Scrapefighter
Round 1 · Early access
Scrapefighter serves the opening of every article to everyone, and the rest only to visitors who behave like people. Search engines and verified AI crawlers still get the full page. Everyone else gets a login card, or a decoy.
No CAPTCHAs. No cookies. No keylogging. Scored at the edge in under a millisecond.
Gate
The first part of the article is public. The remainder arrives when the visitor's hands say human: pointer, touch, scroll and key rhythms, scored on the server, never in the browser. The card fades, the text fades in, the reader never fills a form.
Score
Send a behaviour sample, get back a human-likelihood score and the features that produced it. Use it in your own paywall, comment form, signup flow or rate limiter. Explainable by design.
Deflector
Verified search and AI crawlers receive the full article with standard paywall markup, so rankings and citations stay intact. Unverified fleets on hosting networks receive decoy content instead of yours.
Publishers
A WordPress plugin, or a switch at the edge if your site already runs on Cloudflare. Free for small sites. Your pages stay indexable, your readers stay unbothered, your text stops leaking into other people's models.
isAccessibleForFree markup for search enginesEngineers
A REST endpoint, a browser sampler under 10 KB, SDKs for JavaScript and Python, and a documented trace format if you want to train your own models on the same data we do.
The edge serves the opening sections and holds the rest. The visible share is yours to set.
Coalesced pointer samples, touch positions, scroll cadence and key classes are collected for a few seconds. Key classes only: arrow, page, space. Never the character.
Curvature, speed variation, sample timing, scroll rhythm, press durations and the network context produce a score and a verdict. The rules live on the server, not in the page.
On a pass, the remaining sections fade in block by block and the card fades out. On a refusal, the reader keeps the opening and the login card. Every decision is logged for analysis.
No puzzles, no checkboxes, no cookie banner. Reading is the test.
Addresses are stored only as keyed hashes. Keys are recorded as classes, never as text. Traces hold coordinates and timings, nothing else.
Verified crawlers are identified by published ranges and network ownership, never by user agent alone, and always get the full page.
Scrapefighter was built to defend a ten-language health publisher with more than 1,500 pages against AI scraping fleets. It runs there today, on every page, with every visit recorded for research. Now it is being packaged for everyone else.