A mid-market CFO needs a new FP&A tool. She doesn’t open Google. She opens ChatGPT, types “what software do mid-size manufacturing companies use for financial planning,” reads the response, asks two follow-up questions, and builds a mental shortlist — before your site has registered a single session, before your SDR has any idea she exists.
This is the AI research phase in the B2B buyer journey. It’s new. It’s real. And most $5M–$50M companies have no instrumentation for it whatsoever.
What the AI Research Phase Actually Is
Traditional funnel thinking starts at awareness: a buyer searches, finds your content, enters your ecosystem. The AI research phase sits upstream of that. It’s a pre-awareness stage where a buyer uses an AI assistant like ChatGPT, Claude, or Gemini — or an answer engine like Perplexity — to orient themselves in a category before they start vendor-specific research.
They’re not asking “should I buy X?” They’re asking “how do companies like mine handle X?” and “what are the main approaches?” and “what should I ask vendors?” The AI answers. It may or may not name your company. By the time they reach Google, they already have a frame. And if you’re not part of that frame, you’re competing against a shortlist you were never invited onto.
To be clear about scale: this is not yet the dominant discovery channel. It’s a fraction of overall B2B discovery today — established B2B buyer research still shows search engines, vendor websites, and peer recommendations as the channels most buyers lean on, with AI-assisted discovery a smaller and newer entrant. But “small share” and “doesn’t matter” are not the same thing. The buyers who start in AI tools tend to be the more self-directed, research-heavy buyers — often the ones with real budget authority and low tolerance for a standard sales sequence. They’re worth understanding.
Why Low Volume Doesn’t Mean Low Stakes
In my experience, the tendency in mid-market orgs is to wait for volume before acting. That instinct makes sense for most channel decisions. This one is different for two reasons.
First, the AI research phase compounds. The citations and references that AI models surface today are drawn from content that already exists — articles, analyst write-ups, review site profiles, LinkedIn content, forum discussions. The models don’t update in real time like a search index. If you’re not already visible in the sources they pull from, getting into that rotation takes months of consistent, substantive content. Waiting until the volume is undeniable means starting your content build when the window is already half-closed.
Second, the cost of getting ahead of this is low. This is not a major investment category. It overlaps almost entirely with things a well-run marketing org should be doing anyway: clear positioning, substantive thought leadership, a strong presence on G2 or Capterra, and content that actually explains your category rather than just pitching your product. The incremental ask, if you’re already doing those things, is small.
How AI Models Decide What to Surface
The short version: AI models don’t crawl the web fresh for every query — they lean on training data and, increasingly, live retrieval, and they favor content that’s specific, well-structured, and already referenced by other credible sources. A thin “about us” page doesn’t help; a substantive category explainer, a detailed G2 profile with real customer language, and a consistent LinkedIn presence do. Write to be useful to a research-minded buyer, and you’re writing to be useful to the model answering that buyer.
The deeper mechanics — how an assistant actually chooses which vendors to name, and what makes a specific page citable — are their own topic, covered in How AI Assistants Decide Which B2B Vendors to Recommend.
What a $5M–$50M Company Actually Does About This
No new technology stack required. No major budget reallocation. Here’s where to put energy:
Audit your category content. Search your primary category terms in ChatGPT and Perplexity right now. See what comes up. See who’s named. Read the framing those tools use to explain your category. If your company isn’t mentioned and your competitors are, you now have a concrete content gap to fill — not a vague one.
Write explainer content, not promotional content. The content that gets surfaced in AI research responses tends to be educational. “How does [category] work?” “What are the tradeoffs between approach A and approach B?” “What does implementation actually look like?” These aren’t glamorous topics. They’re the ones buyers are actually typing into a chat window at 8pm before a vendor meeting.
Make your review site presence accurate and specific. G2, Capterra, and TrustRadius profiles are increasingly included in retrieval-augmented AI responses. A thin or outdated profile is a missed signal. Get current customers to leave reviews that use specific, category-relevant language — not just star ratings.
Build your LinkedIn content around category education. If a buyer’s AI-assisted research surfaces a founder’s LinkedIn post explaining a real problem in the category, that’s a trust signal that carries forward into the sales conversation. Consistency matters more than any individual post.
Don’t stop measuring what you already measure. The honest answer is that AI-driven discovery is hard to attribute today. A buyer who started in ChatGPT and arrived at your site through a Google search looks like organic search in your analytics. Don’t invent a fake measurement framework to compensate. Do ask “how did you first start researching this?” in your discovery calls. The qualitative signal is real, even if the quantitative picture is incomplete.
If your revenue engine still runs mostly on relationships and referrals, that’s not a weakness — it’s proof that you’ve built something buyers value. The question is whether the next layer of growth can run on that alone. The AI research phase is one early signal that discovery is getting more fragmented, more self-directed, and harder to influence at the bottom of the funnel when you haven’t shown up at the top. Getting into position now is a measured, low-cost decision. The foundation it requires — clear positioning, substantive content, a credible external presence — is the same foundation a scalable revenue engine needs regardless of where buyers start their research.