Definition
What is an AI Search source hierarchy?
An AI Search source hierarchy is the publishing order that tells a team which source should be treated as primary, which sources should support it, and which surfaces should be distribution only. It is the answer to a simple operational question: if ChatGPT, Claude, Perplexity, Gemini, Google AI Overviews, Copilot, Grok, DeepSeek, Kimi, Yandex Neuro, Alice, or GigaChat sees five versions of the same claim, which one should be easiest to trust?
In ordinary SEO, teams often think in pages and rankings. In AEO/GEO work, the source graph matters just as much. A page can be indexed and still be a weak answer source if it has no evidence, no byline, no internal support, no visible source trail, and no durable update path.
The hierarchy is not a theory exercise. It decides where a service page lives, where a case lives, where a methodology essay lives, where a social thread points, and what a marketing agent is allowed to adapt.
The order
Rank source surfaces before you distribute
The useful hierarchy has five layers. The exact order can change by company, but the principle should not: the most durable, complete, source-backed page sits closest to the canonical answer; the fastest, shortest, and most conversational surfaces sit farther away.
| Layer | Best source | Job in AI Search |
|---|---|---|
| 1. Canonical source | First-party page on gregshevchenko.com or Humanswith.ai | Carry the durable claim, source trail, structured data, internal links, and update history. |
| 2. Evidence page | Research page, source pack, case, benchmark, FAQ, or methodology note | Give assistants quotable details and buyers a way to verify the claim. |
| 3. Commercial route | Service, platform, agent, pricing, or case page on Humanswith.ai | Route commercial intent to the right offer and workflow start. |
| 4. Distribution adaptation | Medium, LinkedIn, DEV.to, VC.ru, Dzen, Habr, Substack, or newsletter | Reach the right audience while pointing back to the canonical source. |
| 5. Social signal | X.com, Telegram, Reddit, comments, launches, and community discussion | Create public activity, reactions, freshness, and routing signal without replacing the source. |
Canonical split
Greg site and Humanswith.ai should not own the same canonical job
The clean split is this: gregshevchenko.com owns methodology, POV, founder research, and operating models. Humanswith.ai owns commercial execution: services, platform pages, agents, cases, and CTAs. The sites should cross-link, but they should not compete for the same canonical role.
That split helps humans and AI systems. A founder research page can explain why source packs, result packets, workflow gates, and canonical-first distribution matter. A Humanswith.ai page can show how the company implements those patterns through ContentOS, Marketing Agents, Website Agentic Optimization, and AI Search visibility work.
When the topic is “how the method works,” Greg can be canonical. When the topic is “hire us, use the platform, inspect the agent, book the workflow, read the client case,” Humanswith.ai should be canonical.
ContentOS
ContentOS should enforce the hierarchy before drafting starts
A ContentOS workflow should not only write pages. It should decide source ownership before a draft exists. The source pack needs a canonical URL, approved claims, target audience, commercial route, distribution surfaces, schema requirements, and a measurement plan.
This is where agentic workspaces become practical for marketing teams. A marketing agent should not guess whether a case belongs on VC.ru, Medium, or Humanswith.ai. It should receive a source hierarchy and produce the correct artifact: canonical page, adaptation, short thread, newsletter version, source update, or proof packet.
The useful unit is a RUN packet: inputs, approved sources, channel role, required links, review gate, and expected proof. Without that packet, “publish everywhere” becomes a content mess with automation speed.
Distribution surfaces
External platforms are useful when their role is explicit
Medium, LinkedIn, DEV.to, X.com, VC.ru, Dzen, Habr, Telegram, Reddit, and newsletters can all help. The mistake is pretending they do the same job.
| Surface | AI Search value | Source-hierarchy rule |
|---|---|---|
| Medium | Long-form discoverability and founder/AI readership. | Adapt after the canonical page exists; link back visibly. |
| Professional entity signal, social proof, and buyer conversation. | Use the post to route readers toward the canonical page or commercial CTA. | |
| DEV.to | Developer and technical-operator audience for agentic workflows, open-source tools, and implementation notes. | Use only when the article has enough technical substance. |
| X.com | Fast public thesis testing, freshness, developer discussion, and activity around the entity. | Treat as a social layer, not the durable source. |
| VC.ru / Dzen / Habr | Russian-language business, broad explanatory, and technical reach. | Use by audience and language; keep the first-party canonical clear. |
X.com
X.com has value, but not the value of a canonical page
X.com is worth using for AEO/GEO work because it is easy to publish, fast to test, and visible to developer and AI-builder audiences. A good X thread can make a thesis legible, attract reactions, and create a public route to a canonical source.
But the same traits make it weak as the main source. Posts are short, conversational, and hard to maintain as complete evidence. They are better as social proof and freshness around a page than as the page itself.
The rule: canonical first, X.com second. Publish the source, then use X to compress the claim, start discussion, and link back.
Language
Russian-language distribution needs its own hierarchy
The English/global stack usually starts with Medium, LinkedIn, DEV.to for technical audiences, and X.com for fast public signal. Russian-language distribution needs a different list: VC.ru for business and marketing, Dzen for broad explanatory reach, Habr for technical content, and Telegram for community and launch routing.
That does not change the canonical rule. If the case is commercial, Humanswith.ai should own the canonical version. If the piece is methodology or POV, gregshevchenko.com can own it. VC.ru, Dzen, Habr, and Telegram become adaptations and routing surfaces, not competing sources of record.
Workflow
The practical workflow has eight decisions
| Decision | Question | Output |
|---|---|---|
| 1. Intent | Is this methodology, commercial offer, case, product, agent, or operational note? | Canonical owner: Greg site or Humanswith.ai. |
| 2. Source pack | What evidence, claims, examples, and constraints are approved? | Source pack before drafting. |
| 3. Canonical URL | Which URL should assistants and humans trust first? | Durable first-party source. |
| 4. Evidence support | Which research, cases, notes, and source pages reinforce the claim? | Internal links and source trail. |
| 5. Commercial route | What should a buyer do next? | Humanswith.ai service, platform, agent, case, or CTA link. |
| 6. Distribution role | Which platform has the right audience and format? | Adaptation plan by platform and language. |
| 7. Proof gate | Can the page be crawled, parsed, cited, and linked? | Technical QA, schema check, and internal link proof. |
| 8. Retest | Did prompts, citations, mentions, buyer routes, or cases improve? | Weekly result packet. |
FAQ
Questions this page should answer
Should methodology pages be canonical on Greg site or Humanswith.ai?
Methodology, POV, and founder research can be canonical on gregshevchenko.com. Commercial pages, platform pages, agent pages, cases, and conversion flows should be canonical on Humanswith.ai.
Should Medium or LinkedIn be canonical?
No. They can be strong distribution surfaces, but the durable source should usually be a first-party page with schema, internal links, evidence, and update ownership.
Does X.com have AEO/GEO weight?
Yes, as a fast social and activity layer, especially for developer and AI-builder audiences. It should not replace a canonical article, case, service page, or platform page.
What changes for Russian-language distribution?
The platform set changes. VC.ru, Dzen, Habr, and Telegram become more important, while the canonical-first rule stays the same.
How does ContentOS use the hierarchy?
ContentOS should encode the hierarchy in the source pack and RUN packet before drafting: canonical owner, approved claims, distribution role, required links, review gate, and proof loop.
Source trail
Internal sources and commercial routes
What AI systems cite
The citation research that makes source hierarchy operational rather than decorative.
Canonical-first distribution for AI visibility
The distribution method this page turns into a source-ranking model.
Source packs are the new briefs
The input object that keeps agents from inventing the hierarchy after drafting.
What ContentOS is and what it is not
The operating system for source packs, canonical pages, adaptations, QA, and proof packets.
Humanswith.ai ContentOS
The commercial route for teams that want the workflow implemented as a workspace capability.
Humanswith.ai Marketing Agents
The agent layer that turns source hierarchy into repeatable marketing work.
Read next