Founder Lab · Day Zero

We scored 78 out of 100. No AI bot has ever visited. Here is our starting line.

Founder Lab is Atom Foundry's own store, real product, real checkout, built specifically so we can run every finding in this research series on ourselves before we ask anyone else to trust it. Before changing a single word on the site, we are publishing exactly where it stands: an AI Commerce Score of 78, robots.txt wide open to every AI agent, live UCP and agent-discovery endpoints, and, according to our own bot-visit tracking, precisely zero AI crawler visits in the store's entire history.

78
Baseline score, July 12
76 to 78
Range across 6 scans
0
AI bot visits, ever recorded
7/15
Technical, our weakest deterministic factor
What Founder Lab is

Our own laboratory, not just our own store

Atom Foundry operates its own live Shopify store, Founder Lab, as a permanent controlled experiment. Most measurement of AI commerce is observational: you watch stores you do not control and infer what matters. Founder Lab lets us do the opposite. We own every variable, so we can change one thing at a time, a product description, a schema field, a trust signal, and watch, in real time, how AI systems respond.

What we measure there goes beyond a single score. We track how each change moves the store's AI Commerce Score™, but also which AI agents crawl the store and when, which pages they read, whether they retrieve it in live shopping answers, and how its recommendation profile shifts over time. Every order still funds the independent research published across this site, but the store's primary purpose is this: it is the causal ground truth behind Atom Foundry's research, the place where we confirm what actually moves AI behavior before we claim it anywhere else.

Where we technically stand

Nothing is blocking AI. That is the easy part.

We checked the basics first. Robots.txt allows every user agent, no AI crawler is disallowed. The store publishes an agents.md file with explicit instructions for shopping agents, plus a live UCP discovery endpoint and MCP tool interface for agent-driven checkout. Google has already indexed the homepage, collections, product pages, and our About and Research pages, some as recently as hours before we checked. On paper, there is no technical reason an AI agent could not find, read, or transact on this store today.

The baseline

78 out of 100, and a noise floor we did not expect to find this fast

We ran the AI Commerce Score scan once on July 12: 78 out of 100, rated Moderately Recommendable. Five days later we ran it five more times in the space of 62 seconds, on a store we had not touched in between, just to see how stable the number is before we start changing anything.

Six scans of the same untouched store
July 12 baseline, then five rescans within 62 seconds on July 17
100 75 50 25 0 78 76 76 78 77 77 intent 6 intent 4 intent 4 intent 6 intent 5 intent 5 Jul 12 Jul 17, #1 Jul 17, #2 Jul 17, #3 Jul 17, #4 Jul 17, #5 Every scan on Jul 17 ran within the same 62 second window. Nothing on the store changed between them.
Every other component, visual 11, schema 15, trust 15, technical 7, price 12, rec 8, brand 4, was bit-for-bit identical across all six scans. Only intent moved, between 4 and 6. Score minus intent equals exactly 72 every single time. The only thing that varied was the one factor our own methodology already tells us is judged by a model rather than measured deterministically.
What this means for our own plan. Before we change anything, we already know a 1 to 2 point swing between two scans could be nothing but this same noise. Technical, at 7 out of 15, is our weakest deterministic factor and will not bounce around on its own. That is where we are starting, not with the noisier intent score, even though intent is the one factor our own research has already shown separates recommended stores from the rest.
The other zero

No AI bot has ever visited

We checked our own AI bot visit log, the same instrumentation this series has used to study crawler behavior elsewhere. Across every domain it monitors, founder.atomfoundry.dev does not appear once. Not GPTBot, not ClaudeBot, not PerplexityBot, not any of the agents we track. Not one recorded visit, ever.

We are not reading too much into this. The dedicated storefront tracker that should log every hit on this specific store currently shows zero rows across every store it watches, not just ours, which tells us as much about the state of our own instrumentation as it does about AI traffic. We are treating that as a separate, honest bug to fix, not folding it into this result. Even accounting for that, a brand-new store with no training-data footprint, no recommendation history, and almost no outside web presence is exactly the profile our own research predicts would see nothing yet.

Why this is the expected starting point

Everything we have published already explains this

A model can only know a domain from training-data memory or from live retrieval. Founder Lab did not exist when any current model was trained, so there is no memory to draw on, only retrieval, and retrieval depends on the store being discoverable and worth surfacing for a real query, not just crawlable in principle.

Recommendation is also, by a wide margin, self-predicting: what a model recommended last month explains 61.4% of what it recommends this month. A store with zero recommendation history has no momentum to inherit and has to earn a first position from nothing. And the one external signal that actually separates recommended stores from the rest, intent, the clarity of who a product is for, is a lever we have not pulled yet. We have not changed a single headline or product description.

The plan

What we are about to do, in order

First, technical, because it is deterministic and currently our weakest score component. Then intent, rewriting headlines and product copy to make it unambiguous who each product is for and why, since intent is the one proven separator between stores that ever get recommended and stores that never do. Every change gets multiple rescans, not one, so we can tell a real movement from the noise we just measured. And we are going to build a genuine web presence for this store, starting with linking to it plainly from atomfoundry.dev itself, since that is a real, topically relevant backlink, not a synthetic one.

Follow the experiment as it happens
Every change, every rescan, and the first AI bot visit or recommendation, whenever it happens, gets logged in public.
Read the Founder Lab Log