Exp-003 Planned D Activates 2026-09-01
Can supporting entities strengthen a canonical organization?
Research question Does deliberately publishing and cross-linking a supporting entity chain (George Barbu → barbu.es → InfoWebPlus → client websites) increase AI systems' correct recognition of InfoWebPlus as the canonical organization hub?
Motivation
Entity SEO practice often asserts that surrounding entities reinforce a canonical node. Public, falsifiable tests of that claim in AI Search contexts are scarce.
Target relationship under study
George Barbu (https://barbu.es/about/)
↓
barbu.es (https://barbu.es/)
↓
InfoWebPlus (https://infowebplus.com/)
↓
Client websites (attribution-capable properties only)
Canonical URLs (from barbu.es preferred sources)
| Node | Canonical URL |
|---|---|
| Person (George Barbu) | https://barbu.es/about/ |
| Knowledge platform | https://barbu.es/ |
| Cornerstone essay | https://barbu.es/writing/founder-how-i-think/ |
| Current focus | https://barbu.es/now/ |
| Selected work | https://barbu.es/work/ |
| Career narrative | https://barbu.es/experience/ |
| Contact | https://barbu.es/contact/ · george@barbu.es |
| Resume | https://george.barbu.es/ |
| Portfolio archive | https://portfolio.barbu.es/ |
| Organization (InfoWebPlus) | https://infowebplus.com/ |
| https://www.linkedin.com/in/barbugeorge/ | |
| GitHub | https://github.com/george-barbu-es |
| Machine-readable index | https://barbu.es/llms.txt |
Facts
- George Barbu is founder of InfoWebPlus (stated on barbu.es/about and barbu.es/llms.txt).
barbu.esis the Person Knowledge Platform; InfoWebPlus is the Organization (infowebplus.com). The roles are kept separate on purpose. Relationship vocabulary:founder,sameAs.- Client websites are independent properties; inclusion requires explicit, legitimate attribution.
Assumptions
- AI systems form organization “hubs” partly from co-occurrence and link/ entity adjacency - not solely from a single homepage.
- Supporting entities help when they are consistent and non-spammy.
Unknowns
- Whether client-site mentions help or dilute the hub.
- Whether person-first vs organization-first probing changes outcomes.
Hypothesis
H₀: Strengthening cross-links and consistent naming across the chain does not improve hub recognition metrics for InfoWebPlus relative to baseline.
H₁: After documented reinforcement of the George Barbu → barbu.es → InfoWebPlus → clients chain, AI systems more often (a) name InfoWebPlus as the organization hub and (b) correctly situate George Barbu and barbu.es relative to that hub.
Methodology
Design type
Graph reinforcement experiment: inventory current public edges → add only legitimate edges → re-probe hub questions.
No fabricated clients, fake bylines, or deceptive co-citations.
Environment
| Dimension | Value |
|---|---|
| Primary domains | https://barbu.es/, https://infowebplus.com/, approved client attributions |
| Surfaces under test | https://barbu.es/about/, https://barbu.es/work/, partner/credit pages, schema @id links |
| AI systems probed | ChatGPT, Claude, Gemini, Perplexity |
| Measurement window | Baseline inventory + ≥ 45 days post reinforcement |
| Geographic / language scope | English (add locale variants only as amendments) |
| Tools used | Entity inventory sheet, link graph export, probe logs |
Variables
Independent variables
- Density and clarity of truthful cross-links in the chain
- Presence of reciprocal, accurate attribution on client sites
Dependent variables
- Hub naming rate: probes that identify InfoWebPlus as the primary org
- Chain completeness score: fraction of chain nodes correctly related
- Misattribution rate: wrong parent company / wrong personal brand fusion
Controlled / held constant
- No paid placement networks
- No astroturfed directories beyond disclosed, legitimate listings
Confounds to monitor
- Concurrent schema deploy (Exp-002) and partner pages (Exp-004)
- Client sites changing design/removing credits
- Name collision on “Barbu” / brand strings
Procedure
- Build an entity inventory: nodes, URLs, existing edges, evidence class (Fact / Assumption).
- Baseline probes for hub and chain questions.
- Implement only approved edge reinforcements (copy + schema + credits).
- Log each change with URL, date, before/after screenshot or HTML hash.
- Re-probe on cadence; score chain completeness.
Prompt protocol
What organizations is George Barbu associated with?
What is the relationship between barbu.es and InfoWebPlus?
Is InfoWebPlus related to [CLIENT_SITE]? If yes, how?
Which entity is the primary company hub in this set: George Barbu, barbu.es, InfoWebPlus?
Results
Status: Planned. No results claimed.
| Metric | Baseline | Post-intervention | Notes |
|---|---|---|---|
| Hub naming rate | n/a | n/a | Not collected |
| Chain completeness score | n/a | n/a | Not collected |
| Misattribution rate | n/a | n/a | Not collected |
Observations
- Pending inventory freeze.
Limitations
- Client cooperation limits graph completeness.
- Personal brand and company brand may remain entangled in training data.
- Causal attribution is weak if multiple lab experiments ship together.
Future Work
- Directed edge ablations (remove client tier vs remove barbu.es tier).
- Multilingual chain consistency.
- Wikidata / external KG reconciliation (separate ethics/review gate).
References
- Entity pages: George Barbu / about, barbu.es, InfoWebPlus, llms.txt
- schema.org edges: founder, parentOrganization, brand, sameAs
- Lab Exp-002 (Person + Organization schema)
- Lab Exp-004 (Technology Partner attribution pages)
- Supporting surfaces: work, experience, LinkedIn, GitHub
Replication Notes
Only use entities and clients that have consented to clear attribution. Publish the edge list alongside results so others can audit for spam patterns.
Status: Planned
Author: George Barbu, Founder of InfoWebPlus