B2B Buyers Now Use AI Chatbots More Often Than Google

68% of U.S. Google searches ended without a click from January through April 2026, while less Google traffic reaches the open web.

A 2026 State of B2B AI Discovery study with AirOps tracked roughly 380,000 AI answers across 81 enterprise-software topics on six surfaces: Google AI Overviews, AI Mode, ChatGPT, Claude, Gemini and Perplexity. Product pages led citations, but trusted sources differed.

What does this change for B2B software teams?

You need visibility on two networks. Google’s AI features draw from live Search; ChatGPT, Claude, Gemini and Perplexity run their own retrieval. A brand can rank in Google and vanish in ChatGPT. A blended score hides that split.

Vendor, integrator and directory citations ran from 52% on Perplexity to 72% on Gemini; reviews added 9% to 21%. Google surfaces used 7% social citations; ChatGPT, Gemini and Claude, 0.3% or less.

• Reddit: AI Mode 52%, AI Overviews 39%, ChatGPT 21%, Claude none.
• YouTube: AI Mode 47%, AI Overviews 38%; essentially zero for ChatGPT and Claude.
• LinkedIn: AI Mode 13%, AI Overviews 9%, Perplexity 6%; none from ChatGPT, Gemini or Claude.
• G2: 23% to 30% on five of six surfaces.

Your pages remain the base, but they compete with rivals and directories. Ranking no longer guarantees a citation; AI Overviews draw fewer citations from top-ranking pages.

Google handles roughly three-quarters of desktop searches. AI Overviews reach 2.5 billion monthly users; another billion use AI Mode. Yet G2 says 51% of B2B software buyers now start research in an AI chatbot more often than Google, up from 29% a year earlier.

Build the stack where citations live

Start with pages you own. Gemini sends 72% of citations to vendor and product domains but cites G2 in 14% of answers. Comparison content dominates: 84 of the 100 most-cited URLs were 59 comparison pages or 25 listicles, drawing more than 66,000 citations and defining the category.

Fix those pages before another blog post. Identify your category’s 10 most-cited comparison pages and listicles, check how each frames you, then publish against the top three competitors. James Scherer, Foundation’s VP of Strategy, put it: “Editorial gets the budget; product pages get the citations.”

Build community demand

Reddit earned 159,000 citations across 12,500 unique URLs, the most-cited external domain. Its logged-out audience, mostly arriving from search, grew 27% year over year in Q2 2026 to 77.7 million, outnumbering 52.6 million logged-in users.

Threads rank on Google, draw readers and feed Google’s AI answers. Citations can vanish: ChatGPT’s Reddit rate fell from 21% in July to 10% in August while Google kept citing Reddit. Community citations ranged from 5% for SEO platforms to 14% for EDR/XDR tools.

Spread work across many threads and subreddits, favoring 50 steady useful answers over one viral thread. Track community citations weekly; monthly reporting would have missed ChatGPT’s drop.

Use LinkedIn to define categories

LinkedIn is where executives and practitioners define the category narrative Google’s AI features can pick up. Social is the second-largest B2B software citation source after brand and product content, at 6%. Post category questions weekly from both groups, then measure AI Overviews and AI Mode; don’t chase likes.

Make video for Google’s search features

Even Gemini, also Google-owned, cites YouTube in just 0.3%, roughly 17 times less than AI Overviews. Google’s Search features use query fan-out to pull in ranking videos. Ship comparison, alternatives and best-of videos for pre-vendor questions, titled as buyers ask.

Keep review profiles active

Claude and Perplexity take one in five citations from reviews. Claude’s third-party citations are almost all reviews: G2 appears in 30% of source-citing answers, Capterra in 13%, and Gartner and PeerSpot in about 9% each. Review citations run from 8% in managed databases to 25% in ATS and recruiting software.

Buyers back this. G2’s 2026 research put review sites at 38%, just ahead of AI chatbots at 37%, as the top source shaping B2B software shortlists. Give your G2 profile a named owner and monthly target; for Claude and Perplexity, it is live retrieval infrastructure.

Set budget and owners

Ask which AI tools buyers use on sales calls and during onboarding. For Google users, prioritize community, video and social; for ChatGPT or Claude users, prioritize your pages and review profiles. Pull the category source mix before budget: review citations swing 3.3 times across categories, and community citations 2.5 times.

In managed databases, reviews earn 8% of citations and communities 13%, so copying an HR-tech playbook would misallocate money.

• Solo: own pages plus the category’s strongest layer. The base works across every surface; an unmaintained second layer goes stale.
• Three to five staff: add one layer from the other internet. Pair Reddit with review profiles for standalone engines; give reviews an owner because they stale fast. A second layer on the same side only strengthens visibility you already have.
• Eight or more: one owner per layer and one citation reporter; watch drift, which a blended score hides.

Measure by surface

Run 100+ fixed prompts on a schedule, track citation rate and share by surface, and alert on any 20% week-over-week swing. AirOps’ 100-point sentiment scores ranged from 81 on Perplexity to 69 on Claude. Tie results to branded searches, referral traffic and content-sourced pipeline.

Give it 12 months. Quarter one repairs old pages and sets baselines; by month 12, refreshed pages, profiles, threads and videos can draw citations as buyers ask questions. Weekly alerts catch platform changes that week.

Search gets you found; the stack gets you chosen.