I asked ChatGPT to recommend my product. It had no idea it existed.
12 days ago, I typed my own product into ChatGPT and asked: "What's the best tool for [my category]?"
It recommended 5 competitors. Mine wasn't on the list.
I asked Perplexity the same question. Same result — my product didn't exist in AI's world.
That was the moment I realized: Google SEO is slowly becoming irrelevant. 37% of people now start searches in AI tools instead of Google. And if AI doesn't know your product exists, you're invisible to a growing chunk of potential customers.
The Problem I Found
I'm a solo founder (UX background, zero coding skills — literally building everything with AI coding agents). I went looking for a tool that could tell me: "Does AI recommend my product?"
I found 24 GEO (Generative Engine Optimization) tools. Every single one was built for enterprises. The cheapest was $29/month. Most were $300+/month. None were built for indie creators selling digital products on Gumroad, Etsy, or Notion.
What I Built
So I built one. In under 2 weeks. With Claude Code as my only developer.
How it works: Enter any product name → get an AI Visibility Score (0-100) in 10 seconds. Free. No signup required.
It checks 5 dimensions:
- Web Presence (are you mentioned online?)
- Source Authority (are high-quality sites covering you?)
- Recommendation Signals (do "best of" articles include you?)
- Community Validation (Reddit, forums, reviews)
- Competitive Context (how do you compare to alternatives?)
Plus it directly asks ChatGPT and Perplexity: "Do you know this product? Would you recommend it?"
The Data That Surprised Me
I ran my own product through it. Score: 12 out of 100. Ouch.
Then I checked products that AI does recommend:
- Figma → 100/100 (Gold: "Picked by AI")
- Notion → 95/100
- A random Notion template with good SEO → 70/100
The gap between "AI knows you" and "AI has no idea you exist" is massive. And most digital creators are on the wrong side of it.
4 Things I Learned Building This
-
AI recommendations are not random. They follow patterns — web presence, authority sources, community mentions. You can influence them.
-
Most creators score under 25. If you're selling digital products and haven't done any AI optimization, you're likely invisible.
-
The fix isn't hard, just unknown. Things like creating comparison content, getting mentioned in roundup articles, and having an
llms.txtfile actually move the needle. -
This problem will only get bigger. AI search traffic is growing 500%+ year over year. The creators who figure this out early will have a massive advantage.
What Happened After Launch
Honest update: I launched 12 days ago. I have 1 real user (beta). I posted on Reddit (5.3K views, 1 conversion). My Twitter account got permanently banned because my AI agent auto-posted without my knowledge (lesson learned the hard way).
But here's the interesting part: I ran the tool on my own product again this week. Score jumped from 12 to 32 in 11 days.
What moved it: 13 blog posts, one Reddit thread that got 5.3K views, and 6 directory submissions. Perplexity now recognizes us in 8 out of 10 probes. ChatGPT still has no idea — but that was expected (training cutoff).
I asked 5 different AI models to evaluate my product. All 5 agreed:
- The idea is right (timing, market gap)
- The product works (clean UX, fast results)
- But I'm solving a problem people don't know they have yet
The biggest feedback: "You're giving a diagnosis but no prescription. When someone gets a score of 18, they feel bad and leave. Show them exactly what to do next."
What I'm Doing Now
I'm not building more features. I'm listening.
I'm onboarding 100 founders into the beta — first 100 get lifetime free access as we build. If you want to be part of this early, grab a spot here.
If you sell digital products (Gumroad, Etsy, Notion templates, online courses), I'd genuinely love your input:
-
Does "AI visibility" matter to you at all? Or is it too abstract compared to "how do I get more sales this month?"
-
If you checked your score and it was low — what would make you come back? What would you actually pay for?
-
Want me to check your product? Drop the name below and I'll run it for you.
The tool is free, no signup: pickedby.ai
Same pattern. ChatGPT naming you and ranking on Google are two different systems. Third-party pages and Reddit get repeated. The homepage often does not.
What has worked for me is logging the same five buyer questions across a few engines. You. A competitor. Nobody. One chat is a mood. A few weeks of the same questions is a measurement. Then one page that answers the buyer question in the first hundred words.
This matches what I saw. One chat is noise — I only started trusting it after 127 runs of the same prompts.
I log the same buyer questions across three engines (ChatGPT, Perplexity, Gemini) rather than five. Fewer engines, more repeats, was the trade that made the numbers stop moving randomly.
The "one page that answers the buyer question in the first hundred words" part matches my data too. That was one of the changes that moved my own score from 10 to 45.
Same wall from the other side. I run AI-visibility audits and kept meeting founders who had just learned their score and had no idea what to do with it. Diagnosis tools are multiplying; the prescription is the scarce part.
What has worked for my clients: a fixed checklist, run in order, no scoring. 47 checks across ChatGPT, Perplexity, and Gemini, each with the exact fix attached. I put the whole thing free on GitHub for exactly the indie/Gumroad crowd you describe: github.com/reachalexwynn/answer-engine-self-audit
Your scorer and that checklist are the same product split in half - you tell them they're sick, the checklist is the treatment. If you want to point your 18-scorers at it as the "what to do next," I'd consider that a win.
Diagnosis without prescription is exactly the gap, and you named it better than I did in the post.
Bookmarked the repo. The split you describe is real — I measure, and the next question every user asks is "so what do I change." I have been answering that by hand, one site at a time, which does not scale.
Happy to point people at the checklist as the "what to do next" step. If you ever want a second data source for it, my measurement runs are repeated over weeks rather than one-shot, so the before/after on a given fix is something I can actually show.
This is a really interesting problem space you're tackling. I happen to know a few solo founders selling digital products who'd likely be happy to answer your questions about their challenges and what would make them come back.
Yes, I would take that.
What I want to learn from them is narrow: whether an AI-visibility number is something they would check again a month later, or a one-time curiosity. That is the part I cannot answer from my own usage.
Introductions whenever convenient — no rush on my side.
“Super interesting build — especially the jump from 12 to 32 in just 11 days, that clearly shows AI visibility is a real lever.
The ‘diagnosis without prescription’ gap also makes total sense — if I saw a low score, I’d want exact steps to fix it immediately.
Quick thought — I’ve been seeing early-stage founders validate ideas using small, capped experiments (like $19 entry, 100 spots max, but with a strong upside). It’s surprisingly effective for both traction and real user intent.
Feels like something like this could actually help you get both users + validation faster right now. If you’re open, happy to share what I’m referring to.”
Thanks for reading it closely.
On the paid capped experiment — I am going to hold off for now. The tool measures three engines and I am still building the fix side, so charging entry before there is a prescription attached would be selling half a product. I would rather keep it free while the measurement data accumulates.
Happy to hear what you are referring to anyway, for later.