I build AI tools that do real marketing work.
Not prompt screenshots — working tools, solving problems I actually hit as a marketer: getting recommended by AI engines, reading competitors’ winning creative, and finding the one fact that qualifies a lead. Every result here is real. Where a demo is cached or a method has limits, it says so.
Three tools, three real problems
Each has a working demo, the marketing judgment behind it, and an honest limits section.
AI Visibility Checker
Does ChatGPT, Gemini, or Claude recommend your brand when a buyer asks? This runs the questions a real buyer would ask, across three engines, and scores where you show up — and where a competitor shows up instead.
Why it’s smart: The question set is the product: it ladders buyer intent from problem-aware to the comparison and decision questions where deals are actually won.
Ad Longevity Finder
Advertisers kill ads that lose money. So the ads still running after months are the ones that work. This pulls a category's live ads from Meta's public library and ranks them by how long they've survived.
Why it’s smart: Ad longevity is a free, public, revealed-preference proxy for performance — and Meta gives you no way to sort by it. This does.
Giving-Platform Prospector
The one field that qualifies a church-giving SaaS prospect — which platform they use today — exists nowhere you can buy it. This 11-stage pipeline reads it off each church's own site, with the evidence attached.
Why it’s smart: Built to distrust itself: a verbatim tripwire threw out 241 of the model's own answers so it could ship zero hallucinations.
The build is the easy part. The judgment is the point.
Anyone can paste a prompt into ChatGPT. The value in each of these is a marketing decision: which questions a buyer actually asks, why a surviving ad is a better signal than a clever one, which single field turns a name into a qualified lead.
They’re also built to distrust their own output. AI is a confident, unreliable pattern-spotter, so every tool here verifies the model instead of trusting it — reporting frequency instead of a single yes/no, throwing out answers it can’t prove against the source, and telling you where it’s weak.
A tool that’s honest about its limits is worth more than one that sounds impressive. That’s the whole thesis of this page.