E-commerce
When an AI agent shops your category, does it buy from you?
Agents compare price, delivery and returns in seconds, then try to check out. Do they find you? Do they pick you? Can they get through your checkout? We measure your share of agent purchases against competitors, then fix where you lose them.
Spec-heavy categories first: home, outdoor, electronics and DIY.
Buying task: Find a waterproof family tent that arrives before the weekend, with free returns, and buy it.
-
Surfaced Searching stores
- Fellcraft (you)
- Competitor A
- Competitor B
-
Chosen Comparing delivery, returns and specs
ChosenFellcraft
Trekker Family Tunnel Tent
Arrives Friday · Free returns · Waterproof rating listed
-
Completed Checking out
- Basket
- Delivery
- Payment
Order placed
The problem
The agent channel is one funnel with three leaks.
Agents surface some stores, choose fewer, and complete purchases with fewer still. Most businesses can’t see where they lose, because AI visibility tools only measure the first step and agent-testing tools only the last.
- Surfaced 34 of 100 buying tasks
Does the agent find your products for a buying task?
We don’t know if shopping agents even find our products.
- Chosen 12 of 100 buying tasks
When it compares price, delivery, returns and data, does it pick you?
When an agent compares us, we can’t see why it picks a competitor.
- Completed 7 of 100 buying tasks
Can it actually get through your site or quote journey and buy?
Our bot protection might be turning away real buyers’ agents.
Across all three “The board asks about agentic commerce and we have no numbers.”
How a test runs
Watch agents shop your category.
We run the same buying tasks through ChatGPT, Claude, Gemini, Copilot, Perplexity and browser agents, from several locations at once. Every run is recorded, and every result lands in the funnel: surfaced, chosen, completed.
A simulated run of 100 buying tasks by ChatGPT, Claude, Gemini, Copilot, Perplexity, ChatGPT agent, from London, Manchester, Dublin, Amsterdam, Berlin, Paris. Fellcraft was surfaced 34 times, chosen 12 times and completed 7 times. Illustrative example.
Simulated run · 6 agents in parallel
What we fix
Fix the data agents compare.
Agents choose on facts they can read: specs, delivery, returns, stock. When those are missing, vague or out of reach, the agent picks a competitor or gives up.
| Field | Before | After |
|---|---|---|
| Title | Before: Trekker – Green Vague | After: Fellcraft Trekker Family Tunnel Tent, Forest Green |
| Key specs | Before: In a PDF spec sheet Vague | After: Capacity, packed weight and waterproof rating as structured attributes |
| Delivery | Before: Missing | After: Delivery dates by postcode, in the feed and on the page |
| Returns | Before: On the FAQ page Vague | After: Returns window and cost in the product data |
| Stock | Before: Shown after adding to basket Vague | After: Live availability for every variant |
| GTIN | Before: Missing | After: Set for every variant |
| Agent access | Before: Unknown agents blocked at checkout Blocked | After: Verified agents allowed; bad bots still blocked |
How it works
Benchmark, optimise, iterate.
-
Benchmark
Real AI agents run realistic buying tasks or quote journeys. Get insights from every agent session to improve your funnel.
-
Optimise
We fix the biggest leak first: whatever stops agents seeing your products, choosing them over a competitor’s, or completing checkout.
-
Iterate
Agents re-run after every site release, plus AI-driven traffic and sales attribution, so gains compound and regressions get caught.
A continuous optimisation loop: every release is tested, every gain is measured.
Why Veroscale
One view of the whole funnel.
| GEO agencies | Feed managers | Agent-testing audits | Systems integrators | Veroscale | |
|---|---|---|---|---|---|
| AI visibility | No | No | No | ||
| Audits and recommendations | No | No | |||
| Feed management | No | No | No | ||
| Agent protocols and integration | No | No | No | ||
| Buying-journey exploration by real agents | No | No | No | ||
| Why agents choose competitors | No | No | No | No | |
| Ongoing measurement and optimisation | No | No | No | No |
Everything you need to be agent-ready, in one place. Visibility, product data, protocols and buying-journey testing from one team, measured and improved over time. Neutral across agents and protocols.
Free, no obligation
See where agents lose you, free.
What you get
- 10 buying tasks or quote journeys, run by 3–4 AI agents on your site and two competitors
- A funnel scorecard: surfaced, chosen, completed
- The three fixes worth doing first
Who you’ll work with
Senior specialists, not a junior account team.
Veroscale is run by experts with decades of combined experience across marketing, product management, engineering and data, including e-commerce product data and AI products.
FAQ
Questions we get asked
Is this GEO?
GEO is one part: being surfaced. We also measure and improve whether agents choose you and can actually buy.
Which agents do you test?
ChatGPT, Gemini, Copilot and browser agents such as ChatGPT’s agent and Perplexity Comet. The mix changes as agents evolve.
We’re on Shopify. Aren’t we covered?
Shopify connects you to agent protocols. It doesn’t tell you whether agents choose you over competitors, or where they fail.
Which platforms do you work with?
Shopify, Adobe Commerce and Salesforce Commerce Cloud first. The free test runs on your public site, so it works whatever you’re on.
Do you need access to our systems?
Not for the free test: it runs on your public site. Fixes and attribution need access, agreed per engagement.
Isn’t letting agents in a security risk?
Blocking all bots also blocks real buyers’ agents. We help you let verified agents in while keeping bad bots out.
How do you measure impact?
Funnel rates before and after, agent success, steps and cost per purchase, plus AI-driven traffic and sales where your analytics allow.