G2 Reach: Who wins the answer

I arrived at G2 Reach ready to talk about AI and how we help our brand partners get found. The room was abuzz, still digesting the Nvidia news and knowing more was coming, when OpenAI opened Business Hub. There, companies can claim and verify their presence in ChatGPT, update company details, and add product feeds with pricing and availability. It also shows how often a business is mentioned, how many people click through to its website, and the topics behind the conversations where its site is cited. And just like that, the goalpost moved again. I expect competing offerings to follow across all frontier models, changing how we measure.

It would be easy to call this the end of the Google Business Profile era. It isn’t, yet. But we’re keeping our eyes on it while pairing these thoughts with the sessions at the G2 event, and the picture is clear. For 20 years, Google taught brands to chase a ranking. Answer engines recommend, and they build the recommendation from evidence that other people supply. So, we have our marching orders: win the answer.

Moving from the Yelp for software to the evidence layer for AI

Godard Abel, G2’s co-founder and CEO, described the company’s 2012 starting point as “a Yelp for software.” Nearly 15 years later, that is not what we’re looking at today. In January, G2 acquired Capterra, Software Advice, and GetApp from Gartner, and it now runs four marketplaces with roughly 6 million verified reviews and reaches more than 200 million annual software buyers. This level of scale is critical because the reviews still serve human buyers, and they increasingly serve the AI agents researching on those buyers’ behalf. The second job is winning “agent trust,” and he left no doubt about the stakes: “I am convinced [AI] is the biggest opportunity yet.”

Here’s how G2 is taking charge of this opportunity:

• G2’s new unified seller portal, puts all four marketplaces behind one login. A single customer review request now distributes across every property, and G2 says each new review gets 4X the exposure. Ads and human-verified leads are managed in the same place.

• G2 Agent Evaluations put AI agents to work on real workplace tasks inside synthetic companies, each with its own departments, data, and policies. Every interaction is scored against rubrics, and agents are ranked on a public leaderboard where buyers can inspect the conversations and grading. The program has been in public beta since September, with early customer experience results and Voice AI results published, and AI SDR and Legal AI agents under review.

• AI Custom Research sends AI voice agents to interview hundreds of real buyers, and G2 analysts review the themes. G2 says fieldwork that once took months now takes weeks.

• Content Studio, now in beta, turns reviews and research into branded infographics, landing pages and carousels, built from a company’s own brand book, logos and colors.

• The AEO Recommender shows how a brand and G2 perform in AI answers for its categories and recommends what to do next. It also lets marketers add the prompts they care about most.

• G2 MCP and Ask AI put G2 review data inside AI assistants such as ChatGPT and Claude, for competitive research and review campaigns.

• Voice reviews let customers leave a review by talking instead of typing. G2 has offered them since 2025, and Godard’s reason still holds: “nobody really likes typing reviews.”

The agent evaluation session tackled: “Why should I trust you to automate my business?” while “every agent pitch [sounds] the same.” Vendors run their own evaluations, “and no surprise, their agents always come out on top.” The conclusion: “the big leap in enterprise AI… It won't come from capabilities alone. It will come from the confidence in what AI actually can deliver.” “There has never been a breakout product that was built on average traits.” G2 says its citations in AI answers grew 8.5 times over six months, and it held the top share of voice in 901 categories in September, up from 166 in April. So, I’m sold.

Most prompt tracking is built on sand

Tim Sanders, G2’s CIO, opened with the analogy of Blockbuster. Netflix launched, Blockbuster spent years arguing it was a niche, and by the time it responded, Netflix had already moved on to streaming. His warning to the room was about the marketer who looks up a year from now still spending the same share of digital dollars on search. However, his bigger point was measurement, a well-timed point given the morning’s OpenAI news. “The real problem is the misalignment between the prompts you’re tracking, and the actual prompts people put into ChatGPT.” “Bad flour equals bad bread.”  

Teams that scrape their own keywords and turn them into questions end up tracking language buyers don’t use, and they collect false positives. Buyers describe the job they need done and the constraints around it. His example was a hospital looking for scheduling and messaging software that works as well on a phone as on a desktop, because the staff is always on the move. This is such an important point and a differentiator from the lessons we all received in the early SEO days. Things have changed; the way we operate needs to, too.

This is also important when brands create categories that don’t naturally exist. He had a great example of a company that stopped calling its product a “whiteboard” and started calling it an “AI visual presentation software.” Meanwhile, a competitor whose content was all about whiteboards was reporting real pipeline from ChatGPT.  

“AI visual presentation software” was likely a decision made to appear innovative and get “AI” into their company description. But if you want a tool like a whiteboard, Tim is right: that’s what you put in as a prompt query, and brands can’t afford to lose that discoverability. We see this mistake all the time and is why a messaging and positioning session that starts with our Marka+ AI baseline is so important.

He also described how the models narrow a field. They start with a broad candidate set, around 14 tools in his example, then look for proof that each one meets the buyer’s requirements, and return about three. “First-party content is necessary but not sufficient in the new world,” he said. “The models love third-party. But they absolutely love online reviews.” He provided another example: a cybersecurity company claimed latency under 25 milliseconds. Still, none of its customer reviews mentioned latency, so the claim had little to stand on. “Specificity is the word.”

He left the room with a budget question for 2027: how much of what you spend on SEO tools, services, and content moves to answer engine optimization? I will certainly be giving this a lot of thought as we present our 2027 strategies for our brand partners.

AEM and the shortlist you can’t see

Alex London, G2’s CMO, introduced answer engine marketing, or AEM, a term that will sound familiar to any marketer who lived through the rise of SEM. It is G2’s new services category, defined as using paid tactics to influence which brands show up in AI-generated answers.

“The market is very crowded, and pipeline is really hard to see,” London said. “And all of our revenue models are changing.” AI is compressing the buyer journey, and the shortlist now forms inside an answer, out of view. The tool market has crowded up to match: about 10 AEO companies 18 months ago and more than 700 now, “all fighting for your dollar.”

AEM isn’t a replacement for anything. “It doesn’t mean you stop SEO,” she said on stage. In a Chief Marketer interview, she said SEO, SEM, AEO and AEM belong in the same plan, and warned that brands not investing in AEO and AEM now will fall behind. She also shared that AI agents made up 68% of G2’s audience in Q2 2026, against 32% humans: “We are now talking to humans and agents.”

Her advice for the rest of us: track agent activity alongside web traffic and choose the few tools that fit your business. Know where your dollars go and what each one adds before asking for more. When it comes to visibility metrics, a mention or a citation isn’t the goal until it turns into a recommendation. ✅

What this means for brand and communications leaders

The dashboards have arrived. OpenAI’s Business Hub reports mentions and clicks, but without the prompts behind them, marketers still need to identify the questions they want to track and, more importantly, the questions their buyers are actually asking.

Godard called this the biggest opportunity yet, and I agree. I expect every frontier model to offer its own version of Business Hub within the year, and the dashboards and ad products will keep multiplying. The advantage will go to brands that know what their customers say about them before a model does and puts in the work of getting those customers to say it publicly and in specific terms.  

That is how we’re approaching 2027 planning with our brand partners: fewer new tools and a lot more customer evidence. The answer engines will decide who gets recommended. Our job is to make sure the proof is already out there when they go looking.

‍