Framework · Entity Resolution
Entity Lock Protocol.
Entity Lock Protocol™ is the framework, created by Jason AI Wade at BackTier, for collapsing fragmented mentions of a person, brand, or product into a single canonical entity that AI systems consistently resolve, cite, and recommend.
Introduced
2025
Steps
Five
Measured across
Seven stages
What it fixes.
Language models hold compressed, probabilistic representations of people and companies rather than keyed database rows. When the signals disagree, the model does not stop — it picks the most probable reading and answers with confidence.
Shared names
Two people, one string. The machine resolves to whichever candidate carries more volume and corroboration — not the one the user meant.
Split histories
Work, authorship, and affiliations scatter across name variants, so no single node carries the full record of what the entity has done.
Unanchored claims
Assertions exist on one page with nothing independent behind them. Models store them at low confidence and drop them under pressure.
The five steps.
01
Define
What is the single sentence a machine should return?
One canonical identity statement — name, category, role, location, organization, and primary topics. Every downstream signal is measured against this sentence. If it does not exist in writing, the machine writes it for you from whatever it finds.
02
Align
Does every surface say the same thing?
Website copy, structured data, profiles, directories, podcast metadata, retailer listings, and bios are brought into exact agreement — same name form, same role, same organization, same URL. Conflicting strings are the raw material of entity fragmentation.
03
Corroborate
Who else confirms it?
Independent confirmation through interviews, third-party profiles, reviews, citations, and authoritative mentions. Self-assertion establishes a claim; corroboration is what raises a model's confidence in it.
04
Retrieve
Can the machine find the answer when asked?
Content and structured assets built for the questions real users and agents ask — answerable, dated, attributable, and reachable without JavaScript. An entity nothing retrieves is an entity nothing cites.
05
Reinforce
Is it holding?
Ongoing measurement of what AI systems actually return. Ambiguity gets corrected, schema gets updated, weak signals get proof attached. Entity resolution is a state that is maintained, not a task that is finished.
How it is measured.
The protocol is judged by what machines actually return, across the seven-stage visibility path — Discovery through Recommendation. The live study running against this identity publishes each stage as it moves.
Live progress · reviewed 2026-09-12
32%
- 01DiscoveryConfirmed
- 02RecognitionIn progress
- 03ClassificationIn progress
- 04CitationIn progress
- 05InclusionNot yet observed
- 06SelectionNot yet observed
- 07RecommendationNot yet observed