Why I built Optimly — and what I learned auditing 5,000+ AI brand responses
by•
Hey PH community 👋I'm Apurva, founder of Optimly. Before I ask for your support tomorrow, I wanted to share the insight that started all of this.While working at Microsoft, Meta, and Discord, I kept noticing something strange: AI models were describing our products with confident, detailed — and often completely wrong — information. Wrong category. Wrong features. Wrong competitive positioning.The kicker? 59% of AI brand misrepresentations aren't missing information — they're wrong categorization. AI models aren't ignoring your brand. They're misunderstanding it.That's a very different problem than "we need more content." It's a comprehension problem, not a supply problem.Optimly audits how ChatGPT, Claude, Perplexity, and Gemini actually understand your brand — and gives you a prioritized fix plan.My question for the community: Have you ever caught an AI model saying something wrong about your product or company? What was it?
38 views
Replies
This is a bigger enterprise problem than most realize. In B2B especially, procurement teams are now using AI to pre-screen vendors before any human ever visits your website. If an AI miscategorizes your product in that zero-click moment, you're not just misrepresented, you're filtered out of consideration entirely. You never even knew the conversation happened. The 59% misclassification stat hits differently when you realize the pipeline damage is invisible.
From a technical perspective, I'd push it one level further: it's not just misclassification, but rather it's inference poisoning. LLMs fill gaps with the nearest confident cluster. So if your brand sits between two well-understood categories, the model doesn't say "I'm unsure", it picks the louder neighbor and argues for it. The fix isn't more content; it's sharper semantic anchoring so the model's inference path resolves to you, not to your closest competitor. Optimly's audit surfaces the specific anchor points that are drifting.
@anoopdixith love that perspective - and why you are truly @Optimly's secret super power! Would you be willing to share what drew you to this problem and what you're excited to build in the coming weeks as we launch Optimly to the world?
@apurvaluty For me, the problem resonated because miscategorization in AI isn't just a marketing issue. It's a revenue leak that's nearly impossible to trace back to source. Most teams never connect a lost deal to how an AI described them three steps upstream. What I'm most curious to see Optimly build toward is temporal, longitudinal tracking. Model comprehension shifts every time these systems retrain, so the audit almost needs to be a continuous monitor of longitudinal snapshots. And the longer term play is fascinating. As AI interfaces inevitably move toward sponsored placements, brands that have already established strong comprehension baselines will have a structural advantage. They will know exactly how to position themselves, what conversational signal turns to a trigger, and which misperceptions to correct before they bid. Looking forward to that future whose glimpses we are already seeing.