Analyzing LLM referral traffic has opened my eyes to intriguing trends regarding volume, growth, citation shifts, and an impressive 18% conversion rate.
Discussing LLMs and their impact on website traffic has become a staple in my client consultations. I’m often asked about current trends, potential improvements, and established best practices.
For brands eager to navigate these waters, my advice is straightforward: begin with the data you can rely on.
To understand how LLM traffic influences key metrics, I thoroughly analyzed 13 months of LLM prompt referral traffic within Google Analytics from our customer base (Jan. 1, 2025, to Feb. 7, 2026).
We concentrated on traffic from various LLM models to brand sites and the conversion events that align closely with substantial business outcomes, such as purchases or lead generation.
Our analysis unveiled four significant insights:
LLM referral traffic remains modest.
LLM traffic is growing rapidly.
Sources mentioned in responses are evolving.
LLMs have a high conversion rate compared to other channels.
LLM Referral Traffic is Still Small
Our dataset reveals that LLM referral traffic constitutes less than 2% of total referral traffic. This means that fewer than 2 out of every 100 site visitors come from an LLM source.
The figures vary between 0.15% and 1.5%, with sources like ChatGPT, Perplexity, Gemini, and Claude.
Though a hot topic, it’s not yet the top concern for immediate financial impacts for many businesses.
… (The rest of the content should follow the same structure, formatted as Gutenberg paragraph blocks) …
In this rapidly evolving space, I believe staying focused, driving innovation, and leveraging data can give brands a strategic advantage over competitors.
I interact with LLMs daily, both at work and in my personal projects. For many of us in tech, leveraging these language models has become second nature.
It’s well-known that folks in the tech sector, like me, engage with LLMs at twice the rate of the general population. In my case, LLM usage often exceeds a full day each week.
Even as regular users, we sometimes find ourselves frustrated when an LLM doesn’t quite deliver the responses we expect. Here’s how I effectively communicate with LLMs during vibe coding sessions. These insights are just as valuable when navigating extended interactions with an LLM UI like ChatGPT.
Choosing My Vibe-Coding Environment
Vibe coding is the art of co-creating software with AI. I lay out my vision, the AI generates code, and together we refine it to match my intent. However, the process isn’t always smooth sailing.
The first step in my workflow involves choosing a coding environment. This space serves as a hub for interacting with the LLM, drafting, and executing code. I’m partial to Cursor, having started on their free Hobby plan, but I’ve since upgraded to the Pro+ account due to my extensive usage.
For those interested, here are some environment options:
Cursor: Widely used by vibe coders for its customizable interface.
Windsurf: An alternative that executes terminal commands independently.
Google Antigravity: A unique option favoring agent-driven development.
In my examples, I’ll be using Cursor, but the principles are applicable across platforms. Even if you’re simply delving deep into LLM conversations, the same guidelines apply.
Why Prompting Alone Isn’t Enough
You might ask why we’d even need a tutorial for vibe coding. It’s true—the basic idea is simple: specify an outcome, and the LLM delivers. However, once the complexity increases, especially when dealing with multifile systems or tools, context management becomes crucial.
The context window is a pivotal concept. It’s the memory scope LLMs use to handle input/output data, a window defined by token limits. For example, GPT-5.2 allows a 400,000-token window, while Gemini 3 Pro goes up to 1 million. Understanding this helps in avoiding token overflow, which can diminish retrieval accuracy.
Expert commentator Matt Pocock explains the nuances of context windows well—view his YouTube video for more insight. For now, keep in mind that effective planning minimizes verbosity and assumes clear window management.
One team, one dream. Divide projects into manageable phases, clearing LLM memory regularly between tasks.
Do your own research. While you don’t need exhaustive detail, grasp general methods and potential build paths.
Trust but verify during troubleshooting. Get clarifications from the LLM and cross-check details externally.
Tutorial: Creating an AI Overview Question Extraction System
To produce high-ranking content in AI Overviews, address the questions they respond to. This tutorial guides you in developing a tool to extract such questions, not just to provide a use case but also to demonstrate effective system development via vibe coding. It’s not a guaranteed path to AI prominence but offers strategic insights.
Step 1: Planning
Before diving into Cursor or any other tool, identify your goals and necessary resources. Although it’s early days, using generative AI for initial brainstorming can be beneficial. I often start by articulating my end goal in a sentence or two, alongside requisite steps, in AI tools like Gemini or ChatGPT. Missteps here are okay—this stage is about outlining thoughts, not finalizing builds.
For instance, I could outline:
I’m an SEO, aiming to leverage Google's AI Overviews to inspire our authors' content. We need to extract implicit questions addressed by AI Overviews. Proposed steps include:
1 – Choose a keyword target.
2 – Run a search and collect the AI Overview.
3 – Deploy an LLM to derive underlying questions from the AI Overview.
4 – Preserve questions in an accessible format.
With a clear direction, select your preferred LLM. While I’m partial to Gemini for chats, modern models with robust reasoning will suffice. Initiate a session, state your intent to build an AI Overview extractor, and share your planning prompt.
Step 2: Laying the Foundation
Cursor offers diverse models which I find advantageous. For this task, start in Plan mode, allowing for structured discussions and informed decision-making.
Kick off discussions with our defined project prompt.
Making modifications is crucial, so carefully review the LLM’s plan to ensure alignment with your vision. Address any disparities through collaborative discussions with the model.
Consider seeking insights into possible project failure points and implement preventive measures accordingly. For efficiency, I tend to request models to generate outline files for improved context window management, validating internal consistency before proceeding.
Step 3: The Build
With the foundation laid, shift to Agent mode using your selected model—in my case, Gemini 3 Pro—to execute the building phase. Keep an eye out for required approvals during script execution to ensure a smooth process.
Once script development is complete, proceed with library installations via the provided requirements.txt file. For organized dependency management, setting up a virtual environment is recommended.
Running your first script execution often surfaces unforeseen challenges. Tackle these by leveraging comprehensive diagnostic feedback, ensuring issues are resolved before moving forward.
Troubleshooting and Improvements
My initial run revealed a lack of expected AI Overview detection—a misstep rectified through close inspection of terminal outputs, model adjustments, and informed re-execution.
Embrace troubleshooting as a key growth component in the vibe coding journey, enhancing reliability and performance as you fine-tune system components.
Employ Weave for maintaining organized records of query inputs and LLM outputs. This robust tool aids in both immediate log assessment and long-term query-trace reference.
Use the analyze_query trace to monitor pivotal data points, fostering awareness of the direct connection between query intentions and AI Overview content insights.
Structure Over Vibes: A Strategic Approach
Across my years of vibe coding, I’ve learned structure creates reliability—increasing complexity demands methodical workflows, ensuring sustainable success. Remember to keep the vibes in your collaborations strong, united by a shared purpose and approach.
I’ve noticed how search is evolving far beyond the typical blue-links framework. Now, discovery often happens within AI-generated answers—whether it’s Google AI Overviews, ChatGPT, or other LLM-driven platforms. It’s clear to me that visibility is no longer just about rankings, and influence doesn’t always lead to a click.
Traditional SEO metrics like rankings, impressions, and CTR seem to fall short as search becomes more recommendation-driven and attribution becomes increasingly opaque. Clearly, a new measurement layer for SEO is needed.
This is where LLM consistency and recommendation share (LCRS) steps in. It helps measure how reliably and competitively my brand appears in AI-generated responses. It’s a modern equivalent to keyword tracking, tailored for the LLM era.
Why traditional SEO KPIs are no longer enough
Traditional SEO metrics worked well when visibility was tied directly to ranking positions and user interaction pivoted on clicks. This relationship weakens in LLM-mediated searches. Even if my page ranks at the top, it may never appear in an AI-generated answer.
LLMs might favor another source with lower traditional visibility, exposing a flaw in conventional traffic attribution. Here, brand influence might occur without a measurably corresponding website visit. The impact exists but isn’t reflected in the traditional analytics landscape.
At the heart of this change is something that traditional SEO KPIs were not developed to handle:
Being indexed means my content is available for retrieval.
Being cited means it serves as a valuable source.
Being recommended highlights my brand as an active solution or answer.
Traditional SEO analytics often stop at indexing and ranking. However, in a world dominated by LLM-driven search, the true competitive edge lies in recommendation—a dimension current KPIs struggle to quantify. This is where the gap between influence and measurement creates a space for new performance metrics.
LCRS: A KPI for the LLM-driven search era
With LLM consistency and recommendation share, I can gauge how reliably my brand surfaces and is recommended by LLMs during search and discovery processes.
LCRS answers a crucial question that traditional SEO metrics can’t: When users look to LLMs for guidance, how often and consistently is my brand part of the conversation?
It evaluates my visibility across three dimensions:
Prompt variation: Different user ways of asking the same question.
Platforms: Various LLM-driven interfaces.
Time: Consistent appearances over time, not just one-shot mentions.
LCRS is less about isolated citations and more about establishing a repeatable, comparable presence, enabling me to benchmark against competitors and track changes.
Although it’s not a replacement for established SEO KPIs, LCRS enhances them by addressing zero-click search scenarios where recommendations determine visibility.
Breaking down LCRS: The two components
LCRS comprises two primary elements: LLM consistency and recommendation share.
LLM consistency
In LCRS, consistency measures how reliably my brand appears across similar LLM responses. High consistency means my brand surfaces across numerous, semantically similar prompts rather than relying on a single high-performing query.
Considerations like prompt variability, temporal variability, and platform variability come into play. Consistency reflects durable relevance beyond transitory exposure.
Recommendation share
While consistency focuses on repeatability, recommendation share assesses competitive presence. It examines how frequently LLMs recommend my brand relative to others in the same category.
Not all appearances count as recommendations; it’s about how often my brand is positioned as a primary choice against competitors, reflecting the portion of recommendation space occupied.
How to measure LCRS in practice
To effectively measure LCRS, a structured approach is necessary, one that replaces anecdotal observations with repeatable sampling reflective of actual user interactions.
1. Select prompts
I start with choosing prompts representing my category, ensuring they include variations in phrasing to capture natural language nuances.
2. Confirm tracking
The choice between brand-level and category-level tracking hinges on focus. Most insightful at the category level, LCRS shows which brands LLMs choose to highlight.
3. Execute prompts and collect data
Since managing data volumes is a challenge, I rely on programmatically executing prompts and parsing responses to identify which brands are recommended.
4. Analyze the results
Automated data capturing is key, though human review is crucial for interpreting nuanced information. Tracking analysis over time is essential for stable directional signals.
Use cases: When LCRS is especially valuable
LCRS is particularly valuable in environments where synthesized answers shape decisions. In marketplaces, SaaS, YMYL industries, and comparison searches, LLMs significantly influence visibility.
Limitations and caveats of LCRS
LCRS offers directional insight rather than definitive certainty, given LLMs’ non-deterministic nature. Short-term volatility is expected, so evaluating trends over time is vital.
This metric isn’t a replacement for traditional analytics but complements them by addressing influence areas without direct attribution.
What LCRS signals about the future of SEO
More than a ranking tool, LCRS signals a shift toward brand presence engineering in the LLM-driven discovery space. Brand authority is becoming crucial, with successful SEOs adapting to optimize for retrievability, clarity, and trust.
The shift from position to presence
As LLM-driven search reshapes discovery, expanding from ranking positions to presence and recommendation is crucial. LCRS allows me to explore this gap and complement existing performance metrics for a comprehensive visibility strategy.
My journey with LCRS shows that adapting SEO strategies for evolving landscapes boosts both visibility and influence within LLM-driven search experiences.
I’ve been following the lively debate around creating separate markdown pages for LLMs, and it appears that both Google and Bing are advising against this approach.
Recently, I noticed that representatives from Google Search and Bing Search have specifically recommended not to create separate markdown (.md) pages designed exclusively for LLMs. This practice involves presenting different content to the LLMs compared to what users see, which can be considered a form of cloaking—a direct violation of Google’s policies.
The question arose when Lily Ray inquired on Bluesky about the prevalence of creating markdown or JSON pages targeted at bots.
“Not sure if you can answer, but starting to hear a lot about creating separate markdown / JSON pages for LLMs and serving those URLs to bots.”
Google’s stance, as explained by John Mueller, is clear. He replied to Lily’s query saying that LLMs have always interacted with standard web pages and don’t require separate markdown pages.
“I’m not aware of anything in that regard. In my POV, LLMs have trained on—read & parsed—normal web pages since the beginning, it seems a given that they have no problems dealing with HTML. Why would they want to see a page that no user sees? And, if they check for equivalence, why not use HTML?”
John Mueller even criticized the whole idea, stating:
“Converting pages to markdown is such a stupid idea. Did you know LLMs can read images? WHY NOT TURN YOUR WHOLE SITE INTO AN IMAGE?” Of course, converting your entire site to a markdown format is an extreme measure.
I’ve collected many of John Mueller’s remarks on this topic, which you can find here.
Bing’s perspective is shared by Fabrice Canel from Microsoft Bing, who suggested that creating duplicate, non-user content isn’t effective.
“Lily: really want to double crawl load? We’ll crawl anyway to check similarity. Non-user versions (crawlable AJAX and like) are often neglected, broken. Humans eyes help fixing people and bot-viewed content. We like Schema in pages. AI makes us great at understanding web pages. Less is more in SEO!”
Why this matters to us: Many of us are tempted by shortcuts to improve search engine performance. Yet, these shortcuts often backfire or yield short-lived benefits. As Lily Ray remarked on LinkedIn, managing duplicate and differing content for bots violates established search engine policies.
“I’ve had concerns the entire time about managing duplicate content and serving different content to crawlers than to humans, which I understand might be useful for AI search but directly violates search engines’ longstanding policies about this (basically cloaking).”
Analyzing nearly two million LLM sessions across nine industries throughout 2025 was a fascinating journey for me. I began with the assumption that ChatGPT would dominate and that AI usage patterns would be relatively uniform with minimal impact.
The findings, however, were surprising.
While ChatGPT does indeed control 84.1% of the trackable AI discovery traffic, it’s primarily serving as a broad-market tool. This discovery significantly impacts strategic approaches.
In today’s landscape, relying solely on a single discovery strategy is not viable. A multi-platform approach that aligns with how and where users find productivity is essential.
Brands must now discern which platforms are empowering productivity rather than merely supporting initial discovery phases.
Various LLMs are excelling in different sectors, often with stark differences. The key takeaway for 2026 is more complex than simply focusing on ChatGPT.
Here’s what I’ve discovered from the data.
The Growth Rate Divergence: ChatGPT vs. Competitors
Throughout 2025, major LLM platforms exhibited significant growth discrepancies:
ChatGPT: 3x growth
Copilot: 25x growth
Claude: 13x growth
Perplexity: 1x growth
Gemini: 1x growth
Although ChatGPT grew, Copilot and Claude experienced much more rapid growth. Platforms like Perplexity and Gemini remained steady, reinforcing specific workflows.
These numbers highlight strategic priorities:
Satya Nadella celebrated Copilot reaching 100 million monthly users.
Dario Amodei revealed that Anthropic’s revenue grew from $100 million to $8–10 billion in under two years.
Aravind Srinivas noted significant interest in Perplexity Finance.
The focus on growth is crucial because it signals true user value:
Copilot excels in the Microsoft ecosystem.
Claude appeals to developers.
Perplexity thrives among finance professionals.
Different LLMs are thriving in various industries at markedly different rates.
Pattern 1: Copilot’s Striking Growth
Copilot’s remarkable 25x growth is indicative of its premier position in B2B environments reliant on Microsoft tools.
SaaS
ChatGPT: 2x growth
Copilot: 21x growth
The rapid adoption mirrors modern SaaS practices, embedding LLMs directly into workflows.
Education
ChatGPT: 6x growth
Copilot: 27x growth
Copilot benefits from educational settings fostering knowledge sharing and synthesis.
Finance
ChatGPT: 4.2x growth
Copilot: 23x growth
Finance aligns with Copilot due to automation needs and context dependency.
Copilot’s growth is most pronounced in industries where professionals are deeply integrated with Microsoft tools.
Instruments like Excel transform into data interpretation powerhouses with Copilot, eliminating the need for external searches.
Implications
For work-centric audiences like SaaS, finance, and education specialists, AI discovery is shifting into LLMs embedded in workflows.
Pattern 2: Perplexity Shines in Finance
While Perplexity has flat growth overall, it stands strong in finance with a 24% market share, unlike in other sectors where it has diminished.
SaaS: down to 7.3%
E-commerce: down to 3.4%
Education: down to 5.2%
Publishers: down to 3.6%
Finance demands accuracy; thus, traceable sources make Perplexity vital in this sector.
Partnering with Benzinga, FactSet, and others, Perplexity offers in-depth data vital for financial decisions.
Trust and verifiability are crucial in finance, and that’s where Perplexity excels.
Implications
In finance, selection of platforms that integrate with licensed data and credible sources is critical. Success hinges on being part of these authoritative ecosystems.
Pattern 3: Claude’s Dominance in Analysis
With just a 0.6% share, Claude might appear to be an underdog, but it thrives in specialist sectors like publishing and finance.
Publishers: 49x growth
Education: 25x growth
Finance: 38x growth
SaaS: 10.3x growth
Claude’s strength lies in standalone, strategic thinking rather than integrated tools like Copilot.
Publishing professionals and financial analysts use Claude for its substantial context window, enabling complex and strategic queries.
Implications
Target audiences that require in-depth analysis should focus on creating structured and detailed content. Claude’s user base is smaller but highly influential.
Pattern 4: Challenges in Tracking Gemini
The data concerning Gemini is puzzling, showing both growth and declines. This could be attributed to issues with attribution rather than an actual decline in users.
Education: −67% tracked traffic
SaaS: +1.4x growth
Finance: +1.3x growth
E-commerce: +2.7x growth
Gemini’s interaction model keeps users within its ecosystem, making measurement challenging.
The reality is that usage might still be robust, but the tracking systems need to catch up with user behaviors.
Implications
As AI-assisted conversions increasingly occur, traditional last-click attribution models need reconsideration.
Monitor brand search performance and invest in broader visibility strategies.
Strategizing Your LLM Approach
AI discovery is diversifying rather than converging. Tailoring strategies based on your audience’s preferences and behaviors is crucial.
Enterprise Audiences: Focus on Copilot integration for SaaS and B2B environments.
High-Stakes Decisions: Consider Perplexity’s reliability in providing traceable data.
I remember when link building was the cornerstone of SEO. While it’s still relevant, its role has evolved as Google set clearer standards, focusing more on quality, relevance, and intent.
Today, in our AI-driven search world, the focus has shifted towards brand mentions, which have become a critical SEO initiative. Brand mentions provide references similar to citations, but in AI search, they explain how brands appear in LLMs (Large Language Models).
Brand mentions are now influential factors for AI search strategies and are gaining more weight in traditional SEO algorithms. Focusing on them should be a priority in 2026 to ensure lasting organic visibility.
Let me guide you on how we can prioritize and benefit from brand mentions.
How and Why to Prioritize Brand Mentions
Brand mentions have become essential in our AI search environments, moving beyond just backlinks. LLMs focus on analyzing mentions, context, and the recurring links between your brand and your target topics.
These mentions form a competitive advantage, especially as they accumulate over time, creating a protective ‘ranking moat’ when competitors don’t invest similarly.
To properly prioritize, ensure your brand’s technical and content fundamentals are solid. This includes crawlability, structured data, and clear on-page content. Afterward, focus on brand mentions before engaging in large-scale content production without an existing citation footprint.
When seeking impactful brand mentions, it’s crucial to examine their sources. My agency goes beyond standard tools, looking for opportunities through systems like Profound that highlight relevant brand mentions aligned with key topics.
We also review AI Overview links for SEO queries and dive into top-ranking Reddit threads to identify frequently mentioned entities related to important keywords.
You can uncover links to source articles in AI Overviews by selecting the chain-link icon, enhancing your brand’s topical visibility.
Driving Passive Brand Mentions
Passive brand mentions come when your content naturally fills an informational gap. The aim is to become the go-to reference for certain topics, achieving this by creating assets that are easily referenced.
These can include original data, insightful reports, or highly scannable explanatory pages. By establishing your brand as the primary source, you’re better positioned for more mentions.
Actively Soliciting Brand Mentions
For proactive outreach to earn brand mentions, focus on building genuine relationships and providing valuable information. Start by sharing assets that offer clear benefits, without immediately asking for something in return.
When contacting journalists or content creators, make your pitches relevant and timely, with a clear angle that increases your inclusion chances. Combining outreach with thought leadership, through podcasts or panels, enhances discovery possibilities.
Our goal is to establish a robust outreach engine, nurturing relationships so that those individuals may naturally reference your brand in the future, potentially leading to collaborative content opportunities.
Deciding When to Engage a PR Resource
PR support is particularly beneficial when you have compelling stories or data but face distribution challenges. It’s also crucial for quick scaling of brand mentions, especially during fundraising, launches, or when competing in aggressive markets, like health or AI.
However, if foundational SEO or assets are lacking, focus on establishing those first. Once ready, PR will accelerate visibility across search engines and LLMs.
The core tenets of link building still apply: aim for quality over quantity and avoid low-impact sources. By keeping a clear focus on key sources and strategy, your brand can achieve significant improvements in search visibility.
Hey there! I’ve been diving into ways to develop an effective AI-ready content strategy that’s perfect for large language models (LLMs) to parse, trust, and cite. It’s fascinating how the focus has shifted from just getting clicks to ensuring understanding through visibility. Let me walk you through my journey of crafting this strategy.
Imagine building a content framework where AI tools not only recognize but also rely on the information you provide. This is where content tailored for LLMs comes into play. It’s all about providing data that these models find credible and resourceful. Essentially, visibility is now measured by how well the content communicates rather than just its ability to attract clicks.
As I started building my strategy, I focused on ensuring that the content is structured and detailed enough for LLMs to easily process and extract valuable insights. This involves more than just surface-level content optimization but delves into creating comprehensive narratives that AI can effectively utilize.
Every year, Black Friday offers a unique glimpse into how consumers search, compare, and decide. This year, it added another layer: it became a real-world arena to see how AI models comprehend commerce amidst genuine demand.
I embarked on a journey to test major large language models (LLMs), analyzing 10,000 responses to understand how these systems perceive the retail landscape and the signals that shape their responses.
As I dissected the dataset, a pattern was unmistakable: Black Friday acts as a genuine stress test for AI-driven discovery.
The sheer number of queries and the diversity of categories reveal the sources, structures, and behaviors LLMs rely on for reasoning about products, retailers, and consumer intent.
The outcomes offer a sneak peek into how AI search is transforming—and how this will impact the broader commerce ecosystem.
TLDR; LLMs lean heavily on a limited range of external domains with YouTube, large retailers, and U.S. review media leading the charge.
Generalist retailers dominantly capture nearly half of all retail citations, serving as the recurring funnel LLMs use to address shopping queries.
Social and user-generated content see an 8.1% surge during Black Friday, as conventional retail and media sites experience a decline.
Off-page signals like Reddit, YouTube, Amazon, and Consumer Reports are vital, equally important as on-page content for shaping LLM comparisons and recommendations.
Structured comparison content wields significant influence, far surpassing branded assets.
The behavior of LLMs differs not only from Google but also from each other, with each platform like Gemini, OpenAI, and Perplexity offering unique formats, lengths, and reasoning patterns.
Unlike traditional search, where the process begins with a query leading to a list of ranked results, AI search reverses this. It starts with a model’s internal web of relationships, sources, and signals to construct a response.
In our review of the top 50 most-cited domains across 10,000 LLM responses—all centered around deals, reviews, and product recommendations—the distribution was notably skewed:
YouTube led with 1,509 citations, followed by Best Buy with 950, Walmart with 885, Target with 477, TechRadar with 355, RTings with 342, and Consumer Reports with 325.
This cluster shapes much of the commercial “knowledge” from which LLMs draw. It gravitates towards large retailers, global media outlets, and platforms specializing in comparisons and reviews.
In analyzing 10,000 responses, I compared the week leading up to Black Friday with the event itself. Pre-Black Friday, responses reins focused on planning behavior.
Retail and brand domains: 59.6%
Media: 23.4%
Social and user-generated content: 17%
When Black Friday commenced, the mix rapidly evolved. Social and UGC content jumped to 25.1%, gaining significant share, while retail and media slightly retreated.
This shift within the models mirrors consumer behavior but also highlights the models’ reliance on conversation-driven content for in-the-moment decision cues.
One of the most transparent insights is the weight third-party domains carry on AI reasoning. Today’s LLMs thrive by absorbing as much human interest in products as possible. Huge volumes of consumer insights, reviews, product demos, sentiment, and structured data guide how models reason and decide.
An analysis revealed key off-page signals LLMs depend on:
Reddit: 34%
YouTube: 19.5%
Amazon: 15.5%
Business Insider: 9.2%
Walmart: 8.9%
Each domain influences different aspects of the model’s decision-making. Across the board, LLMs lean on content that captures human interest, organizes consumer options, and mitigates uncertainty through verifiable data.
While third-party domains reign supreme, brand websites still hold measurable sway. They are vital for any consumer brand aiming to excel in AI discovery.
A site’s architecture plays a crucial role in how a model interprets a brand. Homepages account for 40% and serve as the primary identity layer—establishing tone, positioning, and offering quick semantic signals to models.
Blogs and product pages clarify brand definitions and long-tail context, providing the factual details models need.
Brands that rely too heavily on promotional copy, weak hierarchies, or thin product content risk sacrificing major visibility.
Across the entire dataset, certain retailer categories led the charge in model responses.
Generalist retailers hold 48% of the conversation. Walmart, Target, and Best Buy capture almost half of all retail citations. Their range, familiarity, and content depth make them central figures in LLM commerce reasoning.
Electronics specialists grasp 23% of the share. Best Buy leads, trailed by Newegg and Micro Center, with tech-focused queries often directing models toward these sources.
Other verticals lag behind. Despite strong category leaders, sectors like fashion, beauty, and home capture smaller portions due to the content volume disparity compared with generalist retailers.
Reviewing the platforms uncovered another pattern: major LLMs not only offer different answers but exhibit distinct thinking styles. Each platform has its own rhythm, structures, and styles for presenting commercial information.
Gemini provides the most detailed responses, with essays averaging 606 words, using lists and headings extensively.
OpenAI stands in the middle, averaging 401 words per response, with high list usage and balanced headings.
Perplexity shifts towards brevity with an average of 288 words, favoring short summaries akin to executive briefs.
These differences define unique retrieval and reasoning methods, shaping how each platform interprets brands, categories, and commercial intent.
The data presents a clear direction: AI search is forging its ecosystem, driven by familiar SEO inputs, source quality, content structure, and off-page signals, all interpreted to deliver precise answers.
If your content isn’t well-structured and present across the web, it risks becoming invisible to AI platforms delivering answers or product suggestions.
As this new environment evolves, it’s crucial for retailers and brands to rethink their communication strategies across the entire digital landscape.
On-page actions that matter:
Develop semantically coherent homepages that convey the brand, product categories, and relevance to core queries. LLMs prioritize clarity over cleverness.
Strengthen product pages with factual content, clear specifications, and Q&A sections aligned with user research intents.
Establish educational content clusters tied to core product themes, serving as reusable frameworks for AI models.
Off-page actions that matter:
Foster comprehensive review ecosystems and discussion forums to validate trust signals LLMs recognize with product quality.
Ensure visibility in media driven by comparisons and recommendations. Regularly appear in “best of” lists, product roundups, and influencer content.
Invest in rich media showcasing product value, particularly on YouTube and TikTok. Video content helps train LLMs on product use cases, reflecting sentiment, and experiential value.
Maintain accurate, indexable product data in marketplaces like Amazon, Walmart, and Etsy to enhance AI discovery pathways.
OpenAI’s Shopping Research announcement escalates the stakes. With ChatGPT, OpenAI tracks real-time consumer research behavior, turning preferences into a user-trained targeting engine for commerce.
This isn’t just AI learning about your product. It’s AI absorbing consumer shopping behavior, revolutionizing discovery through an active AI participation model.
Brands not infused into these AI systems risk invisibility during AI-driven consumer journeys.
What Black Friday revealed was more than top-selling products; it showed how LLMs operate under real demand, revealing their reasoning, referencing, and prioritizing patterns.
The advent of AI-native visibility requires structured, semantically rich content, adequately represented across the right off-page ecosystems to align with major AI models’ reasoning.
Black Friday might be the stress test, but the real transformation is only just beginning.
See how collaborating with LLMs can transform your content by converting customer, expert, and competitor data into actionable insights.
When I think about large language models (LLMs), one major discussion point is their ability to scale content creation. It’s a tool we’re all tempted to lean on heavily. However, balancing efficiency with creativity is key.
With our busy schedules, boosting productivity is essential. Imagine using tools like Claude and ChatGPT not just for speeding up processes, but also for adding a personal touch to your website and making your day-to-day tasks easier, all without sacrificing creativity.
This journey explores how to:
Analyze customer feedback and questions comprehensively.
Streamline the gathering of detailed insights from subject matter experts.
Conduct competitive analysis effectively.
These tasks, often done manually, can be remarkably enhanced with automation, giving you an edge by rooting your approach in customer and market realities instead of working in a vacuum.
By tapping into this information, I can better connect with my audience, avoiding the pitfalls of an echo chamber.
Analyzing Customer Feedback at Scale
One outstanding feature of LLMs is their scalability in processing data, identifying patterns, and uncovering trends—tasks that might otherwise take me or a colleague days or even weeks to complete.
If you’re not part of a global enterprise with a dedicated data team, LLMs are your next best ally to substitute those capabilities. Focusing on customer feedback, for instance, could mean the difference between success and redundancy. The thought of sifting through thousands of NPS surveys doesn’t sound appealing to me, and I doubt it does to you either.
Utilizing raw data uploads into a project knowledge space and having my LLM of choice run its analysis is one way to go. However, I prefer uploading this data into something like BigQuery, using LLMs to write relevant SQL queries for in-depth analysis, ensuring integrity and accuracy.
This approach not only lets me peek behind the analytical curtain, learning SQL by osmosis but also serves as a safeguard against potential inaccuracies or hallucinations often seen with direct LLM data uploads.
The separate handling of data fosters a more reliable, accurate, and actionable insight, preventing the wild goose chases that could arise from misleading automated responses.
Practically speaking, unless overwhelmed by enormous datasets, BigQuery is a free resource (setup might require a credit card, though). And fear not if SQL is new to you; with an LLM, you’re set for success with full query support in place.
Here’s a glimpse into my workflow:
Generate SQL functions using the LLM.
Debug and validate data entries.
Feed LLM with results from SQL queries.
Create visualizations either with the LLM or via further SQL queries.
Frustrations abound when attempting to secure time with subject matter experts, whose schedules often leave them stretched thin.
Why would they want to regurgitate information they’ve already discussed ad nauseam with the manufacturing team? Yet, for marketing purposes, I still need this information to clearly present new features on our platform, offering customers precise details beyond mere specifications.
How to get this coveted expertise? By crafting a customized GPT that can assume the role of interviewer, asking the right questions.
Be advised: customization may vary depending on the launch, product, or service in question. A ChatGPT Plus subscription should suffice for this task.
The guidelines should entail the following:
Role and tone: Define the interviewer’s persona.
Context: Clarify learning objectives and rationale.
Interview structure: Outline initial topics and follow-ups.
Pacing: Implement a structure of query-response dynamics.
Closing: Craft a concluding summary or call to action.
Testing it myself, I pretended to be a subject matter expert to refine this tool, always seeking to fit within their limited downtime.
The responses provided can then be further analyzed or converted into draft articles thanks to an LLM.
While potentially tricky, the strategic examination of competitors can yield profound insights regarding the competitive landscape and personal business gaps.
Here’s a few things I’ve found valuable when dissecting competitor data:
Aggregating competitors’ reviews helps identify common themes, benefits, and problem areas.
An analysis of their web copy gives clues into the type of audience they’re targeting and their unique positioning. Combine this with the Wayback Machine to track how messages have evolved over time.
Job postings can highlight strategic priorities or areas of potential experimentation.
Social media engagement data can provide insight into customer satisfaction and desire, revealing potential gaps in their customer service.
Using LLMs alongside extensive datasets allows me to remain grounded in customer realities while being swift in delivering specific, actionable insights through pair programming.
The methods explored within are just starting points. Consider other useful data sources you might already have access to:
Call transcripts from sales teams.
Query data from Google Search Console.
Insights from on-site searches.
Heatmaps tracking user interactions.
A note of caution—while analytics data is tempting, sticking to qualitative, customer-focused data rather than quantitative metrics leads to richer insights.
I’m always fascinated by how technology evolves, especially when it comes to AI models. Recently, I stumbled upon some compelling data showing how these AI systems are reshaping brand hierarchies and influencing buyer decisions at an unprecedented speed.
AI models like ChatGPT, Gemini, and Claude have become a part of our daily interactions, from search to content creation and product recommendations.
According to a survey conducted by Responsive, a significant 80% of tech buyers now use generative AI to research vendors just as often as they use traditional search methods. This shift in how buyers build trust with AI-driven discovery tools quietly determines which brands stay top-of-mind and which fade into oblivion.
At Previsible, we’ve been analyzing this intriguing phenomenon through what we call LLM perception drift. It’s a new metric revealing how AI models are dynamically organizing brands within specific categories over time. (Disclosure: I am the CEO and co-founder of Previsible.)
Our case study on project management software, comparing data from September to October 2025, highlights just how quickly AI brand perception can change. This volatility is set to become the next major metric for SEO strategies.
Key insights
The concept of LLM perception drift is emerging as a crucial visibility metric in SEO and B2B marketing.
Brands like Atlassian gained prominence, while others like Trello and Slack saw declines, indicating the dynamic nature of AI perception.
Understanding AI brand perception is pivotal for marketers aiming to grasp authority and relevance in language models.
By 2026, maintaining digital visibility will hinge on AI brand signal stability as LLMs rapidly evolve.
A subtle shake-up inside the AI mind
Evertune’s AI brand score provides insights into how likely a model is to recommend a brand without specific prompting. It measures both visibility and ranking within AI responses.
September to October shifts highlight considerable changes in the internal brand landscape of AI models. Notably, Slack saw a significant decline, while Atlassian experienced a boost.
This seemingly simple reshuffle reveals a deeper transformation in AI’s nonspecific brand awareness, altering how the model discerns and prioritizes brands despite market stability.
The meaning behind the drift
We’re seeing two main forces driving these shifts:
Category entanglement
Rather than declining, categories are blurring — project management tools are being integrated into broader conceptual frameworks.
Operations
Digital transformation
Workflow orchestration
Enterprise productivity
IT consulting
Names like Deloitte and KPMG rise alongside Smartsheet and Atlassian.
Ecosystem advantage
Brands with multi-product ecosystems are getting noticed more. Atlassian’s lift, for example, stems from its robust documentation and integration abilities. Brands like Microsoft, Google, and Amazon are also seeing positive movement.
Models increasingly prefer brands that span multiple ecosystems, echoing entity-based SEO patterns but at a faster, more volatile pace.
We observe emerging trends in newer brands like Celoxis and Workfront, showcasing how fine-tuned LLMs draw from diverse datasets.
SaaS directories
GitHub repositories
Technical documentation
Reviews
Community content
For smaller B2B brands, this represents a gateway to visibility without needing to dominate traditional SEO metrics.
Why this shift matters for B2B discovery – and why it’s speeding up
Traditional SEO focuses on visible search results, whereas LLMs synthesize knowledge based on associations and contextual richness.
This means that brand recall in AI systems relies on deeper semantic connections, and these can fluctuate significantly over short periods.
Understanding and leveraging this LLM perception drift is crucial — being consistently recognized in AI outputs is now as vital as traditional search rank.
A new AI optimization KPI: AI brand signal stability
In working with B2B clients, we’re focusing on AI brand signal stability as an emerging metric — tracking how consistently a brand’s presence is maintained in AI outputs.
Fluctuations suggest fragile brand perception, influenced by data changes and model retraining, while stable scores indicate strong semantic grounding.
In coming years, AI brand signal stability will be essential alongside share of voice and traditional SEO metrics.
From project management to every B2B vertical
This transformation isn’t limited to project management — it’s happening across all B2B sectors.
The recalibration of category contexts by AI models alters the buying journey, influencing brand appearance in AI-generated content.
The rise or fall of brand attention affects which brands occupy summative or comparative outputs, making AI memory a new realm of marketing focus.
This shift marks SEO’s evolution — from focusing on search indices to emphasizing model memory optimization. Our goals now include measuring how AI interprets and recalls brand identity.
It’s about ensuring that AI systems correctly interpret and represent brands across their expansive digital landscapes.
This demands new strategies and tools tailored to how dynamic perception systems function, rather than treating them as static outcomes.
Evertune’s dataset highlights more than monthly position changes — it showcases a quick shift in AI’s category perception, which marketing teams must monitor to stay competitive.
By 2026, brand appearance in AI-generated summaries will play a bigger role in decision-making than traditional metrics like pageviews or clicks. Brands that effectively manage their model-driven visibility will set themselves apart as AI becomes a mainstay in digital research.
