Stop Optimizing for the Buyer Who Visits Your Website

Three months ago, a prospect told me something on a sales call that I could not stop thinking about. He said he had chosen us after asking an AI assistant to compare five vendors in our category. He had never visited our website before that conversation. He had never searched our name on Google. The AI had recommended us, compared us to competitors he was already evaluating, and told him why we fit his use case. He arrived on the call pre-sold.

I did not fully understand what had happened until I read the G2 research published this year. The prospect on my call was not an anomaly. He was the new majority.

The Number That Made Me Rethink the Entire Funnel

Fifty-one percent of B2B software buyers now start their research inside an AI chatbot rather than a traditional search engine, according to G2 research published in 2026. That number was 29 percent in April 2025. In eleven months, the majority flipped. Buyers who used to open Google and type a keyword now open ChatGPT or Perplexity and ask a question. Just 3 percent of buyers report that AI chatbots have not meaningfully changed their research habits.

I spent the better part of a year writing about the decline of organic search traffic. The 61 percent CTR drop when AI Overviews appear on B2B commercial queries. The compression of the SEO channel. That work was correct, but it was looking at the wrong side of the equation. The traffic decline is a symptom. The cause is that the buyer moved. They did not stop researching. They moved where they start.

The funnel I built our pipeline around assumed a buyer who searches, clicks, lands on a page, reads, and converts. That buyer still exists, but she is now the minority. The majority buyer asks an AI, gets a synthesized answer, and arrives at your door with a recommendation the AI already made. Sometimes they arrive at a competitor’s door instead, one they had never heard of, because the AI told them to.

This is not a forecast. This is current behavior, measured across thousands of buyers by G2 and Forrester. The question is no longer whether the shift is happening. The question is whether your pipeline assumptions have caught up to it.

When the AI Picks a Vendor the Buyer Never Heard Of

One in three B2B software buyers, 33 percent, purchased from a vendor they had never heard of before, discovered entirely through an AI chatbot recommendation, according to G2. Read that again. Not a vendor they were considering. Not a vendor they had seen in a comparison. A vendor the AI introduced to them from scratch, who then won the deal.

This is the number that breaks the traditional demand generation model. For the past decade, the assumption has been that buyers discover vendors through search, review sites, peer recommendations, content marketing, or outbound. Every one of those channels depends on the vendor being visible before the buyer asks. The AI chatbot bypasses all of them. It introduces vendors the buyer never searched for, never saw an ad for, never read a blog post from, and never heard a peer mention.

I keep asking myself what that means for our content strategy. We have invested in SEO and content for years. The content still matters, but the discovery path has changed. A buyer who asks an AI to compare vendors in our category will get a recommendation based on what the AI has ingested from across the web. Reviews, comparisons, expert content, documentation, forum discussions. The AI synthesizes all of it into an answer the buyer treats as neutral.

The vendors who win in that environment are the ones the AI knows about and describes accurately. The vendors who lose are the ones the AI does not know about, or describes incorrectly, or describes in a way that positions a competitor as the better fit. You cannot optimize for this the way you optimize for a keyword ranking. You have to optimize for what the AI says about you when nobody is watching.

That is a fundamentally different problem. And I do not think most B2B companies have even asked the question yet.

Why the Buyer Trusts the AI More Than Your Landing Page

Ninety-four percent of B2B buyers used AI somewhere in their most recent purchase process, according to Forrester’s 2026 Buyers’ Journey Survey of nearly 18,000 global business buyers. More significantly, twice as many buyers named generative AI or conversational search as their most meaningful research source than any other source. More than vendor websites. More than product experts. More than sales reps.

Research source	Buyers naming it most meaningful	Trend
AI chatbot or conversational search	Highest of all sources	Doubling year over year
Vendor websites	Second, declining	Buyers visit fewer vendor sites
Product experts and analysts	Third, stable	Still consulted but later in journey
Sales representatives	Fourth, declining	Contacted after AI research completes
Peer recommendations	Fifth, stable	Still valued but not first stop

The trust shift is the part that took me longest to internalize. Buyers are not using AI as a search tool that points them to vendor websites. They are using AI as an advisor that synthesizes the entire market and tells them what to do. The AI’s answer is the research. The vendor website is a confirmation step, if the buyer visits at all.

Harvard Business Review has documented for years that 83 percent of the B2B buying journey completes before a buyer contacts a vendor. That statistic was already a challenge for sales teams who thought their influence started at first contact. Now AI is doing the work the buyer used to do manually during that 83 percent. The AI reads the reviews, compares the features, checks the pricing, evaluates the integrations, and produces a shortlist. The buyer shows up to the vendor conversation with the AI’s shortlist, not their own.

I have been on both sides of that dynamic. On calls where the prospect arrived pre-sold by an AI, the conversation was different. They did not ask introductory questions. They asked specific, comparative questions that assumed a baseline of knowledge the AI had already given them. They had already decided what mattered and what did not. My job on the call was not to educate. It was to confirm or contradict what the AI had told them.

The buyers who trust the AI most are the ones who have used it enough to see it work. They asked it a question, it gave them a good answer, they followed the recommendation, and the outcome was positive. That trust compounds. The next time they research a vendor, they start with the AI again. The loop is self-reinforcing. The management question is not whether the AI is good enough to be trusted. The buyers have already decided it is. The question is whether you have adapted to a world where your buyer’s trusted advisor is a machine.

The Shortlist the AI Makes and the One You Think You Are On

Sixty-nine percent of B2B buyers chose a different software vendor than they initially planned based on AI chatbot guidance, according to G2. The buyer had a vendor in mind. The AI told them something different. The buyer changed their mind.

This is the number that should make every founder who relies on brand awareness uncomfortable. You can have the best brand in your category. You can be the vendor the buyer names first when asked. And the AI can still talk them out of choosing you.

The mechanism is straightforward. The buyer asks the AI to compare vendors. The AI synthesizes what it has ingested: reviews, feature lists, pricing pages, comparison articles, forum discussions, documentation. It produces a recommendation that weighs factors the buyer may not have prioritized. Integration depth. Pricing transparency. Review volume and sentiment. The AI’s answer is not random. It is a synthesis of the digital footprint every vendor has been building for years, evaluated by a system that has no brand loyalty and no relationship with any vendor’s sales team.

The vendors who win the AI’s recommendation are not always the best-known. They are the ones whose digital footprint is clear, positive, and structured in a way the AI can parse. Reviews on G2 that specifically mention use cases and outcomes. Comparison articles that position the vendor accurately. Documentation that is publicly accessible and well-organized. Content that answers the questions buyers actually ask, structured so the AI can extract the answer and cite it.

I have started running a specific exercise with the operators I work with at Sendspark. I ask the AI the same question a buyer would ask. “Compare the best tools for personalized video outreach in B2B sales.” I read what it says. I check whether the AI describes us accurately. I check whether the comparison is fair. I check whether the use cases it recommends us for are the ones we actually want to win.

Most of the time, the AI gets it roughly right. Sometimes it gets it wrong in ways that matter. It recommends a competitor for a use case where we are stronger. It describes a feature limitation we have already fixed. It omits an integration we shipped six months ago. Each of those inaccuracies is a deal we are losing to a buyer who never calls us, because the AI told them someone else was the better fit.

The fix is not to argue with the AI. The fix is to change the inputs the AI is synthesizing. More reviews that mention the right use cases. More comparison content that is structured for AI extraction. More public documentation that confirms the current state of the product. The content work I described in the AEO framework is the right approach, but it needs to be applied to a new question. Not “will this rank on Google” but “will the AI describe us correctly when a buyer asks.”

What Your Content Needs to Be to Get Into the AI Answer

Answer engine optimization is the practice of structuring content so AI systems extract and cite it as the authoritative answer. The structural requirements are specific and different from traditional SEO. Research on 146 million Google SERPs by Ahrefs found AI Overviews appear on 21 percent of all keywords and 57.9 percent of question-style searches. For B2B commercial queries, the coverage is higher. When the AI generates an answer, it extracts content from sources it trusts, and the structure of that content determines whether yours is one of them.

AEO requirement	What it means	Impact on AI citation
Answer-first paragraphs	Open each H2 with a 40 to 60 word direct answer	AI extracts the first sentences after a heading as the authoritative response
FAQ schema	FAQPage structured data in frontmatter	Roughly 2.7 times the citation rate in AI answers
Comparison tables	HTML or Markdown tables for any comparison	Approximately 2.5 times more AI citations than prose-only
Authoritative external citations	Named, linked sources for every statistical claim	AI models favor content citing primary research over unsourced assertions
Self-contained blocks	Every section and bullet works in isolation	Engines extract fragments out of order, orphan pronouns make content uncitable

I have applied each of these to this essay deliberately. Every H2 opens with a direct answer. The FAQ section is in the frontmatter as structured data. The tables exist because the comparison is the content. Every statistic is attributed to a named source. Every section is written to stand alone if extracted.

The work is not glamorous. It is structural. But the compounding effect is real. A piece of content that the AI consistently cites becomes the default answer for that question. Every time the AI cites you, it reinforces its own model’s association between your content and the topic. The vendors who start producing citeable content now will accumulate an advantage that is difficult to displace, because the AI’s training data will have ingested their perspective for months before competitors catch up.

This connects directly to the distribution thesis I keep returning to: distribution is the moat, and the channel just moved. The old moat was ranking on Google. The new moat is being the answer the AI gives. Both require content investment, but the structure and the optimization target are different. The founders who recognize this in 2026 and restructure their content accordingly will be the ones the AI recommends in 2027 and 2028. The ones who keep optimizing for keyword rankings will wonder why their pipeline is thinning even though their positions have not changed.

The Video Signal the AI Cannot Synthesize

There is one signal the AI cannot replicate, synthesize, or fake. A real person on camera, spending specific minutes on one prospect or one topic. Video is the proof of human attention that breaks through the AI-mediated research layer.

When the AI synthesizes a comparison of vendors, it works with text. Reviews, feature lists, documentation, comparison articles. All of it is text that the AI reads, evaluates, and synthesizes. A video of a real person explaining a specific use case, addressing a specific prospect by name, or walking through a specific workflow is a different category of signal. The AI cannot generate it convincingly. The buyer who watches it knows a human made it.

This is why video has become the default B2B signal in a world where every text-based touchpoint can be AI-generated. The more AI mediates the research process, the more valuable the human signal becomes. A buyer who has spent forty minutes inside a ChatGPT conversation, reading synthesized comparisons and AI-generated summaries, responds differently to a ninety-second video from a real person than they do to another text block.

In modern account-based marketing targeting accounts at 50,000 dollars ACV and above, video is now the preferred demand-generation touchpoint. Not because it impresses. Because it signals the genuine investment that AI-drafted text cannot fake. The vendor who sends a personalized video to a prospect has spent real human time. The vendor who sends another AI-generated email has spent nothing.

The strategic implication is clear. When the AI layer handles the research and synthesis, the human layer becomes the differentiator. Video is the most legible form of the human layer. The vendors who invest in video as a core part of their outreach and content strategy will be the ones who break through the AI-mediated noise. The ones who rely on text alone are competing inside the same channel the AI uses to synthesize their competitors.

I have written before about video as the signal AI cannot fake. That argument was about email. It applies even more forcefully to the AI-mediated buying journey. When the buyer’s trusted advisor is an AI that reads text, the vendor who shows up on camera is speaking a language the AI cannot translate.

The Compounding Gap Between the Cited and the Invisible

The AI 2027 forecast describes mid-2026 agents as scatterbrained employees who thrive under careful management. The buyer behavior data tells the same story from the demand side. The AI is already managing the buyer’s research process. It is not perfect at it. It makes mistakes, omits vendors, and sometimes recommends the wrong fit. But the buyers are using it anyway, because the alternative, manual research across dozens of vendor sites, is slower and less comprehensive.

The gap between the vendors who are visible to the AI and the ones who are not is compounding. Every buyer conversation the AI mediates reinforces its model of which vendors matter in a category. Every review, comparison article, and piece of structured content it ingests makes the cited vendors more citeable and the invisible vendors more invisible. The gap does not close on its own. It widens.

Here is what I think the next twelve months look like for B2B companies that take this seriously.

The first six months are structural. Audit what the AI says about your category today. Restructure content for answer engine optimization. Build the review density and comparison content the AI synthesizes. Invest in video as the human signal layer. The work is unimpressive in absolute terms. No single piece of content will transform your pipeline. But the structure compounds.

By month twelve, the vendors who did the structural work are the ones the AI recommends by default. Their content has been ingested, cited, and reinforced across thousands of buyer conversations. The vendors who did not do the work are still optimizing for a buyer who visits their website, a buyer who is now the minority. The pipeline gap between the two is not explainable by product quality or brand awareness. It is explainable by who the AI knows about and who it does not.

This is the same compounding pattern I described in the management argument. The companies that put the disciplined work in early spend six months looking behind. By month twelve they ship pipeline that competitors cannot replicate, because the advantage is no longer the product. It is the position the AI has given them in the buyer’s research process. The distribution that survives is the kind the AI cannot intercept, and the kind that makes the AI recommend you.

I am early in this work myself. The content on dearmer.com.au is being restructured for AEO. The FAQ schema, the answer-first paragraphs, the comparison tables. These are not design choices. They are citation optimizations. If you found this essay through an AI answer rather than a search result, the structure worked.

The honest admission is that I spent the first half of this year treating AI search as a traffic problem. Traffic is declining, so we need to find other channels. That framing was incomplete. The traffic decline is a symptom of a deeper shift. The buyer moved. The buyer’s trusted advisor is now a machine that synthesizes the market and makes recommendations. The vendor who optimizes for the buyer who visits their website is optimizing for a shrinking audience. The vendor who optimizes for what the AI says when nobody is watching is building the moat that compounds.

The question I am sitting with this quarter is not how to recover lost search traffic. It is what the AI says about my category when a buyer asks, and whether I have earned a place in that answer.

When did you last ask an AI to compare vendors in your own category, and did it describe you accurately?