Human response
Representative human requests must receive the intended human page.
Bifurcation · How It Works
How DualWeb routes and synchronizes content.
The Bifurcation production contract requires that a human request and AI crawler request follow separate delivery paths. A conforming installation must separate request classification from content synchronization and response delivery so each boundary can be tested before public claims are approved.
The request flow
- 01
Receive the request
Preserve the requested page identity, path, query behavior, and the evidence needed for classification.
- 02
Classify conservatively
Select a supported crawler class only when the configured signals pass; unknown traffic must follow the documented default.
- 03
Select an isolated representation
Use cache and variation rules that prevent responses crossing between human and AI traffic classes.
- 04
Return or fall back safely
The installation must deliver the synchronized representation only when healthy; otherwise it must use the approved fallback and record the failure.
The synchronization flow
- 01
Discover approved source pages
Start from the authorized public scope and stable canonical page identity.
- 02
Transform for extraction
Preserve facts while presenting headings, passages, lists, links, and structured data in server-readable HTML.
- 03
Validate parity
Compare identity, material facts, commercial terms, canonical references, and content coverage.
- 04
Publish with freshness evidence
Record successful synchronization and make stale or failed states visible to operators.
What the approved installation suite must verify
Crawler response
Each supported classification must receive the intended synchronized representation.
Unknown response
Unclassified traffic must follow the documented default.
Cache isolation
The suite must prove that traffic classes cannot receive each other's cached content.
Failure fallback
Routing and synchronization failures must follow the approved safe path.
Parity and freshness
Both representations must preserve page identity and core facts within the documented freshness target.
Treat crawler delivery as production infrastructure.
A marketing-page change is not a substitute for classification, cache-isolation, synchronization, parity, and fallback tests against the operational environment.
Discuss the delivery stack