Your B2B brand probably ranks on page one for dozens of high-intent keywords. According to recent Search Engine Land analysis, that ranking now means less than it did six months ago: B2B brands appear in just 3% of AI Overviews, even when they hold top organic positions. The gap between "ranking" and "being recommended" is the story David Bell has been documenting all year, and it carries direct implications for how you model pipeline contribution from organic channels.
Bell, co-founder of Previsible and a featured contributor at Search Engine Land, has built one of the more rigorous data sets on AI-driven discovery. His work sits at the intersection of SEO measurement and revenue forecasting, which is exactly where marketing executives need clarity right now.
The 92% Number That Rewrites Channel Mix
Previsible's latest AI Traffic Study, released this week, analyzed 6.77 million AI-driven sessions across 166 websites spanning SaaS, e-commerce, finance, legal, health, insurance, education, and publishing. The headline finding: ChatGPT commands 92.4% of trackable standalone LLM referral traffic. That concentration matters for budget allocation. If you're running experiments on "AI visibility" without isolating ChatGPT as the primary surface, your test design is contaminated before it starts.
But the more operationally useful insight is what Bell calls the "Google-first" reality. AI discovery happening inside Google, through AI Overviews and AI Mode, represents more AI-influenced traffic than every standalone LLM assistant combined. The implication for your forecast model: don't treat "AI search" as a separate line item competing with organic. Treat it as a filter that sits upstream of your existing organic funnel, compressing the consideration phase and shifting where intent signals appear.
Where the Traffic Actually Lands
Bell's earlier 2025 State of AI Discovery Report tracked 1.96 million LLM sessions and found that AI traffic doesn't distribute evenly across your site. Industry pages showed 1.14% AI penetration, tools pages hit 0.95%, and pricing pages reached 0.46%, all significantly higher than the 0.13% site-wide average. That concentration pattern tells you something about buyer behavior: by the time someone clicks through from an LLM, the decision is largely made. They're not browsing. They're validating.
For pipeline forecasting, this changes how you weight page-level conversion rates. A pricing page visit from ChatGPT isn't equivalent to a pricing page visit from a branded search. The intent density is higher, but the volume is lower, and the attribution is messier because most AI-influenced sessions arrive without clean referrer data.
The 3% Visibility Gap
The Walker Sands B2B AI Search Visibility Benchmark quantified what Bell's reporting has been circling: AI Overviews appear in nearly 50% of search results pages where enterprise B2B brands rank, yet the median enterprise B2B brand is cited in just 3% of those AI Overviews. That's a 47-point gap between "showing up in organic" and "being recommended by the AI layer."
For a CMO building a board deck, this gap creates a measurement problem. Your organic traffic numbers may look stable while your actual influence on buyer decisions is eroding. The fix isn't to abandon organic investment. It's to add a new KPI: citation rate in AI-generated responses for your priority keywords. Bell's work at Search Engine Land has been pushing this measurement shift for months, and the tooling is finally catching up.
Content That Gets Cited vs. Content That Ranks
Bell's June 2026 piece on AI Overviews citing self-serving listicles surfaced an uncomfortable finding: Google's AI Overviews recommend competitors 69% of the time, even when citing your content. The model pulls facts from your page but routes the recommendation elsewhere. This isn't a bug in your content strategy; it's a structural feature of how retrieval-augmented generation works.
The operational response is to audit your content for "citability" rather than just "rankability." Does your page contain the specific, verifiable claims that an LLM would extract? Does it answer the question directly in the first 200 words, or does it bury the answer behind context-setting paragraphs? Bell's research on deep-research AI agents found that a 13-word edit can steer what these systems recommend. That's not hyperbole; it's a test design insight. Small, targeted changes to how you structure claims can shift citation behavior in ways that large-scale content production cannot.
The Forecast Adjustment
Here's the math that matters for your next planning cycle. Forrester estimates that AI-generated traffic is currently 2% to 6% of B2B organic traffic and growing at 40% or more month-over-month. At that growth rate, AI-influenced discovery will represent a material share of your top-of-funnel by Q4. But the traffic itself may not show up in your analytics as "AI referral" because Google doesn't separately attribute AI Mode and AI Overview clicks in Search Console.
Bell's contribution to the field is making this invisible influence visible. His data sets give you the benchmarks to model what percentage of your organic traffic is actually AI-mediated, even when the referrer string says "google.com." For a CFO reviewing your channel efficiency, that distinction matters. You're not losing organic performance; you're losing visibility into how organic performance is being shaped.
The Pilot You Should Run
If you're taking Bell's research seriously, here's a two-week test worth running. Pull your top 20 keywords by pipeline contribution. Check which ones trigger AI Overviews using a tool like Semrush or Ahrefs. For the keywords where AI Overviews appear, manually query ChatGPT and Perplexity with the same intent. Document whether your brand is cited, mentioned, or absent.
That audit gives you a baseline citation rate you can track over time. It also surfaces the specific pages where small structural edits might shift your visibility in AI-generated responses. Bell's work suggests the ROI on those edits is higher than the ROI on producing new content that ranks but doesn't get cited.
The brands that will win the second half of 2026 aren't the ones producing more content. They're the ones present on the decision-making surfaces LLMs actually rely on. Bell's data makes that case with the kind of specificity a board can act on.
