Why this founder failed
Benjamin Houy has pulled the plug on Lorelight, his GEO platform. The reason? He built something the market didn’t really need. And that’s the big lesson.
I came across a very candid blog from founder Benjamin Houy, detailing why he shut down Lorelight, the generative engine optimisation (GEO) platform he launched just months ago. It’s a straightforward admission that he built something the market didn’t really need.
“The tool worked. The technology was solid. Customers signed up,” Houy said. “But then they’d churn. Not because the tool was bad, but because the insights had zero impact on their behavior.”
Compelling pitch
Lorelight entered the market with a compelling pitch. As ChatGPT, Claude and Perplexity reshape how people find information, brands need optimisation tools for these AI platforms. The startup tracked brand mentions across AI responses, monitored visibility and analysed what drove citations in generative search results.
The concept seemed timely. With companies scrambling to understand their presence in AI-generated answers, a dedicated GEO tool felt like an obvious gap in the market.
But Houy discovered something unexpected when he analysed hundreds of AI responses across different platforms. “Brands with high AI search visibility all had the same characteristics,” he explained. “Quality content, mentions in authoritative publications, strong reputation, genuine expertise. Sound familiar?”
It should. Because it’s the exact playbook that’s driven SEO, PR and brand building for decades.
The realisation
“There was no secret formula. No hidden hack,” Houy admitted. “The AI models are trained on the same content that builds your brand everywhere else. They cite the same authoritative sources.”
This realisation forced an uncomfortable question: why build a specialised tool for something that doesn’t require specialisation?
Get FirstFeature.one’s stories in your inbox
Join Medium for free to get updates from this writer.
Sure, Lorelight could track mentions and monitor competitor visibility. But the action items were always identical:
- create better content;
- build relationships with authoritative publications;
- establish expertise.
“I had to ask myself: why am I building a tool for something that doesn’t need a tool?” Houy said. “The insights from Lorelight couldn’t change what customers needed to do. They would pursue the same brand-building fundamentals whether they had the data or not.”
Back to other project
Rather than iterate or pivot, Houy is returning his focus to French Together, the language learning business he’s run for a decade. It generates $10,000 in monthly recurring revenue without specialized tools or growth hacks, just solid fundamentals.
“French Together works because I focused on creating genuinely useful content, building trust through consistent value delivery, and establishing expertise through years of teaching,” he explained. He plans to expand additional languages, rebrand and launch a mobile app.
The big lesson
Houy’s experience highlights a trap many founders fall into, building solutions for problems that don’t actually exist.
“I could have kept going. I could have convinced myself that I just needed better marketing or more features,” he reflected. “But customers were churning because the product didn’t change what they needed to do.”
His advice to other founders is blunt: “Before you build that next tool, ask yourself: are you building something that will genuinely change what your customers need to do? Or are you creating a solution for a problem that doesn’t actually need a separate solution?”
It’s the kind of honest post-mortem that’s rare in startup culture, where pivots are rebranded as strategic evolution and failures are buried in press release jargon.
“Whether you want to rank on Google or get mentioned in ChatGPT responses, the fundamentals remain the same,” Houy concluded. “You don’t need a specialized tool for that. You need good marketing.”
For now, he’s betting on building brands the old-fashioned way, with quality, consistency and genuine value.
