Trust Publishing Institute examines dual publishing on MedicarePlans.com using WebMEM and the Data-to-Action Hierarchy for Answer Engines, a model developed by David W. Bynon. The field study asks whether publishers should provide resolution-ready knowledge for AI alongside human-facing content. Lear more at: https://trustpublishing.org/html-structured-memory/publishing-knowledge-alongside-content/
What if Answer Engine Optimization is still optimizing the wrong thing?
Not because AEO doesn't matter. It does.
But almost everything we're doing in AEO starts with the same assumption we've had since the beginning of the web:
The page is the publication.
We write a webpage for people. Then we help an AI system find it, parse it, extract information from it, infer the relationships inside it, figure out the context—and hopefully produce the right answer.
But there's a strange problem with that model.
In many organizations, the knowledge existed before the page did.
Think about Medicare.
Before somebody writes a sentence about Medicare Advantage plans in Arizona, the underlying information already exists as plans, contracts, counties, benefits, premiums, enrollment, Star Ratings, identifiers, relationships, and government datasets.
We take all that structured knowledge, turn it into a webpage for a human being...
...and then ask the machine to reverse-engineer the knowledge from the webpage.
Why?
That's the question behind a new Trust Publishing Institute field study on MedicarePlans.com.
We're testing a different architecture:
One web surface. Two publications.
The human gets the explanation.
The machine gets the underlying knowledge.
Here's a real example.
A human-facing page can say:
“There are 16 standard Medicare Advantage plans available in Mohave County, Arizona, for 2026.”
Perfectly good sentence.
But the fact is bigger than the sentence.
What does “Medicare Advantage” mean here?
Are Special Needs Plans included?
Which 16 plans?
What year applies?
What county?
What source establishes the count?
On the experimental machine publication, those boundaries are explicit.
2026. Mohave County. Standard Medicare Advantage. SNPs excluded. Sixteen plans. Eleven PPOs. Five HMOs.
And then there's a separate membership index identifying the 16 individual CMS Plan IDs that actually make the number 16 true.
That's the difference we're interested in.
The machine isn't being asked to infer the structure behind the sentence.
The structure is published.
The study uses something we call the Data-to-Action Hierarchy for Answer Engines to describe what happens next.
It moves through seven levels:
Strings. Things. Facts. Relationships. Context. Resolution. Action.
And that gives us a very useful boundary.
The publisher doesn't need to answer the user's question for the machine.
The publisher's job can end earlier.
Publish the entities. Publish the facts. Publish the relationships. Publish the scope. Publish the membership. Publish the source and version.
In other words:
Publish resolution-ready knowledge.
Then stop.
The answer engine still has to reason.
It has to decide what applies. It has to evaluate context. It has to resolve ambiguity. And ultimately, it has to produce the answer.
Then a person—or increasingly, an autonomous agent—can decide what to do with it.
This isn't an argument that AEO is obsolete.
It's asking whether AEO is solving only one part of the problem.
AEO says: make the page easier for machines to extract from.
Dual publishing asks: why make the machine reconstruct knowledge the publisher already has?
MedicarePlans.com is now the live field environment where Trust Publishing Institute is testing that question. The machine-facing implementation uses WebMEM, a structured publishing protocol developed by David W. Bynon.
But this isn't a claim that WebMEM causes rankings, citations, or AI visibility.
We don't know that.
The study is observational.
We're watching crawling, indexing, search visibility, answer-engine citations, source selection, and query resolution over time.
The bigger question is architectural.
The web taught us to publish documents.
Search taught us to make those documents discoverable.
AEO is teaching us to make those documents easier for machines to consume.
But maybe the next step is publishing the knowledge itself.
That's what we're testing.
And if the answer turns out to be yes, the next era of publishing may not be about choosing between content for people and data for machines.
It may be about publishing both. Trust Publishing Institute City: Bullhead City Address: 1800 Club House Drive #93 Website: https://trustpublishing.org/