Mike Holownych

Mike HolownychAug 3, 2026

I audited 73 landing pages running paid ads. Every single one failed the same check. Not most. Every one. The check takes 30 seconds. Most founders have never run it. Swipe to see what I found, and what to fix before you spend another dollar on ads. → If this is your page right now, drop a 🔥 in the comments and I'll audit it live. If you know a founder running ads with a page like this, tag them; this might save them their next month's ad budget. ♻️ Repost to help someone fix this before they blame the ad.

Mike HolownychJul 25, 2026

Most AI visibility tools cannot defend the numbers they sell. I built one that refuses to produce the number. Citable must not produce a verified conclusion it cannot support with evidence. That means: • Missing facts must return blocked, not a plausible guess. • Expired evidence must invalidate the claim it supported. • Technical findings must be traceable to captured source state. • Heuristics must be identified as heuristics. • Observed citation behaviour must remain separate from predicted “AI visibility.” • Another person must be able to reconstruct and challenge the result. So I am publicly launching the Citable Review Challenge. I am looking for independent reviewers willing to test one or more of these questions: Can Citable’s findings actually be reconstructed from its evidence package? Does it refuse conclusions when the required evidence is unavailable? Are its documented limitations materially honest? Does the evidence model improve the audit, or merely create procedural overhead? Can you make it confidently assert something it cannot defend? Is the entire system overengineered for the problem it claims to solve? That final question is explicitly in scope. I am not looking for favourable coverage. Negative findings can be published without my permission. I will not request quote approval, ratings, backlinks, talking points, or editorial changes. The only requirement is that the tested version or commit is identified so the result can be reproduced. I am particularly interested in reviewers from three groups: • Technical SEO, AEO, or GEO practitioners • Coding-agent and developer-tool reviewers • Engineers who work in evidence, testing, measurement, governance, or adversarial validation Citable is Apache-2.0 and publicly available: Repository: https://lnkd.in/gHFBD9G6 Reviewer overview: https://lnkd.in/gHQCEdZw The interesting question is not whether Citable finds more issues than another audit tool. It is whether its conclusions survive scrutiny better. If you review technical tools and want a bounded review protocol, comment REVIEW or message me directly.

Mike HolownychJul 18, 2026

Your AI Visibility Score is probably made up. Not fake. Just... impossible to defend. Here's the question nobody asks the vendors: "Show me the evidence behind this score." Not your methodology page. Not your marketing deck. I want the actual audit. • The robots.txt that was captured. • The headers that were observed. • The schema that actually existed. • The prompt that was used. • The model's response. • The timestamp. • The checksum proving none of it changed afterward. Can they produce it? Most can't. Because the score was never designed to survive scrutiny. It was designed to look authoritative in a dashboard. Here's the uncomfortable reality: Nobody can guarantee ChatGPT, Gemini, Claude, or any other answer engine will cite your content. You don't control citation. You control eligibility. That means: • Is your content crawlable? • Are your claims supported by evidence? • Is your entity information consistent? • Can every important statement be reconstructed months later? Everything after that is probability. So I built Citable differently. It doesn't give you an "AI Visibility Score." It gives you something much harder to argue with: An evidence package. Every finding is traceable. Every claim can be challenged. Every limitation is documented. One of my favorite files in the repository is known-limitations.md. If the tool cannot know something, it says so. No fabricated certainty. No confidence theater. Just evidence. Because after more than twenty years in regulated infrastructure and incident management, I've learned one lesson that applies just as much to AI as it does to production systems: If you can't reconstruct the claim, you can't defend the claim. If you'd like priority access to the repository before it's widely circulated, we'll need to be connected on LinkedIn. Send me a connection request, then comment AEO on this post. I'll prioritize sending the repository to people who are connected and who help get this discussion in front of more practitioners by sharing the post. The industry doesn't need another score. It needs evidence.