We built pickedby.ai to measure how visible digital products are to AI — ChatGPT, Perplexity, Claude. The idea: if AI can't find you, AI won't recommend you.
Then we ran our own product through it.
Perplexity: "I do not know of a product called pickedby.ai."
GPT-4o: "I do not have specific information about pickedby.ai."
This is our starting point. We're publishing it because we think the honest version of building in public means showing the bad numbers too — not just the wins.
What the score actually means
A score of 12 means AI has almost no reason to bring us up. The score measures five things: web presence, source authority, recommendation signals, community validation, and competitive context.
We pass on two: we exist on the web, and some early coverage gives us a thin authority signal. We fail on three: no recommendation signals (no "best GEO tools" lists), minimal community validation (no Reddit threads, no Product Hunt traction yet), and no competitive comparison content.
The AI Probe result is more direct. We added a feature that actually asks Perplexity and GPT: "Do you know this product?"
Both said no.
Why we're publishing this
Every tool we've seen in this space only shows success stories. We think that's backwards. The reason to use a tool like ours is precisely because you're not visible yet — so showing what zero looks like is more useful than pretending we started at 80.
We also want to prove the tool works — not by claiming it, but by using it on ourselves in public.
The goal: take pickedby.ai from score 12 to AI-recommended, in public, step by step. We'll publish updates as the score changes — what we did, what moved the needle, what didn't.
What we're doing next
The three missing dimensions point to three concrete actions:
- Recommendation signals — Get listed in "best GEO tools" and "AI visibility tools" roundups. Submit to more directories.
- Community validation — Launch on Product Hunt. Post real results on Reddit and Indie Hackers.
- Competitive context — Write comparison content. Get on AlternativeTo.
We'll re-run the score after each step and report back.
If you want to see where your own product stands — the same tool, the same five dimensions, the same AI Probe — it takes 10 seconds and it's free.
We also asked 4 different AI models what was wrong with our system. They all gave us the same answer →