Most B2B SEO advice still runs on rankings and backlinks, because those are the things a report can show you. Position, referring domains, organic sessions: all real, all measurable, and all blind to the thing that is actually costing B2B companies deals right now. A growing share of B2B buyers get their answer from an AI assistant before they ever reach a page you could rank on, and there is no line in Google Search Console, Ahrefs or Semrush that tells you when that happens.
I saw exactly how blind those reports can be while running an AI visibility audit for a client, a technical consultancy whose buyers vet a specialist carefully before they ever pick up the phone. Nobody wants to be the one who hired the wrong expert, so they ask around first, and increasingly “asking around” means opening ChatGPT.
The result was not close, and it is the reason this post exists.
What the audit found
I wrote eight questions this client’s actual buyers would ask when vetting a specialist in their field, the kind of questions a real prospect types in before ever contacting a company directly. I ran each one through ChatGPT, Perplexity, Gemini and Google’s AI Overviews, a fresh session every time, and I never introduced the client’s name first. A model will discuss a company generously once you name it for them. The point of an audit like this is to see who comes up unprompted.
Across those eight questions and four engines, 32 runs total, the client was recommended zero times.
That alone would be worth knowing. What made it more useful is what happened on the runs where the engine clearly knew the client existed. On several of them, the assistant searched the client by name, pulled up their history, weighed them against a competitor by name, and still left them out of its final answer. That rules out the easy explanation. This was not an obscure company an AI model had never heard of. It was a company the model could describe accurately and chose not to recommend anyway.
I measured this in June 2026. AI answers are not fixed the way a page-one ranking can feel fixed. Models retrain, providers tune their answer logic, and the same eight questions could return a different result in three months. This is a snapshot of what four engines said that month, not a permanent verdict on the company. But a snapshot is still a real measurement, and this one showed a specialist firm being actively passed over by the tools its own buyers were starting to use for exactly this kind of decision.
To be clear about what this is: it is a diagnosis from an audit, not a result I delivered. I did not fix this for the client. I found it. What they do with it is a separate conversation, and as of this writing it is still an open one.
If you want the process itself, methodology, prompts, engines, and how I score the answers, I run this as a standalone AI visibility audit separate from ongoing SEO work.
Why a rankings report cannot see this
Google’s own numbers explain why this gap is widening rather than staying a curiosity. At Google I/O 2026, Sundar Pichai said AI Overviews now has over 2.5 billion monthly active users, and AI Mode passed a billion monthly users in its first year. That is not a niche channel for early adopters. Billions of people now routinely see an AI-generated answer above, or instead of, a list of blue links.
The click-through consequence is measurable too. Ahrefs’ own research, published in 2025, found that AI Overviews can cut click-through to the top organic result by roughly a third, with informational queries hit hardest. A B2B buyer researching “how to evaluate a [specialty] vendor” is exactly that kind of informational query. The searcher reads the AI-generated summary, gets an answer, and never scrolls to the ten blue links your SEO report has been tracking for years.
None of this shows up as a problem in a standard audit. Your rankings can hold. Your backlink profile can grow. Your organic sessions can even be flat instead of declining, because the query volume hasn’t dropped, only where the answer comes from. Every number a client-facing SEO report was built to show can look fine while the thing that actually decides whether a buyer calls you happens somewhere the report never looks.
What this changes in practice, and what it doesn’t
Keyword research, content depth, technical SEO, backlinks: none of that becomes optional. AI answer engines still lean heavily on the same trust signals search engines have rewarded for years, established sites, clear expertise, other credible sources referencing you. If you have not done the SEO fundamentals, you are not going to get named in an AI answer either. Think of this as an additional layer on top of the work you are already doing.
What changes is the finish line. The old version of B2B SEO success was “rank on page one for the terms our buyers search.” The 2026 version has to include “get named when our buyers ask an AI assistant instead of typing a query.” Those are related goals, but they are not the same goal, and you can hit the first one completely while missing the second.
In practice that means treating your buyers’ actual questions as content you need to own, alongside the search keywords you already target. An industrial coatings manufacturer’s buyer might type “best coating for saltwater exposure” into Google, but they might also just ask ChatGPT “which coatings companies specialize in marine environments” and expect a direct answer. If your content only exists to rank for the first phrasing, you have covered half the field.
It also means testing your own visibility the same way I tested the audit above. Pick the handful of questions your actual buyers ask before they call. Run each one through two or three assistants in a fresh session, without naming yourself first. See who shows up. If a competitor gets named and you don’t, that is worth more diagnostic weight than another month of rank tracking.
I go deeper on where AEO, GEO and SEO actually differ, and where that distinction is mostly semantic, in a separate post on AEO vs GEO vs SEO. This one stays with the strategy shift.
An honest example, using this exact page
I can make this concrete without leaving the room. This page, in its previous version, had zero impressions and zero clicks in Google Search Console over the trailing 365 days. “B2B SEO” as a target term is a reasonable guess based on the slug, but I have no measured demand behind it, and I’m not going to dress that up as a validated keyword strategy. If you’re reading this and you also run SEO reports for a living, you already know how often a keyword target gets chosen this way: a plausible guess standing in for real data, because real data was never pulled.
That is a smaller version of the same problem the audit surfaced. A page can exist, be well-intentioned, and still be functionally invisible, whether the cause is an untested keyword or an AI assistant that never mentions you. The fix in both cases starts with actually measuring the gap instead of assuming the absence of complaints means things are fine.
Where this leaves B2B SEO strategy
If you run B2B SEO and your reporting stops at rankings, backlinks and sessions, you are reporting on a shrinking share of how your buyers actually reach an answer. That doesn’t mean stop reporting on those things. It means add the one metric no platform hands you automatically: whether the AI tools your buyers are already using would name you if asked, unprompted, today.
You can see examples of how this plays out across different accounts on our results page, or if you want someone to run the eight-question test against your own buyers’ actual language, that’s a conversation worth having before your next rankings report tells you everything looks fine.
Being ranked and being recommended used to be close enough to the same thing that nobody needed to separate them. They aren’t anymore, and the gap between them doesn’t announce itself. It just quietly decides who gets the call.
If you want to know where you actually stand, get in touch and we can run the same test on your buyers’ real questions.