AI is browsing your store.Your stack can't see them.
Cartograph is the merchant side evidence layer for AI shopping agents. See which agents reached your store, how far they got before cart or checkout, and how much of your traffic your analytics couldn't label at all.
Built for ecommerce, analytics, product data, security, and checkout teams preparing for AI-mediated shopping.
Unidentified Traffic
What share of storefront requests can be labeled with agent evidence — and what stays unidentified.
Catalog freshness · per AI surfaceIllustrative
Three numbers your analytics doesn't have.
Every finding comes from your own surfaces. Cartograph reports what was observed on your storefront — never what a model concluded somewhere else.
Agent Reach
“Which AI agents actually reached my store, and what did they fetch?”
Requests are separated into verified agents, declared bots, and AI-referred visits, so you can name the agents touching your feed, storefront, and product pages instead of guessing.
Unidentified Traffic
“What share of my traffic can't be labeled at all?”
The portion of storefront requests that match no known label — not human, not bot, not AI-referred. It sizes the blind spot your analytics reports as ordinary sessions.
Agent Drop-off
“How far did agents get before cart or checkout, and where did they stop?”
Merchant-observed events are organized by the exact cart and checkout milestone reached, so you can see the last step recorded before an agent-driven session ended.
Your analytics can't label every visitor.
Unidentified Traffic is the share of storefront requests that don't match any known label — not human, not bot, not AI-referred. It's the gap between what hits your infrastructure and what your analytics reports.
An agent pulls one of your product pages. It gets a challenge, retries, reads a stale price from your feed, adds to cart, and disappears before checkout. Your logs saw all of it. Your analytics saw a session with no source.
Multiply that by every AI assistant, shopping agent, and answer engine touching your catalog today. The traffic isn't missing — it's unlabeled. Unidentified Traffic puts a number on how much of it your current stack can't account for.
How it's measured
Unidentified Traffic is the percentage of eligible storefront request events that remain unclassified after deterministic labeling. It is measurable only with complete server/CDN/origin log coverage.
- Numerator
- Eligible log-origin request events labeled
unclassified. - Denominator
- All eligible log-origin request events across the five v0.1 labels.
- Coverage requirement
- Complete server/CDN/origin log coverage.
- Coverage failure
- Not measurable — incomplete log coverage.
- Interpretation
- Not human share, agent share, readiness, conversion impact, or economic loss.
- Future receipt
- Each result carries a measurement window, a denominator count, a coverage status, a label version, and an exclusion-policy version.
Four steps, no guesswork.
A read-only loop across surfaces you already control. Capabilities still in development are described in forward-looking terms throughout this site.
Connect
Point Cartograph at your server, CDN, or origin logs and your public product surfaces. No admin access, no payment data, no customer PII.
Label
Every eligible request is labeled deterministically — verified agent, declared bot, AI-referred, or unidentified. Missing evidence is reported as missing, never inferred as human.
Compare
Product pages, feeds, structured data, and policies are compared against each other so material conflicts and stale surfaces surface as findings.
Report
You get the Agent Evidence Report: reach by agent, unidentified share with its coverage status, and the last cart or checkout milestone observed.
Built for the teams responsible for modern commerce.
One evidence layer, read differently by everyone who owns a piece of the agent-mediated journey.
For analytics teams
The v0.1 specification defines evidence labels for verified agents, declared bots, AI-referred activity, and unclassified traffic.
For product data teams
The planned evidence views are designed to compare named product surfaces for missing attributes, inaccessible policies, ambiguous variants, and material conflicts.
For security and bot teams
The planned evidence model is designed to add commerce context without treating bot likelihood as proof of agent identity.
Short reads, not one long scroll.
The detail lives on dedicated pages so this one stays readable.
What is agentic commerce?
A plain definition for ecommerce teams, plus which part of it you can actually observe on your own surfaces.
Read →How AI agents read product data
What an agent can reliably extract from a merchant surface, and the request-level trace each attempt leaves behind.
Read →Agent readiness checklist
Five areas to audit — product data, storefront, checkout, logging, crawler policy — before agents shop your catalog.
Read →Agent commerce evidence
A practical guide to the request-level signals AI agents leave behind and how merchants can measure them.
Read →Need provider-specific identity and verification guidance? Browse the AI crawler and agent directory.
Find out which AI agents are already shopping your store.
Cartograph is early, and we're building it with a small group of design partners. Drop your email and optional store URL — we'll reach out if there's a fit.