a step further by launching 20 brand-new websites and publishing 2,000 AI-generated articles across them in November 2024. Our goal was to discover if AI-written content could gain traction and bring in measurable traffic.

Let’s look at the results of our experiments!

Key results

SE Ranking blog results:

New domains results:

Disclaimer:

We conducted this experiment to get insights into how AI-generated content performs in real-world conditions. Our findings and observations are for informational and educational purposes only. We strongly discourage using the data from this experiment to manipulate search algorithms and rankings. 

Remember that while AI can help with generating content, we don’t recommend relying on it as the sole method for content creation. Always review, edit, and customize AI-generated content thoroughly and add human expertise. Always ensure that AI-created copies align with your brand voice, meet quality standards, and accurately serve your audience’s needs.

Testing AI-generated articles on the SE Ranking blog

We started our experiment by testing how AI-generated articles would perform on the SE Ranking blog. The goal was to evaluate whether AI-written materials could rank well on SERPs and attract traffic. We used the AI Writer to generate texts, published them on our blog, and tracked performance with Google Search Console, Keyword Rank Tracker, and AI Results Tracker. 

In the following sections, you’ll find more details about how we set up our experiment and the results we achieved on our blog.

SE Ranking blog experiment methodology

Selecting topics and generating content

We identified six topics that we believed would resonate well with our readers:

• Seed keywords
• Dashboards vs reports
• Product-led growth
• SEO benchmarking
• Taxonomy SEO
• Local business reviews

These target keywords served as the foundation for individual articles. We added them to SE Ranking’s Content Editor and used our AI Writer to create texts. This feature supports several generation formats:

• One-click article generation: This approach involves adding a target keyword and generating text immediately, without custom settings.
• Article generation via Wizard: Specify initial requirements for the AI Writer to follow when generating texts.

We chose the second option to cover the selected topics comprehensively and ensure articles are on par with our blog’s standards. Before generating articles, we examined each topic to establish a solid outline and specified the following requirements to the AI Writer:

• User intent 
• Target audience
• Content format
• Tone of voice
• Creativity level

We guided the AI step-by-step through the Wizard and provided the following context:

• Critical points to be covered
• Relevant keywords to incorporate
• Specific questions to be answered
• H1 heading
• Heading structure 

Based on this information, the AI Writer produced our article first drafts.

Editing AI-generated drafts

AI-generated drafts served as a starting point rather than as a final product. Our team revised each article for clarity and accuracy. Our SEO team also reviewed selected pieces for factual correctness.

We used our Content Editor’s recommendations to make small revisions to the text and improve the Content Score.

We didn’t proofread these articles. 

To maintain transparency, we included the following disclaimer in each published piece:

“This article was initially drafted using SE Ranking’s AI Writer. However, some parts of the text were written from scratch and the content has undergone thorough revisions, editing, and fact-checking by human editors and subject matter experts to ensure accuracy.”

Publishing polished articles on the SE Ranking blog

Once all adjustments were made, we published AI-assisted articles on our blog on the following dates:

SE Ranking blog experiment results

AI-generated articles got 555K impressions and 2,300+ clicks in a year

We tracked each article’s clicks and impressions in Google Search Console, starting from the publication date. Overall, AI-generated articles received 555K impressions and 2,300+ clicks between June 2024 and July 2025.

Below, you’ll find GSC data showing overall impressions and clicks:

For comparison: in the first 6 months after publication (June–December 2024), the articles received a total of 137,815 impressions and 866 clicks. The second half of the year saw a much more rapid growth in traffic and impressions. As we can see, impressions increased by as much as 4 times, and clicks by 2.5 times. This suggests that if an article is high-quality, it continues to deliver better results over time.

Below you’ll find GSC data showing the impressions and clicks, broken down into 30-day periods:

In the first six months, articles published earlier (e.g., Dashboards vs Reports and Seed Keywords) received more clicks and impressions because they had more time to rank and gain traction. For example, Seed Keywords started around 4,500 impressions in the first 30 days and nearly doubled that by the 6th period. The Dashboards vs Reports article showed even more dramatic growth, jumping from roughly 2,900 impressions in the first 30 days to over 11,000 impressions by the 4th interval and 14,000+ in the 5th and 6th. 

However, this advantage diminishes over time. While the Dashboards vs Reports article still leads in impressions, the second spot is now held by the SEO Benchmarking article, published in September 2024.

Clicks for each article fluctuated for various reasons. For example, Seed Keywords dropped from 28 clicks in the fourth interval to 20 in the fifth, then rebounded to 49 in the sixth before falling to 8 in July 2025. In contrast, the Dashboards vs Reports article showed a more consistently upward trend in clicks.

Interestingly, despite generating only 51K impressions, the SEO Taxonomy article has more clicks than the SEO Benchmarking article, which has nearly three times as many impressions.

3 of the generated articles rank in the organic top 10

We also tracked AI-generated articles and keywords with our Rank Tracker to see how positions changed over time. We chose July 31, 2025, as the end date for tracking positions for each article in our experiment (but we have been tracking positions throughout and continue to do so now).

As you can see from the table below, three articles are currently ranking in the top 10, one post is showing no performance, and two are in positions 19 and 25.

The screenshots from our Rank Tracker (shown below) present position dynamics for each article, beginning at the publication date and ending on July 31, 2025.

2 generated articles made it to the sources section in AI Overviews

We added our target keywords to SE Ranking’s AI Results Tracker to see if they trigger AI Overviews in search. We also checked if our generated articles were cited as sources in these AI responses.

Five out of six keywords triggered AI Overviews, and four articles were linked as sources in AIOs. Included were the Dashboards vs Reports, Taxonomy SEO articles, Seed keywords (since August 1, 2025), and SEO benchmarking. 

The local business reviews keyword triggered AI Overviews, but our related article didn’t appear in the source list. The product-led growth keyword currently doesn’t trigger AIOs.

By the way, if you want to go beyond experiments and start implementing AI in your day-to-day SEO work, check out our AI for SEO guide. It shows how to automate core website optimization tasks using AI, backed by real examples and action steps.

Testing AI-generated content on 20 brand-new websites

With this second iteration, we wanted to see how AI-generated content would perform when published on brand-new domains with no existing reputation. We also looked for any dependencies on the domain’s niche. Our goal was to see how AI content gets indexed, how well it can rank in search results, and whether it has traffic potential.

We accomplished this by testing 2,000 AI-generated articles across 20 newly registered domains. The following sections provide more details on our methodology and results.

New domains experiment methodology

Selecting niches and keywords

We chose 20 popular and broad-ranging niches to cover diverse interests and industries. These included:

• Arts & Entertainment
• Business & Services
• Community & Society
• Computers & Technology
• Ecommerce & Shopping
• Finance & Accounting
• Food & Drink
• Games & Accessories
• Health & Medicine
• Industry & Engineering
• Hobbies & Interests
• Home & Garden
• Jobs & Career
• Law & Government
• Lifestyle & Well-being
• Pets & Animals
• Science & Education
• Sports & Fitness
• Travel & Tourism
• Vehicles & Boats

For each niche, we selected 3–5 top-performing, authoritative websites that ranked for relevant how-to keywords. We used our Competitive Research Tool to extract informational how-to queries and applied filters to get a list of low-search-volume keywords. 

In cases where niches didn’t have enough low-search-volume terms, we supplemented the list with medium-search-volume queries. The monthly search volumes for the selected keywords ranged from 10 to 990, with an average of 106 across all niches.

We reviewed the keywords to match our criteria and cover different topics within each niche. This resulted in our keyword list of 100 how-to queries per niche and 2,000 keywords in total. Find the complete keyword list in this document.

Buying new domains and setting them up

To create a standardized environment, we purchased 20 brand-new domains—one for each niche. These domains had no prior search history, backlinks, or established reputations.

We made WordPress the content management system for each website. To maintain consistency across all sites, we applied the same free WordPress theme and activated only one additional plugin: Yoast SEO.

We didn’t make further customizations or enhancements. No internal linking was employed, and no images were added.

Generating content and publishing it on new websites

We generated content in one click using SE Ranking’s AI Writer. We created 2,000 articles in total and 100 per niche. We didn’t provide extra prompts, guidance, or edits. We also generated titles and meta descriptions for each content piece.

The generated content was used as-is. We generated articles in bulk with the help of our development team.

We then exported these AI-generated articles and uploaded them to the corresponding WordPress sites via SE Ranking’s WordPress API.

Each site received 100 articles during the same period on November 5-6, 2024. We added no further manual adjustments, internal linking, or images.

Tracking approach

After the content went live, we added each site to Google Search Console to monitor its indexing and performance metrics. We submitted sitemaps to encourage indexing without any other indexing or promotional tactics. We conducted no link-building campaigns, created no internal link structures, and promoted nothing on social media. These sites were left untouched because we wanted them to be evaluated based on merit alone.

We also added our keyword list to: 

All data was tracked in the US.

We gave the sites 36 days before checking the initial results, then rechecked them more than half a year later.

Results of the new domains experiment

70.95% of all generated pages were indexed in the first 36 days

Between November 6 and December 11, 2024, we monitored Google as it indexed the 2,000 AI-generated articles published across 20 new domains. The indexing rate reached 70.95% (1,419 out of 2,000 pages), meaning that almost two-thirds of all published pages were successfully included in Google’s search results within a little over one month. 

This is a high rate given that different estimates suggest that only about 15% to 20% of the entire web gets indexed. John Mueller from Google also noted that it’s normal for not all 100% of your pages to be indexed.

• 11 sites saw full indexation, with all 100 pages per site included in Google’s index.
• 3 sites achieved substantial, though not complete, indexation of 78–97 pages.
• 2 sites lagged behind, each having only 18 indexed pages.
• 4 niches struggled to get indexed by Google, each with fewer than 10 pages indexed.

8 websites started to rank for 1000+ keywords in 36 days

We also analyzed how many keywords each site began ranking for within the first month. Instead of the hand-picked “how-to” keywords we targeted during content creation, these keywords represent all terms for which GSC detected organic visibility. 

Our 36-day results show that:

• Eight sites had an impressive start, ranking for over 1,000 different keywords each. 
• Eight sites landed in the mid-range category, visible for between 100 and 1,000 keywords.
• Three sites had more modest visibility, ranking for between 30 and 100 keywords.
• One site barely gained traction, with only 2 detected keywords.

As for the positions these domains achieved across all keywords identified by GSC:

• Five sites had all their GSC keywords ranking between positions 29.7-49.3 on average. 
• Fourteen sites had all their GSC keywords ranking within positions 52.3-73 on average.
• One site had all its GSC keywords ranking in position 112.8 on average.

Generated content accumulated 240+ clicks and 122K impressions in the first month

We examined that early traffic and impression data and metrics across all domains are quite noticeable:

• 244 clicks in total
• 122,102 impressions in total

Several niches accumulated more than 10,000 impressions in total. These were primarily the same niches that ranked for 1,000+ keywords: 

• Hobbies & Interests: 17,425 impressions
• Business & Services: 17,311 impressions
• Travel & Tourism: 13,598 impressions
• Lifestyle & Well-being: 13,072 impressions
• Law & Government: 11,794 impressions
• Games & Accessories: 11,083 impressions
• Vehicles & Boats: 10,677 impressions

For actual clicks, we saw similar trends: niches excelling in impressions also led in clicks.

• Hobbies & Interests: 55 clicks
• Games & Accessories: 41 clicks
• Law & Government: 38 clicks
• Business & Services: 24 clicks
• Lifestyle & Well-being: 23 clicks
• Vehicles & Boats: 18 clicks

The Home & Garden niche, despite not passing the 10K-impression threshold (6,410 impressions), joined the lead in clicks (10 clicks), while the Travel & Tourism domain recorded 13,598 impressions but generated only 9 clicks.

As for underperformers, the Pets & Animals domain, ranked for only two keywords, resulting in just 6 impressions and 0 clicks. The Sports & Fitness niche, despite ranking for 168 keywords, earned 393 impressions but no clicks.

The screenshots below show data pulled directly from GSC (November to December 2024) for each site in this experiment.

Five sites had up to 30% of their keywords ranking in the top 30 on desktop and up to 37% on mobile from November to December 2024

According to data from our Keyword Rank Tracker, five domains had between 20% and 30% of their keywords ranking in the top 30 on desktop from November to December 2024. We checked the “how-to” keywords we used for content generation, not the ones detected by GSC. 

The leading websites were:

• Games & Accessories: 29 out of 100 keywords
• Hobbies & Interests: 29 out of 100 keywords
• Law & Government: 25 out of 100 keywords
• Lifestyle & Well-being: 25 out of 100 keywords
• Vehicles & Boats: 22 out of 100 keywords  

In total: 

• 189 out of the 2,000 keywords (9.45%) across all analyzed domains ranked in the top 30 on desktop.
• 379 queries out of 2,000 (18.95%) ranked in the positions 31-100. 
• 1,432 of the analyzed keywords (71.60%) ranked outside the top 100 on desktop.

In the table below, you can see the distribution of all analyzed keywords by ranking tier across niches in desktop search:

The data for mobile search was better. It shows that the same top five niches (although in a different order) had from 29% to 37% of their keywords ranking in the top 30 on mobile:

• Hobbies & Interests: 37 out of 100 keywords
• Games & Accessories: 32 out of 100 keywords
• Lifestyle & Well-being: 30 out of 100 keywords
• Vehicles & Boats: 29 out of 100 keywords  
• Law & Government: 26 out of 100 keywords

For mobile search:

• 11.85% of analyzed keywords (237 out of 2,000) ranked in the top 30. 
• 16.75% of analyzed keywords (335 out of 2,000) ranked in positions 31-100.
• 71.40% of analyzed keywords (1,428 out of 2,000) ranked outside the top 100.

The table below shows the mobile device-based distribution of analyzed keywords in different ranking groups and niches:

Since February 2025, none of the articles have ranked in the top 100 search results

Although this experiment initially showed promise based on the results above, all of the articles dropped out of the search results on February 3, 2025. 

In practice, they remained in the SERPs and generated minimal traffic for only three months. This outcome was likely driven by two key factors:

1. The sites hosting these articles were new and lacked domain authority.
2. Google tends to deprioritize fully AI-generated content that lacks human input and originality.

Conclusion

Our research shows how AI-written content performs in search and the results it can deliver. AI content that has been edited and refined by our team continues to perform well, while fully AI-generated content has seen no traffic or visibility for the past six months. The takeaway is clear: fully AI-generated content may deliver some initial results, but it’s unlikely to be a good long-term strategy. If you do use AI to create content, it should always be followed by thorough editing, optimization, and other refinements.

We’re continuing to test and explore, analyze the reasons for the drops, and plan to examine how internal links and images affect performance. We’ll also experiment with keyword clusters. In the meantime, you can run your tests by signing up for a free trial of SE Ranking. You’ll get access to tools and data that help you validate any SEO hypothesis, refine your approach, and develop a well-rounded strategy.