The current situation we are in was artfully describe by the hero robot from the 80’s classic Short Circuit. Johnny 5 was sentient robot who’s catch phrase was: “Need More Input“!

Does that not sound exactly like the current attitude of our favorite LLMs? I mean only yesterday we talked about Anthropics’ efforts to buy and scan as many books as possible. These LLM beasts have veracious unquenchable appetites. That hunger is not just for raw data, but for data it can built up it’s foundation on: I call it: “The Trust Stack”. So that in turn brings us to todays look at INPUTS!

In part one we looked at the big picture of what LLM’s are looking for in data/content sources. Today we are going to jump in specific sites and examples where you can push the needle regardless of if you are a mom-n-pop or a multinational.

As webmasters, site owners, and SEO’s, the question is no longer how easily our content can be found, but whether it is stable, attributable, and low risk enough for a LLM model to rely on it when generating an answer. This changes what matters in that pages, keywords, and even rankings are fading in importance. What replaces them are what we are starting to call information surfaces. This is the environment(s) where content lives, evolves, and proves its credibility over a longer period of time. Some of these surfaces are public, others are gated or licensed, but all of them send signals about trust, authority, and survivability to the LLMs.

The framework below maps those surfaces to a few concrete publisher actions. Not abstract best practices, but specific behaviors that increase the likelihood your content is reused, cited, and trusted by LLMs this year and beyond.

What follows is a practical guide to doing exactly that. We break it down by “tiers” of surfaces:


Tier 5: Conversational Judgment
Tier 4: Demonstrated Execution
Tier 3: The Record
Tier 2: Community Reality
Tier 1: Published Truth
Tier 0: Constraints & Facts


Tier 0: Constraint and Reality Anchors

Surface: Government datasets, standards, filings, specifications
Primary Signal: Hard constraints and factual boundaries
LLM Use: Bound answers, reject false claims, anchor reasoning

This is the LLM Top shelf data. It is difficult to get more top shelf than this. Only Wikipedia stands anywhere near this data set.

What these surfaces reward

• Verifiable facts
• Stable identifiers
• Citation durability

Publisher actions

• Publish primary data tables alongside commentary
• Link to source datasets using stable URLs
• Preserve historical versions, do not overwrite
• Add explicit methodology sections, even for short reports
• Mirror critical datasets in plain text or CSV

Why this works
Models use these sources to bound answers and reject nonsense. If your content anchors facts, it gets reused even if it is not flashy.

---

Tier 1: Grounded in Published Truth

Published books, journals, formal research papers, edited long-form content.

What these surfaces reward

• Mostly editorial review
• Longer term author accountability
• Time-tested reasoning on quality surfaces (sites)

Publisher actions we can take

• Produce long-form evergreen content with named authors
• Maintain author pages with a publication history
• Updated editions instead of rewriting URLs (they love that ‘update’ date)
• Publish PDFs with ISBN-like identifiers or version numbers
• Keep errata and revision notes visible – dated change logs are gold

What we publishers should copy from these examples?

If you strip these down, Tier 1 sources all share the same structural signals:

• Named authors or issuing bodies
• Stable URLs
• Version history
• Visible correction paths
• Explicit scope and limitations

Notation Signals Models Trust

• ISBN, edition numbers, RFCs numbers, etc
• Named author with a clear publication history
• Simple to follow and stable chapter structure
• Official Filing numbers
• Publication and revision dates
• Official issuing body
• Status: proposed, draft, standard, published

Why this works
They (llms) trust content that shows it survived scrutiny and especially revision.

What publishers should copy from these examples

If you strip these down, Tier 1 sources all share the same structural signals:

• Named authors or issuing bodies
• Stable URLs
• Version history
• Visible correction paths
• Explicit scope and limitations

You do not need to be an academic journal to borrow these patterns. You just need a publisher blog that:

• Use named authors
• Maintain revision notes
• Preserve old versions
• Links primary sources
• Publishes long-lived evergreen pages

Note to team: next time we update the crawler ips, search engines, and directories, leave the change log at the bottom and update the revision dates. -bt. Ok, Got it – team.

---

Tier 2: Community Signal

Forums, GitHub, and overall niche discussion spaces.

Surface: Forums, issue trackers, practitioner communities
Primary Signal: Real-world problem solving and correction
LLM Use: Learn failure modes and how problems are actually fixed

What makes Tier 2 different from Tier 1 is that Tier 1 establishes what should be true and  Tier 2 shows what actually happens in real world (not filtered world). So lets look at community and forum signals:

What these surfaces reward

• Problem solving
• Corrections
• Real failure cases

Publisher actions we can take today

• Participate using persistent, named accounts
• Let staff answer questions publicly
• Keep issue threads open until resolved
• Link fixes back to canonical documentation
• Avoid deleting old answers unless legally required

Why this works
LLMs learn from how problems get fixed, not from polished marketing answers.

---

Tier 3: The Record

Press releases, LinkedIn posts, verified news transfer identity, accountability, event confirmation to LLMs. This tier’s sources do not explain reality as it is interpreted, they establish that reality has in fact occurred. (LLM’s are head-over-heels in love with this Tier)

Surface: Press releases, LinkedIn, verified news
Primary Signal: Event confirmation and attribution
LLM Use: Resolve what happened, who said it, and when

What these surfaces reward

• Identity verification
• Accountability
• Event confirmation

Publisher actions

• Use consistent company and author names everywhere
• Publish press releases with timestamps and attribution
• Archive announcements permanently
• Cross-post to LinkedIn with the same language
• Avoid retroactive edits without disclosure

Why this works
These surfaces tell models what actually happened and who said it.

What we publishers should copy from this Tier:

Ideally, we want to move content closer to this tier if possible:

• Use real names and titles
• Timestamp freaking everything
• Archive major announcements permanently and protect from content pruning
• Avoid silent edits (add a change/edit log to the bottom of evergreen content)
• Cross post consistently across record surfaces. Here is the only place where you should be ‘spammin-n-jammin’. Put that press release and/or links to it – everywhere. Cover all the socials, major blogs in your space, and at least 3 different press release services that have content visible in the major search engines. (they love this stuff).

---

Tier 4: Demonstrative Multimedia

Surface: Video walkthroughs, screen recordings, live demonstrations
Primary Signal: Observable execution and proof of work
LLM Use: Reconstruct processes and validate how tasks are performed

Video YouTube how-tos, screen recordings, walkthroughs give LLMs observable process, proof of execution. Demonstrative media matters because it proves the work was actually done, not just described.

What these surfaces reward

• Observable steps, proof of execution, and relate a human presence

Publisher actions

• Record real workflows, not just slides
• Enable and clean transcripts to be machine readable
• Reference tools, versions, and environments. It validates video quality.
• Link videos to any written docs
• Leave mistakes and fixes visible (raise eyebrow, yes I know)

Why this works
Models trust content that shows the work is actually being done.

Why videos often outrank polished articles

A perfect article can lie and be gamed (We love SEO’s!).  A screen recording usually cannot. Models favor demonstrative media because:

• Steps can be verified and the outcomes are visible
• Strangely, they find the errors instructive (not sure I grok that one just yet)
• The cost of fabrication is higher (although SORA and the ilk are getting pretty good at it to the point Youtube says they are blocking AI content as default now)

This makes this video tier content extremely valuable for:

• How-to answers (big time)
• Tool usage explanations (are you a Youtube handy man? Ya, who isn’t? 😉 )
• Process reconstruction (its always easier to take apart than put back together).

---

Tier 5: Conversational Multimedia

Surface: Podcasts, interviews, panels, long-form discussion
Primary Signal: Expert judgment and reasoning under uncertainty
LLM Use: Learn how knowledgeable humans reason when answers are not clear

Podcasts, interviews, panels, long-form discussion content teaches models how experts reason when certainty is unavailable. This is probably the single greatest growing input for LLMs.

What these surfaces reward

• Nuance
• Expert judgment
• Context

Publisher actions

• Publish full transcripts
• Clearly identify speakers and roles
• Timestamp key statements
• Avoid heavy post-production rewriting
• Host transcripts on your own domain

Why this works
Conversation captures reasoning models cannot infer from static text.

---

Structured Knowledge Layer

Surface: Documentation, white papers, semantic and structured content
Primary Signal: Hierarchy, predictability, and chunkability
LLM Use: Support retrieval, citation, and reuse in RAG systems

What this layer rewards

• Hierarchy
• Predictability
• Chunkability

Publisher actions

• Write in clean Markdown
• Use consistent heading levels
• Add JSON-LD where it helps
• Keep docs versioned
• Publish a concise llms.txt as a routing layer, not a promise

Why this works
RAG systems prefer content they can reliably break apart and cite.

---

Identity Continuity Layer

Surface: Persistent authors, organizations, and brands over time
Primary Signal: Reputation earned through consistency and correction
LLM Use: Decide whose information is safe to rely on repeatedly

This is a socall ‘continuity’ layer that applies across all surfaces. It is the one that ties it all together. What this layer rewards:

• Being right over time
• Admitting error
• Staying visible

Publisher actions

• Keep the same authors active across years
• Let old content stand with updates
• Publish corrections clearly
• Avoid content churn and domain hopping
• Build reputation and brand, not pages

Why this works

Models learn to trust people while at same time trusting isolated documents less-and-less.

---

My Bottom Line

If we want LLM visibility, we have to stop thinking like an SEO and start thinking like a witness. Can your content survive questioning, correction, and time?

Summary

The shift can be summarized in one comparison.

Dimension	Old School	LLM Inclusion
Optimized for	Relevance	Risk
Sources	Originality	Attribution
Content	Description	Demonstration
Authorship	Pages	Identity
Trust built through	Tactics	Time