AI Agents Now Outnumber Humans in G2's Audience

AI agents made up 68% of G2's audience in Q2 2026, leaving humans at 32%, according to Alex London, CMO of the software marketplace and review platform. "We are now talking to humans and agents. We've never built a playbook for that," London said.

The shift is already measurable in G2's business. The company attributed more than 5% of referral sessions, more than 20% of referral leads and over 10% of referral-attributed revenue to AI discovery after it began separating organic AI-assistant traffic from conventional referral traffic. Paid answer engine optimization, by contrast, accounts for less than 1% of its paid-search clicks, leads and revenue. Total traffic rose 35.5% year over year. G2 uses Profound Agent Analytics and Cloudflare traffic data to identify agent visits and separate user-triggered activity from ordinary crawlers and background requests.

How widespread is chatbot-assisted buying?

A G2 report, "The Evaluation Maze," found that 82% of B2B buyers had sourced software recommendations from an AI chatbot in the prior two years. Those chatbots and the AI agents behind them — systems that research and recommend products on a buyer's behalf — form the second audience London described. G2 sells vendors advertising and buyer-intent products, giving it a commercial stake in how reviews and AI discovery shape software buying.

Because buyers now use chatbots to research products before visiting a vendor's website, a brand's presence across the web matters as much as its own site. AI systems draw on both company content and third-party sources, including reviews, when generating answers or shortlists. London's advice: keep brand messaging consistent across those sources and prioritize useful, accessible information rather than publishing more content for volume. "It goes well beyond just the content," she said. "It's around that trust layer."

She recommends reviewing a brand's digital presence about every 90 days, based on her read of model refresh patterns, though schedules vary by system.

Does AEO replace SEO?

London does not think so. She said search engine optimization, search engine marketing, answer engine optimization and what G2 calls answer engine marketing should work together. "I don't think it's an 'or.' I think [that's] where your investments go," she said.

Marketers should maintain the foundations of search visibility while adapting content for AI-driven discovery, London said. Third-party validation, including reviews, offers information beyond a company's own claims.

"If you don't start to invest in AEO and answer engine marketing, which is really how you're marketing to these models, then that's where you're going to fall behind," she said.

How should marketers measure AI visibility?

London said brands should track whether they appear in the answers buyers use to build shortlists and whether that visibility connects to business outcomes. A mention or citation does not necessarily mean a system is recommending the brand.

"You can't just stop at a mention and a citation. You have to make sure you're winning the answer," she said.

What does G2 run internally?

G2 has built roughly 90 agents across its marketing organization, though London cautioned that companies sometimes use "agent" to describe simpler automated workflows. One example: a paid-search workflow, built by a G2 employee using Claude, that analyzes search-term exports, flags irrelevant queries and identifies keyword expansions. The workflow cut tasks that previously consumed seven to 12 hours of team capacity each week down to minutes — the equivalent of reclaiming more than a full workday per week.

London said teams should give AI tools clear context and brand guidelines, test them before relying on them at scale and measure whether they improve efficiency. The goal should be measurable efficiency gains, not AI use for its own sake. For vendors on G2's platform, the implication is direct: with agents now the majority of its audience, discoverability by machines has become a primary acquisition channel, not a side effect.