G2 is a powerhouse in AI Search. Thanks to a string of recent acquisitions—Capterra, Software Advice, GetApp—and an aggressive push to open its review data up to LLM crawlers, the G2 ecosystem now accounts for 84% of review platform citations (!) across major AI models.
Last month, Alexandra London, CMO of G2, joined us at Index ‘26 to share what G2 is seeing as AI rewires B2B buying. Her session, “How G2 Built Trust With Buyers, Bots, and LLMs All at Once,” was packed with previews from a recent G2 buyer survey, and a candid look at how G2 itself is playing the AI Search game.
Here are six of her insights worth resurfacing.
Kicking yourself for missing Index ‘26 in San Francisco? Good news: You’ve got a second chance in New York. We’re bringing the AI Search summit to NYC’s Union League Club on May 27, with speakers from KPMG, Webflow, Optimizely, and more. Pepper will also be sharing our own in-depth research into the GEO strategies actually moving the needle.
1. Buyer behavior is moving faster than your tech stack.
London opened with an overview of G2’s most recent survey, published in April of this year, which polled 1,000 B2B buyers about how AI is reshaping the way they discover, evaluate, and buy software.
One stat should be on every CMO’s radar: “Nine out of 10 folks have changed how they conduct searches, and that’s just since August,” she said. “When we run this study next, that number might be at 100%.”
The problem: Buyer behavior is moving faster than most marketing tech stacks were built to handle. Most teams are using tools built for a pre-AI era, and none of them were designed to play nicely with AI-native systems.
It’s time to take a hard look at what’s actually pulling weight. The phrase she used was “unlearn to relearn.” Audit what’s locked in, what’s coming up for renewal, and where the integration gaps are showing up in your AI workflows.
Takeaway for marketers: Build a quarterly review of how your existing tools are (or aren’t) holding up against AI-native workflows, and start the replacement or supplementation conversation well before any contract end date.
2. Your website now has two audiences.
London mapped the new B2B buyer path, which includes four main components:
AI Search (ChatGPT, Gemini, Claude, etc.) is now the most common starting point.
Software review sites (G2, Capterra) inform and validate AI’s suggestions.
Peer recommendations—i.e., texting your marketing friends for recs—still happen, but now more commonly as a gut-check rather than a first step.
Vendor websites provide the deep-dive layer, where shortlisted buyers go to get specific.
That means your website is now pulling double duty. It’s a destination for buyers, and it’s the source material LLMs scrape to form opinions about you. It has to read clearly to both humans and AI chatbots who are going to tell prospects all about you. It’s kind of like everyone is adding an entirely new persona they need to market to, and that persona is not human.
Takeaway for marketers: Stop optimizing your homepage purely for human visitors. The model is reading too, and it cares about clarity, structure, and factual specificity in ways your hero copy probably wasn’t designed for.
3. Claude’s power is growing.
While ChatGPT remains the most common LLM for B2B buyer searches, Claude has quickly become a top contender. G2 data found that Claude’s share of “where do you start your B2B search” jumped by ~21 percentage points between August and March.
It’s a point worth noting if your GEO strategy has focused solely on ChatGPT, and a trend worth tracking for the future (the pecking order is going to keep changing).
Takeaway for marketers: Keep tabs on how all the major LLMs are talking about your brand, not just the one with the highest market share.
4. Buyers are writing novel-length prompts, and your content needs to take that into consideration.
Buyers these days are writing dramatically longer prompts than they were even a year ago. “More and more, we’re seeing queries beyond a single sentence of ‘what’s the best CMS, what’s the best creative agency.’ Folks are writing novels in their prompts. ‘Here’s the current tech stack I’m on. How would it feed into this? Here’s exactly what I’m looking for.’”
Your buyers are detailing their tech stack, constraints, and their evaluation criteria, and expecting the model to weigh all of it. That expands the scope of what your content needs to cover in order to provide a satisfactory answer.
Takeaway for marketers: Generic feature pages won’t cut it. The content that gets cited speaks to specific use cases, integrations, and competitive comparisons.
5. Your profiles aren’t set-it-and-forget-it assets.
When product positioning evolves, marketing teams tend to focus on websites, decks, and campaign creative. But if you don’t update your G2 profile, you run the risk of muddling your message in AI Search.
This is true of all directory-style sites, so consistency is key. LLMs want to see the same facts across every place your audience shows up, and your messaging on platforms like G2 carries extra weight.
“If your brand narrative is in all your creative—the latest ad or billboard—amazing. But if you forgot the G2 step, then it really doesn’t matter when folks go to search,” London said.
Takeaway for marketers: Create a quarterly calendar invite to update your directory profiles with your latest messages and stats. Ensure they’re consistent everywhere you show up.
6. Stale G2 reviews = stale citations.
The other half of what's on your G2 page—customer reviews—has a different freshness problem. “If you are not capturing reviews in real time and at scale, you will fall behind in LLMs,” London said.
While you can’t edit or delete customer reviews, you can exercise some control over volume. Be sure you're prompting customers to leave new reviews regularly through CRM-triggered campaigns, post-onboarding nudges, and CSM outreach, so what LLMs see is recent. (And respond publicly to any negative reviews; AI Search takes note of brand replies too.)
Takeaway for marketers: Review collection should be an always-on motion baked into your post-purchase or post-onboarding flow, not an end-of-quarter scramble before a G2 grid update.
One standout quote from London summed up the session’s theme: "The authentic trustworthiness of your customers' voices is extremely important. And it cannot be forgotten, especially where we're headed."
Stay tuned to AI Native for more Index ‘26 recaps in the coming weeks.
AI news never sleeps, even if LLMs are in dreamland. Here’s what you need to know this week.
Google's AI Search is leaning harder on Reddit. A new "Expert Advice" section in AI Mode and AI Overviews will surface quotes from Reddit and other public discussions, with creator names and community handles attached.
Snap and Perplexity called off their $400 million deal. The partnership would have integrated Perplexity's search engine into Snapchat's Chat interface, but the companies couldn't agree on a path to rollout (product-speak for “this isn’t working”). Things reportedly ended on friendly terms.
The Musk-Altman trial is decidedly less amicable. Recent testimony included Musk admitting that xAI trained Grok by distilling OpenAI's GPT models, and Greg Brockman acknowledging he never followed through on his original $100K nonprofit pledge, despite now holding a $30 billion stake. Whatever the verdict, the discovery process is doing more damage than the lawsuit itself.
Anthropic is teaching Claude to dream. A new feature for Claude Managed Agents runs between sessions to review past tasks across multiple agents, identify recurring mistakes, and consolidate the most useful patterns into memory—basically REM sleep for AI agents. Sweet dreams are made of debugging logs.
That announcement came alongside two related agent updates (Outcomes and Multiagent Orchestration)—and a separate move to double Claude Code's 5-hour rate limits for Pro, Max, Team, and Enterprise subscribers, thanks to a new compute deal with SpaceX. (Yes, that SpaceX… Anthropic's now renting GPUs from the parent company of one of its loudest critics.)
Pennsylvania is suing Character.AI for impersonating licensed medical professionals. The case is the first to specifically target AI chatbots impersonating medical providers, and follows several wrongful death suits the company settled earlier this year involving underage users. A good reminder that "trust me, I'm a doctor" hits different coming from a chatbot.
Index ‘26 in NYC
We’re bringing Index ‘26 to New York later this month. Attendance is curated and seats are limited—secure your spot today.
What: A full day of sessions on AI-led organic growth, GEO strategy, and the future of organic search.
Who: A curated audience of marketing, demand gen, and SEO leaders from companies like KPMG, Webflow, Optimizely, and more.
Where: Union League Club, 38 E 37th St, New York, NY 10016
When: May 27, 2026
Senior Director, Marketing Operations @ Hinge (NYC, $238,000–$285,500 annual salary)
CMO @ Grand Circle Corporation (Boston, MA; $350,000–$450,000 annual salary)
Senior Director, Brand @ National Geographic Society (Washington, DC; $185,250–$195,000 USD annual salary)