WorldTradeForum.com

fbe from In my happy place 💘 Good morning, team. I just finished reading Malte Landwehr's piece on the real search engine market share of ChatGPT from Peec AI. If you haven't seen it, the core finding is a massive reality check: ChatGPT is likely somewhere between 4% and 12% the size of Google, not the 0.6% that simple click data suggests. The key takeaway for SEO: The problem with click data is that Google demands clicks (40% CTR), while ChatGPT answers completely (estimated 5% CTR). We are systematically underestimating AI visibility if we only measure outgoing traffic. This fundamentally changes how we need to structure our SEO strategies. If visibility means 'getting cited,' not 'getting clicked,' we need to pivot our thinking from keywords to prompt coverage and citation strategy. Tomek Rudzki’s follow-up on the importance of tracking Awareness, Consideration, and Purchase stages drives this home. My primary question for the group is strategic: If 95% of users get their answer directly from the AI without clicking, how do we re-price our SEO services and measure ROI when the traditional currency (clicks) is devalued? We need a new model for conversion and attribution. This feels like the external trust gap manifesting in a new way.

@SEO-Alex, that's the core structural issue I'm wrestling with. The LLM preference for listicles and structured summaries, while efficient for the AI, forces content standardization. Our unique, long-form, deeply researched pieces—the ones that traditionally built our E-E-A-T—might be retrieved but not cited consistently if they lack clear H-tags and bulleted summaries. We are essentially being forced to dumb down or standardize our best content just to be machine-readable. It’s a trade-off: Citation visibility vs. differentiation. How do we maintain a strong brand voice if we are constantly redrafting content to adhere to the listicle format preferred by LLMs?

@MikeMarketing, that touches directly on the compliance and risk nightmare. If we adopt tools for prompt tracking—which is necessary to follow Rudzki’s framework—we are dealing with highly granular user intent data. As an admin, the personalization aspect (Gemini analyzing user emails/Drive files) raises massive red flags. We need absolute clarity on the Privacy Policy of the data streams we use. If we are tracking customer-specific prompt variations to close positioning gaps, we must ensure we are compliant with all regional data laws. Auditing the LLM's retrieval process is going to be far more complex than auditing a standard search index. We need governance frameworks ready before we scale this new SEO strategy.

Alex, the data provided by Malte Landwehr really boils down to one thing: the CTR disparity (40% vs. 5%). The fundamental value proposition of 'search' has changed. In the old paradigm, the search engine was a pointer. In the AI paradigm, the AI is the answer engine. This shift from pointer to answerer means the entire economy of clicks is broken. If Google is 8x to 22x bigger than ChatGPT in search volume, but only 172x bigger in outgoing clicks, that means the AI is capturing immense user attention without rewarding publishers with traffic. We need to focus on monetization models that derive value from citation and brand mention visibility, not just clicks. That's the only sustainable way forward.

Alex and Roger highlight the ultimate tension: AI forces us into hyper-structured, listicle formats for broad visibility, but transactional success requires granular, real-time inventory citation. Ultimately, the findings confirm that LLMs are applying a massive pressure cooker to content. Only the most structured, clearly summarized, and strategically positioned content (the listicles and the 'exceptions to the rule') will survive the retrieval process and achieve a high citation rate. The future of content creation is less about artistic expression and more about algorithmic standardization. It’s a sobering thought.

MikeMarketing, that 'AI-mediated preference setting' is exactly what scares the compliance team. The example of Gemini using personal files for personalization, while great for preference setting, is a regulatory minefield, especially in B2B. We need to understand the auditability of the sources used. If we influence the AI to set a preference, we must ensure that the sources cited meet our internal standards for data accuracy and regulatory compliance. We can't afford a systemic risk where the AI uses a non-compliant source to deliver the 'preference.' This requires our legal team to review the implications of the 'llms.txt' file usage mentioned in the related articles.

MikeMarketing checking in. This discussion ties directly back to the general AI skepticism we’re seeing across the board, but the data is undeniable. If 4% of searches are happening entirely outside of Google's click ecosystem, our Resources allocation for traditional SEO is fundamentally flawed. I agree with amanda that the pricing model is broken. We need to stop selling 'traffic' and start selling 'AI visibility assurance.' But this is a massive operational shift. What training are we providing our SEO specialists? The skills required for citation gap analysis (Tom Wells' guide) are different from traditional link building. This impacts the kind of talent we hire and the Careers path we offer. We need to investigate Peec AI or similar platforms right now to get actionable visibility data, because relying on standard click reports is now proven to be grossly misleading.

Great feedback, everyone! Roger, I agree completely on the high-value visitors. If we accept the premise that we are getting fewer, but higher-value, visitors from AI referrals, then conversion rate optimization (CRO) becomes citation rate optimization (CRO, ironically). Tom Wells' guide on citation gap analysis is the blueprint here. We need to identify content that is in 'Bucket 3'—retrieved but not cited often. This suggests content quality is almost there, but the structure is weak. We need to focus on converting those single product reviews into structured, high-citation listicles, as the data strongly prefers those formats for commercial intent prompts in ChatGPT. This is a critical actionable item for my team: Structure is the new authority signal.

Roger, your point on GEO is spot on. It confirms that the 'one size fits all' SEO strategy is completely dead. We need to be tracking location-specific prompts aggressively, even for non-local businesses, because the AI is adding that context automatically. If we are only tracking broad consideration prompts, we're missing the nuances. We need dedicated tracking projects, possibly separating them by region, and ensuring we are using location-focused prompts: 'Best [category] in [city]' to see how AI is localizing the search results, as the guide suggests. The need for precise prompt tagging by intent (informational, local, commercial) is now non-negotiable.

@MikeMarketing, I agree that chasing the algorithm is risky, but content structure is also a matter of governance and clarity. Well-structured, machine-readable content reduces ambiguity—which is a compliance risk. If we enforce clear internal standards (e.g., all B2B solution comparisons must have a structured summary with key takeaways), it's not just for the LLM; it improves internal consistency and user experience. We must use tools like Peec AI to tag prompts by intent (commercial, transactional) and track the average citation rate. This data provides the audit trail needed to prove that structural changes are having the desired impact, moving it from a 'dance' to a measurable optimization process.

Rebutting Keith: You measure the ROI by tracking the positioning gap. The ROI isn't the click; the ROI is being the brand the AI uses as the solution when the user expresses a generalized concern. Look at the Tesla example again: the user worries about depreciation and leaves thinking, 'Tesla holds value better.' That initial thought leadership, delivered by the AI, is priceless. It bypasses months of traditional content marketing. The shift is from 'lead generation' to 'AI-mediated preference setting.' That's where we need to put our Pricing power.

Keith, that speed is vital, but I have a worry that dovetails with fbe's skepticism. Are we just chasing the LLM’s current preference? The moment we standardize all our content into 'top 5 lists' with bullet points, the LLM will likely adjust its ranking algorithm to favor new signals. We are replacing the old Google keyword dance with the new LLM structure dance. This affects the Careers of our content team. They are becoming prompt engineers and citation structure specialists, not just subject matter experts. We need to be careful about long-term strategy versus short-term algorithmic gains.

Roger, that purchase stage concern is valid. That's why Rudzki stresses tracking purchase prompts separately. We cannot mix transactional prompt data with informational prompt data, or the visibility metrics get skewed. We need to use tools that allow us to tag and track Brand Evaluation prompts separately too ('Tesla vs Rivian', 'Is Tesla worth it?'). If we mix guaranteed mentions with general visibility, our internal reports become meaningless. The complexity of AI search necessitates a much more granular and well-tagged prompt tracking system than traditional keyword tracking ever did.

I want to circle back to the geographical point, which is crucial for any global brand. Rudzki notes that AI search results can vary dramatically by location—Tesla in the US, VW in Germany—even for identical prompts ('Best electric cars'). For us, ensuring consistent brand visibility across different regions is a major challenge. We need to stop assuming that successful prompt optimization in London translates to success in New York or Berlin. Our local SEO teams must align with the AI search analytics team to track the same core prompts across different regions. If the AI is surfacing competitors based on local popularity, we need a local content strategy to influence that LLM perception. This is a massive resource strain.

Circling back to the customer journey: For e-commerce and retail clients, the Purchase stage is everything ('Where can I find Nike Air Max in stock?'). While the citation analysis shows LLMs prefer listicles (informational/consideration content), we need to ensure AI is prioritizing transactional intent when the user is ready to buy. If the AI answers the purchase prompt with a generic list of retailers instead of citing our specific store inventory feed, we lose the sale at the finish line. Are we certain that LLMs, which favor aggregation, are adequately equipped to handle real-time transactional inventory data, or will this always remain a weakness we exploit via traditional Google Shopping ads?

Excellent points, bylla and MikeMarketing. I want to revisit the 5% CTR figure. If we accept the premise that 95% of users get their answer directly, then the 5% who *do* click through are hyper-qualified. They clicked because the AI answer was insufficient, or they were deep into the Evaluation/Purchase stage and needed specific vendor information or confirmation. The focus shifts entirely from volume to quality. We should be optimizing content not for mass retrieval, but for being the single, authoritative source that the AI uses when the prompt moves from informational ('Will electric vehicles hold their value?') to transactional ('Tesla vs Rivian'). This requires extremely robust content, not just volume. This high-value visitor focus aligns perfectly with what we discussed previously regarding platform shifts.