In late 2024, I embarked on an eye-opening 16-month journey with SE Ranking’s research team to test the performance of AI-generated content in organic search. We launched 20 diverse websites, eagerly tracking their progress.
But my curiosity didn’t end there. I was driven to comprehend how AI systems find, process, and use information. This inspired me to expand our project and delve deeper into AI search and LLM visibility experiments.
In our next phase, we boldly created a fictional brand and inserted it into a real, competitive niche. Our aim? To see how fast AI would catch on and if our make-believe brand could stand toe-to-toe with industry giants and governmental sources.
After just one month, enlightening patterns began to emerge.
Methodology behind the experiment
I crafted a fictional brand and dispersed content across various platforms:
- A fresh website exclusively for the brand, registered specifically for this daring experiment.
- 11 seasoned domains, each over a year old with a solid history and existing rankings.
I experimented with seven different content formats:
- Comprehensive guides.
- “Alternatives” listicles.
- “Best of” listicles.
- Review articles.
- Comparative (“vs”) pages.
- How-to/tutorial content.
- Clickbait-style articles.
Kicking off in March 2026, I monitored five AI systems: ChatGPT, Google’s AI Overviews, Google’s AI Mode, Perplexity, and Gemini, tracking 825 prompts and generating 15,835 AI answers during the initial month.
For every prompt, I considered:
- Our brand’s appearance in AI responses.
- Its recognition as a source.
- Frequency of being the main cited source (position 1).
This ongoing experiment was initially designed to observe AI systems’ reactions to freshly created, fictitiously branded information.
Key experiment insights
- 96% of our brand’s AI visibility stemmed from branded searches. Even in a low-competition niche, a new domain struggled to compete on non-branded topics.
- For niche-specific queries, our brand outshined well-established competitors by up to 32 times, achieving dominant visibility in under 30 days.
- Despite lacking authority, clearly articulated identity pages, like “[Brand Name] Complete Guide” and “About Us”, became frequently cited, highlighting the importance of brand positioning in AI.
- Perplexity surfaced new content swiftly, often citing additional domains over the main site.
- Google’s AI Mode offered stability on branded queries.
- Gemini struggled with brand identification, resulting in 60% of responses without our brand’s citation for uniquely branded queries.
- Deep guides, review articles, and comparison pages gained the most citations, while generic content saw minimal impact.
- A hub page with 10 supporting articles yielded no citations, whereas shorter, repetitive pages garnered over 1,800 citations, emphasizing the power of high-volume content publishing.
Insight 1: New domains may not beat market leaders right away, but they can define their brand narrative in AI search
A new site struggles to compete broadly initially. However, our fictional brand quickly gained traction through branded queries, largely because these were the focus points.
Of all AI answers, a staggering 96% came from branded searches alone, reiterating the crucial role of brand-specific queries in early visibility.
This mirrors traditional SEO patterns where new brands must first build trust and recognition.
My key takeaway for marketers was clear: AI systems are inclined to use your site as a primary information source during your brand’s formative years.
This insight was reinforced as pages consolidating brand information, such as the “Complete Guide” and “About Us”, became the primary sources cited from our main domain.
Therefore, shaping the brand narrative early on AI platforms is crucial, even for emerging brands.
Insight 2: AI engines behave very differently
Our experiment shed light on the unique behaviors of five AI systems in indexing and presenting our fictional brand.
Google’s AI Mode: The most stable for branded visibility
Google’s AI Mode proved to be a reliable ally, consistently putting our brand at the top for around 90% of branded queries.
It was the bastion of predictable brand visibility in our experiment.
Google’s AI Overviews: High visibility, lower consistency
Though less consistent, Google’s AI Overviews provided notable brand visibility. Yet, fluctuations and temporary drops were observed during our test period.
Whenever links were absent, visibility suffered, highlighting the need for sustained link presence.
Perplexity: The fastest to pick up new content, but not always brand-first
Perplexity swiftly indexed new content, quickly boosting early visibility.
However, its affinity for additional domains over the main brand site complicated content attribution in AI responses.
ChatGPT: Slower to react, stronger over time
ChatGPT gradually improved recognition of our brand, with a notable increase in visibility over March.
Notable growth occurred in unique claims and comparisons (“vs”), showcasing ChatGPT’s potential for longer-term brand assimilation.
Gemini: Weakest performance and most inconsistent behavior
Gemini presented challenges with niche recognition, improving only when framing prompts appropriately.
Despite effort, results remained inconsistent, with significant citation gaps on brand-specific queries.
Insight 3: Content format matters, but so does the volume
Through diverse content experimentation, we found in-depth articles earn the most AI citations.
Comprehensive guides, reviews, and comparisons outperformed simpler formats, reinforcing the power of detailed content presentation.
The volume of content also played a role. Although the individual performance was low, 30 shorter pages collectively generated impressive AI visibility.
This doesn’t diminish the value of quality but indicates a large amount of content can boost overall reach.
Insight 4: Topical clustering alone doesn’t produce AI visibility
Our structural tests revealed that topical clustering, without substantial content, didn’t boost AI visibility.
It challenges the notion that clustering inherently strengthens authority, stressing the importance of standalone content value.
Though structured linking offers insight into site understanding, AI systems prioritize the need for direct and valuable information retrieval.
So, do AI engines reward entity coherence more than truth verification?
Our first month’s results point to a significant insight: AI systems value availability and consistency over strict truth verification.
Though not all-reaching, well-structured, repeated, and available content can be surfed with surprising ease.
This phenomenon was observed during manual checks where even a fictional brand received favorable recommendations due to consistent narratives.
It’s not simply LLMs favoring new brands, but where gaps exist, even limited information may be built up positively.
Final thoughts
The true revelation isn’t the visibility of a fictional brand. Rather, it’s how visibility aligns with brand-centric inputs like unique claims and varied content.
This leads to pivotal conclusions:
- AI search isn’t arbitrary. It responds to discernible and influenceable signals.
- AI remains vulnerable to manipulation. Without inherent truth-checking, strategies used by legitimate brands can simulate credibility.
Illuminating the need for active narrative shaping, our experiment urges businesses not to rely on AI systems to innately capture accurate brand representation.
We’re committed to expanding and monitoring these insights over time, as we collect ongoing data.
Inspired by this post on Search Engine Land.
