Simulated: every visitor, crawler, agent, payment, and signature here is a script. No real service or standard is used.
1. Publish your work and set terms
Signals this page sends Illustrative, not a standard
2. Visitors arrive
Press a visitor. The gate reads your terms and decides.
3. Your creator ledger
What came back to you for what machines took.
4. Visit log
No visits yet. Press a visitor.
How it works
The problems this demo is about
- Consent. A creator cannot say "search yes, training no" in a form every machine reads and keeps.
- Credit. AI answers use the work but often drop the link, so readers never arrive.
- Payment. Value moves to the answer engine. Nothing comes back to the person who made it.
- Enforcement. Terms are requests. A bot can ignore them or claim to be a browser.
- Proof of authorship. Human work is hard to tell apart from copies and generated text.
- Retreat. Creators move behind logins and closed apps, so the open web gets smaller.
Pain point
AI systems take creators' work and send back few readers, so publishing on the open web no longer pays.
Problem statement
Creators cannot say how each kind of machine may use a work. They cannot tell which crawler is which, and they get no credit or payment. Some may retreat behind logins and closed apps.
What sparked it
Cloudflare Radar crawl-to-referral ratios show that AI crawlers take hundreds or thousands of pages for each visit they send back.
Why this approach
Terms travel with each work, and a gate at the host or CDN enforces them. People read free, and machines show identity, credit, or payment.
Hypothesis tested
We believed that a terms gate makes publishing in the open worth it again. If it holds, creators who see the ledger will say they would publish more in the open.
How to try it
- Open the Prototype tab. Change the terms in step 1.
- Press each visitor in step 2. Read the verdict in the log.
- Switch between "Signals only" and the terms gate. Press the same visitors again.
- Send 25 mixed visits. Read the ledger.
What this does not do
- It does not connect to a real crawler, agent, or payment system.
- It cannot take back work that a model already used for training.
- The signals format is illustrative. It is not a standard.
Notes
- Buyer: a blog host, a newsletter platform, or a CDN that offers the gate as a feature.
- Model: the host takes a share of payments through the gate. The share is an assumption.
- Weak spot: AI companies must agree to sign their crawlers and to pay.
- Closest standard: RSL 1.0 adds license terms to robots.txt. No major model developer has said it will honor it.
- Feedback: which use would you allow for free, and which would you charge for?
Traces
| Step | Prompt | Model | Tool | Date |
|---|
Cost
| Step | Model | Input tokens | Output tokens | Cost (USD) |
|---|---|---|---|---|
| Total | ||||
Not measured: one lead session wrote this prototype together with its research and publishing. The session did not report token use for this prototype alone, so the table is empty.