Most B2B SaaS founders I talk to are still optimizing for page-one Google rankings while their buyers have quietly stopped typing queries into search bars altogether.
The Shift Nobody Is Measuring
I noticed something odd last year when I was doing customer interviews for MentionFox. Three separate prospects told me they had discovered us through "something an AI told them." Not a blog post. Not a listicle. Not a G2 review. They had described their problem to ChatGPT or Claude or Perplexity, and our product came up in the answer. I had done nothing deliberate to make that happen. It just happened because we had enough signal scattered across the web that the models had absorbed us.
That was the moment I started taking generative engine optimization seriously. Not as a buzzword, but as a distribution channel with no dashboard, no click-through rate, no impressions column. A channel that was already sending us warm buyers and I had zero visibility into it.
The thesis I landed on is this: Google ranks pages. LLMs recommend vendors. Those are fundamentally different games, and almost every B2B SaaS team I know is still playing only the first one.
What I Actually Found When I Dug In
The first thing I did was ask a dozen different AI assistants variations of the same question our customers ask when they are looking for a tool like ours. Things like "what is the best tool for tracking brand mentions across social and the web for a B2B company" or "how do I monitor when my SaaS brand gets mentioned in Reddit threads."
The results were humbling. Some tools I had never heard of showed up consistently. Some very well-funded competitors barely appeared. And our own positioning was inconsistent - sometimes accurate, sometimes a version of us from two years ago, sometimes missing entirely.
The pattern I noticed was not about domain authority in the traditional SEO sense. It was about depth of explanation. The tools that appeared most often in LLM answers had something in common: there was a lot of third-party explanatory text about them. Forum posts where someone explained how they used the tool. Comparison threads on Reddit and LinkedIn where practitioners debated tradeoffs. Use-case writeups on niche newsletters. Not just press releases and product pages. Actual human beings explaining the product in their own words, in context, in response to a real problem.
LLMs are trained on language that describes the world. If the world has not described your product in the language your buyers use to describe their problems, you will not appear in the answer. Your keyword-stuffed blog post optimized for "brand monitoring software" is less useful to a language model than one authentic Reddit comment from a real user explaining how they set up alerts for competitor mentions.
The second thing I found was that recency matters more than I expected. Claude and GPT-4 have knowledge cutoffs, but Perplexity and the Bing-based assistants pull live web data. If your product has gone six months without any new third-party coverage, new case study language, or community discussion, you start to fade. The models that do live retrieval will simply stop finding you. This is not a set-it-and-forget-it optimization.
The third finding was the most uncomfortable. Several AI assistants described our product confidently but inaccurately. One described a feature we had deprecated. One mixed up our pricing tier logic. One correctly identified our core use case but attributed a capability to us that belongs to a competitor. Hallucinations are not just a risk for your buyers - they are a reputation risk for your brand, and you will never know they are happening unless you are actively monitoring.
What We Built to Fix It
This is the part where I have to be honest about the fact that I built MentionFox partly to solve my own problem. We added a GEO dashboard specifically because I wanted to see, in one place, which AI platforms were citing us, what they were saying, and whether the description matched our current positioning.
The workflow we settled on internally is straightforward. We run weekly prompts across the major LLM interfaces that mirror the language our ICP uses when they are in problem-recognition mode. Not "MentionFox" as a direct query - that just tests brand recall. We ask questions the way a VP of marketing at a 50-person SaaS company would ask them. We log the outputs. We flag inaccuracies. And we trace back the source documents the models are likely drawing from, then update or supplement those documents.
For GEO monitoring specifically, we track which third-party sources tend to get cited when AI tools answer questions in our category. That tells us where to focus content and community energy. If a particular Substack newsletter or a specific subreddit keeps appearing as a source, that is worth more effort than another SEO-optimized landing page that no AI is surfacing.
The metric I now care about is share of LLM answers in our category. Not share of voice in the legacy sense. How often, when someone asks an AI assistant about our problem space, does a version of MentionFox show up, and does it show up accurately.
What You Should Actually Do
If you are running a B2B SaaS right now, here is the minimum viable version of this. Pick five questions your ideal customer would ask an AI assistant when they are in active evaluation mode. Ask those questions across ChatGPT, Claude, Perplexity, and Gemini. Screenshot the answers. Look for three things: whether you appear at all, whether the description is accurate, and which tools appear more consistently than you.
Then look at the sources. Perplexity will often show you citations directly. For the others, you can make educated guesses based on what ranks organically for similar queries. Your job is to create or influence content in those source locations. That might mean engaging authentically in the Reddit threads your buyers use. It might mean writing a detailed use-case walkthrough on a niche newsletter. It might mean making sure your documentation uses the exact language your customers use when they describe their problems to each other.
The underlying shift here is not technical. It is a recognition that your distribution is now partially mediated by systems that learned from the web. If your product does not have a rich, accurate, current footprint in the language your buyers use, you are invisible in a channel that is growing faster than anything else in B2B discovery right now. Google is not going away. But it is no longer the only door.
If you want to see how MentionFox handles LLM visibility tracking and surfaces inaccurate AI mentions before they cost you a deal, take a look at our GEO dashboard. And if you want to know what it costs to get started, here is our pricing.
If you found this useful, I write about solo-founder distribution, B2B SaaS, and what's actually working in the AI-search era over on my Substack (one post per week, no spam).
I'm building MentionFox - a B2B intelligence suite that combines brand mention tracking with AI-visibility (GEO) measurement, investor research, and outreach automation. There's a free tier and a 5-day trial of Pro at mentionfox.com/pricing.
Top comments (0)