Prove who picked your paywall lock.
Invisible canary phrases, unique to each visitor, sit inside your paywalled text. When an AI answer engine repeats one, Gatecrash ties it to the logged session that broke in to get it.
- Be first in line
- One email, when we open
- No spam
- Rotating canary phrases0%
- Fingerprinting sessionsQueued
- Probing answer enginesQueued
- Clustering sessions into actorsQueued
- Drafting the evidence packQueued
Catches AI scrapers picking your paywall lock, and logs every break-in for your lawyer.
Each paywalled article carries invisible canary phrases, unique to every visitor. Gatecrash asks AI answer engines leading questions all day until one repeats a canary, then links the hit to the logged session that dodged a login or CAPTCHA to get it. Your own lawyer gets the evidence and decides what to do.
- Step 01
Marks every paywalled page with canaries
A JS challenge and a metered paywall keep forensic logs, and each visitor’s copy of the text carries invisible canary phrases unique to their session.
- Step 02
Probes the answer engines every day
Agents ask the big answer engines and AI search APIs leading questions, and match any canary that surfaces to the session that took it.
- Step 03
Packs the evidence for your counsel
Each month, every act of circumvention: the session, the method, the network, the canary and where it surfaced, with a chain-of-custody log.
Built for paywalled trade, data and newsletter publishers.
An access-control layer with forensic logs
A JS challenge and metered paywall that record each session, with per-visitor canary watermarks in the protected text.
A daily canary probe
We check the major answer engines and AI search APIs for your canaries every day, and log each hit.
A monthly evidence pack
Each break-in with its session, method, network, canary and chain of custody, ready for your own litigation counsel.
First up: We start with a free probe: paste five paywalled URLs and get a same-day report on whether AI answer engines reproduce your text.
The world just changed.
What changed, and where it was reported.
In Reddit v. SerpApi/Perplexity, a New York federal court let DMCA circumvention claims go ahead. Reddit alleged that proxies, fake user agents and ‘mimic human behavior’ features got around Google’s CAPTCHAs.
Loeb & LoebIn Ziff Davis v. OpenAI, Judge Stein ruled that robots.txt is only a ‘keep off the grass’ sign, not an access control, so ignoring it is not circumvention. Only a real lock counts.
Barry SookmanUS law sets DMCA statutory damages at $200 to $2,500 per act of circumvention, and lets the court award reasonable attorney’s fees to the winning side.
Cornell LII
One person and 4 AI agents.
The agents, all day
- Rotating canary phrases
- Fingerprinting sessions
- Probing answer engines
- Clustering sessions into actors
- Drafting the evidence pack
What a person does
Your own IP litigator, hired and paid by you, decides whether to send a letter or sue, and signs everything. An independent forensic expert, never paid on the outcome, reviews our method and signs declarations.
You document real circumvention of your own access controls, and your own lawyer decides what to file. We never claim robots.txt violations alone, never lure anyone onto bait pages, identify every actor by more than its user agent, and never send demands or share a recovery.
Is ignoring robots.txt enough for a claim?
No. A court has ruled that robots.txt is not an access control, so we never count it. Only breaks of a real lock, such as a login, a CAPTCHA or a JS challenge, go in the evidence pack.
Could you blame the wrong company?
That is the main risk. Attackers forge AI-crawler user agents, so a user agent proves nothing. Every actor is tied to network, canary and behaviour evidence before it goes in a pack.
Is the law settled?
No. Circumvention claims against scrapers are new, two district rulings pull in opposite directions, and appeals could close the door. We supply evidence; your lawyer weighs the case.
Be first through Gatecrash’s door.
- Be first in line
- One email, when we open
- No spam