Lucas Mucek
Founder, Poland
I work in e-commerce and online marketing: online stores, product data, search and paid ads. I build CanItBuy with AI tools, and I answer every message myself.
Write to meOne person. An open lab.
CanItBuy checks whether AI shopping agents can actually buy online. The same agent, driven by different AI models, shops in our own reference stores every week. Every run is recorded and the results are public.
- reference stores
- 18
- AI models every week
- 5
- test modes
- 5
- latest run
- 23 Oct 2026
What we have built
Obstacle registry
13 numbered obstacles, each reproduced in its own reference store.
Course 2.0
The reference stores moved to English and US dollars. Ranking results count from this day.
Course 2.1: traps
Four stores check that the agent buys only what the shopper asked for.
The first public ranking
Report #1: which AI model can actually buy online.
How we work
Independent
- Not affiliated with OpenAI, Anthropic, Google, Alibaba, Amazon or any store platform. We pay for the AI models we test, like any customer.
- One set of published rules for every model. No company can pay for a better place. Method and limits.
How we make money
- The free store test and the public ranking cost nothing. We earn from private tests for stores, partners and teams that build AI agents.
- Paid work never changes the ranking. Private results go public only with written consent.
What we never do
- Our agent never pays, never places an order and never types personal or payment data.
- It says who it is, signs its requests and never works around a verification page. Our agent.
- We test only our own stores and stores whose owners agree.
Talk to the person who runs the lab
Questions about the ranking, a test of your agent or a mistake you found: write to hello@canitbuy.com. We reply within 3 business days. Service provider details are in the terms, section 1.