As I dive deeper into the world of AI, I’ve come across something truly fascinating about how query language is changing the landscape of AI citations. In our analysis, Profound looked at an astounding 3.25 billion citations spread across seven AI models and fourteen countries. What the data revealed was mind-blowing: the language used in queries is the main catalyst reshaping citation rates across different AI platforms.
Interestingly, I noted that AI tools like Google AI Overviews and ChatGPT handle non-English prompts in uniquely distinct manners. This variation has far-reaching consequences for brand visibility on a global scale, especially within the realms of AI search. The differences in response patterns not only highlight the power of language but also impact how brands are perceived worldwide.
I remember the days when a Google search was akin to embarking on a quest for information. It was an adventure of navigating various links and forming my own opinions.
Nowadays, tools like AI Overviews, ChatGPT, and Perplexity condense all that information into a single, simplified answer. This transformation often strips away the finer details while amplifying certain perspectives.
This shift has redefined online reputation management. Now, search engines not only present information but shape the underlying narratives. This raises the stakes for brands, as even a top-ranking status doesn’t guarantee influence if AI stories tell a different tale.
For brands, the game has changed. Being number one doesn’t ensure visibility and influence anymore. The underlying narrative holds far greater power.
AI Narrative Formation: Crafting User Answers
AI platforms now utilize what I like to call ‘AI narrative formation.’ This process crafts the responses we receive from various search engines. Let me walk you through how this system works.
Source Pooling
These systems pull content from numerous sources. Contrary to expected reliance on peer-reviewed articles, they gather data from Reddit, YouTube, and social platforms like Instagram and TikTok.
Signal Weighting
Not all sources are equal. Often, a popular yet low-quality source can outweigh a singular, credible entry. A bustling Reddit thread with negative feedback might overshadow a well-researched Wikipedia page.
Narrative Compression
The summarization process compresses diverse inputs, often losing nuance along the way. Complex reputations are simplified into general statements like, ‘Users find this company untrustworthy.’
Continued Reinforcement
These summaries transcend their original context, getting shared and re-shared across social media. As these echoes return as new data, they further entrench the narratives in AI responses.
Unraveling a Finance Company’s Reputation in AI Search
To illustrate AI narrative formation, consider a recent case I worked on involving a financial company, which we’ll call Company X.
Company X’s reputation remained strong on traditional SERPs. High Trustpilot ratings and reputable endorsements were the norm until Google AI Overview threads surfaced a forgotten Reddit forum rife with grievances against them.
The AI Overview skewed the narrative, suggesting Company X had unresolved customer service issues, even though these concerns had been addressed years prior. This created a skewed perception that was hard to counteract.
The Amplified Risk from AI Searches
AI dramatically increases reputational risk through several mechanisms:
The Spread of Negative Narratives: Negative content surfaces faster and more prominently than before.
AI Hallucinations: Despite growing awareness, AI inaccuracies continue to deceive.
The Snowball Effect: Repeated narratives gain momentum, complicating reputation management efforts.
It has become evident that in ORM, repetition often overrides accuracy.
Auditing AI-Generated Narratives: A Step-by-Step Approach
Let’s consider a situation involving an AI-generated narrative challenge faced by CEO X of a well-known SaaS company.
After an out-of-context quote from CEO X’s podcast appearance went viral, AI summarized him unfavorably. Quickly, his reputation transformed negatively across major platforms.
Step 1: Mapping Queries
I initiated a process to understand what queries AI outputs were generating about CEO X. This helped identify the underlying issues.
Step 2: Capturing Outputs
Identifying repeated claims revealed how CEO X was perceived. Narratives from Google AI and ChatGPT were consistently portraying him negatively.
Step 3: Delving Through Sources
The next step involved examining the quality of sources contributing to these narratives, often outdated or lacking accuracy.
Step 4: Analyzing the Narrative Gap
This involved assessing discrepancies between AI narratives and his actual reputation, contextualizing the initial quote, and examining the long-standing perception of CEO X.
Step 5: Correcting and Replacing Sources
Finally, I focused on directly addressing, correcting, and replacing those negative narratives. This involved engaging directly with platforms that contributed to the misinformation and reinforcing positive content elsewhere.
A New Perspective: From SEO to Narrative Management
The focus has shifted from merely achieving top SEO rankings to understanding and adapting to narrative shifts. We must rethink our strategy from content engagement to managing the narratives AI disseminates.
To succeed, it’s important to reinforce AI systems with quality inputs, including crafting high-quality content, pursuing credible mentions, disseminating structured data, and managing misinformation directly.
I’ve recently come across news about a fascinating development from Google: the introduction of the Google-Agent user agent. It’s designed to signal when AI agents complete tasks on behalf of users, marking a significant step towards AI-driven web interactions. I’m eager to share what I’ve learned about this new feature and its implications.
What Happened: Google added Google-Agent to its collection of user-activated fetchers on March 20, and it’s currently rolling out gradually. This intrigued me because it means a novel way of tracking AI interactions is becoming available to us.
The Google-Agent user agent identifies requests made by AI programs running on Google’s infrastructure, which includes experimental tools like Project Mariner. It’s fascinating to see how advanced Google is getting in this space.
How It Works: Google-Agent appears in HTTP requests when an AI agent visits a site to complete tasks initiated by users. Imagine it like a helping hand behind the scenes, orchestrating internet tasks for us.
For example, Google-Agent could be used for browsing pages, evaluating content, or performing actions like submitting forms. This differs from traditional crawlers like Googlebot that operate continuously without user prompts. It’s exciting to think about how this technology could evolve further.
IP Ranges: Google has shared the IP ranges for its desktop agent, and notably for its mobile agent as well. This transparency is helpful as it allows us to better manage and identify website traffic.
Why We Care: With this insight, I can now distinguish between traditional crawl activity and visits spawned by users through AI agents using server logs. This capability will enable me to track agent-assisted conversions, understand emerging user behaviors, and prepare for what might be called ‘agentic search’.
What They’re Saying: According to Google’s announcement, “The Google-Agent user agent is rolling out over the next few weeks, and will be used by Google agents hosted on Google infrastructure to navigate the web and perform actions upon user request.” This statement makes me realize the potential impact on our digital interactions.
What to Watch: While early volumes of activity may be low, now is the ideal time to establish a baseline. Monitoring logs for Google-Agent activity ensures I stay informed, and I need to ensure that my CDN and WAF configurations aren’t unintentionally blocking these IP ranges.
Furthermore, it’s crucial for me to validate that key site actions, including forms and user flows, function smoothly for automated agents, ensuring an optimized experience for users.
Google’s foray into AI Overviews is rapidly transforming the landscape of shopping queries. I’ve discovered that these AI Overviews now appear in 14% of all shopping searches—an impressive leap from just 2.1% in November 2025. This data comes from a comprehensive analysis by Visibility Labs.
Ecommerce brands, which previously seemed shielded from the impact of AI-driven click reductions in search results, are now beginning to feel the heat. This change signifies a shift they can no longer ignore.
Why This Matters to Us. As AI Overviews extend their reach across product searches, the risk for ecommerce brands is escalating. The chance of losing visibility and clicks prior to customers engaging with standard organic or Shopping listings is becoming a real concern.
The Analysis. The Visibility Labs study specifically analyzed product-intent keywords associated with results that included a Shopping box, irrespective of whether they were paid or organic. This included terms like “weighted blanket,” “mushroom coffee,” “protein powder,” and “blue T-shirts.”
Out of this extensive research, involving 20,900,323 shopping-related keywords, 2,919,229 keywords featured an AI Overview. This equates to a 14.0% penetration rate.
Expert Opinion. Jeff Oxford, the founder and CEO of Visibility Labs, emphasizes:
“Focusing on AI SEO is no longer a luxury; it’s becoming a necessity. Ecommerce sites must look beyond traditional SEO tactics and start weaving AI SEO best practices into their search optimization strategies.”
I recently came across fascinating research revealing how diverse AI platforms like ChatGPT, Google AI, and Perplexity cite their sources. It’s intriguing to see how each platform approaches sourcing information and the implications for their visibility.
The study highlights substantial differences in citation patterns among these major AI players. This variation in sourcing methods significantly affects how each platform is perceived in terms of reliability and authority.
Understanding these citation patterns can offer valuable insights into the competitive landscape of AI visibility. As we explore this further, it becomes clear why recognizing these differences is crucial for anyone interested in AI optimization.
As I dive into the ever-evolving world of AI search engines, I find myself asking: which one should my brand optimize for first? The options are plentiful, with ChatGPT, Google AI Overviews, Perplexity, Bing, and others vying for attention. The goal is clear: prioritize AI visibility leading into 2026, but the path there is not so straightforward.
Each of these AI platforms offers unique features and potential benefits that can cater to different business needs. It’s crucial for me to assess their capabilities and align them with my brand’s strategic objectives. Whether it’s the conversational prowess of ChatGPT or the data-rich insights from Google AI Overviews, the choice has to drive brand value.
In the process of optimization, understanding the nuances of each platform helps to leverage their full potential. By comparing these engines, I can tailor my approach, ensuring my brand stays ahead in AI visibility, making informed decisions today that will resonate in the future.
Have you ever wondered what it would be like if Google knew exactly what you wanted to search for even before you started typing? Well, that’s the future Google is aiming for.
Currently, Google is pushing this innovation onto our devices with small AI models that rival much larger ones in performance.
What’s happening. In a recent research paper presented at EMNLP 2025, Google researchers have introduced a groundbreaking approach. By dividing “intent understanding” into smaller, manageable steps, they have enabled small multimodal LLMs (MLLMs) to deliver results comparable to more powerful systems like Gemini 1.5 Pro. These models operate faster, at a lower cost, and crucially, they keep data processing on the device.
The future is intent extraction. Presently, most large AI models infer intent from user behavior via the cloud, leading to speed, cost, and privacy issues. By dividing the process into two straightforward steps, Google addresses these concerns effectively with on-device models.
Step one: Each interaction is individually summarized. The model records what appeared on the screen, what action the user took, and a preliminary guess of their intent.
Step two: Another model reviews these summaries, focusing solely on factual information. It dismisses guesses and formulates a concise statement outlining the user’s overall goal for their session. This targeted approach prevents the common pitfalls when smaller models are asked to process long chains of actions at once.
How the researchers measure success. Success is determined with Bi-Fact, where small models employing the step-by-step strategy consistently outperform other small-model methods, as evidenced by their F1 scores.
Models like Gemini 1.5 Flash, despite being only 8B, match the performance of the Gemini 1.5 Pro on mobile data. Errors diminish since unfounded guesses are removed, speeding up operation and reducing costs compared to large cloud-based models.
How it works. Intent is analyzed by breaking it down into distinct facts, identifying missing or fabricated details. This process reveals how and where understanding fails, offering insights into how systems misinterpret meaning and miss crucial information.
The research further shows that noisy training data impacts large end-to-end models more significantly than this structured approach. The decomposed system remains robust against the unpredictability of real user behavior.
Why we care. For Google to develop tools that suggest actions or answers before a query is entered, understanding user intent from behavioral patterns across apps, browsers, and screens is essential. This research is a major step towards that vision. Although keywords will remain important, optimizing for clear, logical user paths will take precedence over mere query inputs.
Hey there! If you’re like me, you’re probably always looking for ways to make your content more effective, especially in today’s AI-driven world. I’ve discovered nine crucial changes that can transform your content, making it AI-friendly. This means platforms like Google AI Overviews, ChatGPT, and Claude will be able to parse, trust, and cite your pages more efficiently.
First, let’s talk about understanding how AI algorithms work. It’s essential for ensuring that your content is optimized for AI parsing. I’ve found that using structured data and schema markup can significantly enhance the way AI understands and displays content.
Another vital aspect I focus on is creating concise, informative headings. These help both readers and AI systems grasp the main points quickly. Remember, clear and direct headings often lead to better AI interpretation and can enhance your SEO performance.
I’ve also made it a point to ensure my content is easily accessible. This includes optimizing for mobile users and ensuring fast loading times. Not only does this appeal to AI algorithms, but it also improves overall user experience, which is a win-win!
Moreover, I pay close attention to the language used in my content. Simpler, jargon-free text is easier for AI to process. This approach not only makes my content more understandable for AI but also broadens its readability for a wider audience.
Integrating relevant keywords is another strategy I use to ensure my content is AI-friendly. These keywords help AI platforms accurately categorize and display my content, increasing visibility and reach.
Finally, I always review my content for accuracy and relevance. Keeping information up-to-date ensures that AI systems can trust and effectively utilize the content I produce, which is crucial for maintaining authority and credibility online.
I’ve been diving deep into the fascinating world of Generative Engine Optimization, or GEO, as it’s reshaping the $80 billion SEO market. With insights from the renowned Andreessen Horowitz, I’m excited to explore how AI search engines like ChatGPT, Perplexity, Google AI, and Apple’s Siri are evolving and impacting our strategies.
The surge of AI-powered tools is transforming how we approach digital marketing. In such a rapidly changing environment, staying updated with GEO strategies is crucial. Thankfully, A16Z provides invaluable guidance to navigate these changes effectively.
As someone passionate about SEO and AI, I find the integration of AI in search engines like ChatGPT and Google’s AI Overviews captivating. These tools not only enhance user experience but also demand nuanced optimization tactics.
Apple’s Siri and AI-driven searches are continuously pushing the envelope, making it vital for us to adapt our SEO strategies. Leveraging these insights can significantly elevate our digital marketing efforts and ensure we remain competitive.
Join me as I delve into these transformative insights from A16Z, exploring how we can refine our GEO strategies for a future dominated by AI-driven search engines.
Upon evaluating a whopping 10,000 keywords, I’ve discovered an intriguing insight: pages that successfully rank for Google AI Overview ‘fan-out’ queries are significantly more likely to be cited. In fact, they account for more than half of all citations on these platforms.
From my analysis, it’s clear that pages leveraging these queries dramatically increase their chances of being referenced. As data from Surfer SEO suggests, these pages offer more citation opportunities compared to those focusing solely on the main search query.
An analysis of these 10,000 keywords revealed a strong correlation—precisely, a Spearman of 0.77—between the volume of fan-out queries a page ranks for and its likelihood of citation in Google’s AI Overviews.
Diving into the numbers. I found that pages ranking for fan-out queries are 161% more likely to be cited than those ranking exclusively for the main query. Consider this:
76% of the keywords evaluated triggered AI Overviews.
Through Gemini, I extracted 33,000 fan-out queries.
Pages ranking for both the main query and at least one fan-out constituted 51% of AI Overview citations.
In contrast, pages ranking solely for the main query accounted for just under 20%.
Fan-outs outshine the main query. Recognizing the power of ranking for fan-out queries, I noticed such rankings were 49% more likely to earn citations than merely ranking for the main term. When the AI Overviews chose to reference organic results, here’s what stood out:
Approximately 20% of cited pages ranked only for the main query.
Conversely, around 30% ranked exclusively for fan-out queries.
Most AI citations skip top ranks. Fascinatingly, about 68% of cited pages didn’t appear among Google’s top 10 results for either their main or fan-out queries. However, for the top three most prominent citations, this figure dropped to roughly 46%.
But there’s more. It’s crucial to understand that correlation doesn’t equate to causation. Additionally:
Achieving a ranking for fan-out queries alone won’t guarantee an AI Overview citation.
User context and personalization affect fan-outs, with only about 27% remaining constant across test runs.
Normal SEO practices don’t fully determine citation selection.
Why this matters to us. If your goal is to be cited in AI Overviews, striving for broader topic authority might be the answer. Surfer SEO advises crafting extensive topical content around core subjects, creating content that naturally responds to a variety of related questions, and allowing AI Overviews to recognize your pertinence across different fan-outs.
