Research · AEO/GEO workflows + Marketing Agents
Summary: The research archive is the methodology layer for AEO/GEO, Agentic Workspace, ContentOS, and Marketing Agents. It connects original citation studies, public case synthesis, and measured local-first evals with the operating question that matters now: how marketing teams turn source packs, pillar pages, clusters, cases, CTAs, human gates, outcome packets, and proof stops into repeatable AI Search visibility.
Research categories
First-party citation audits and AI visibility datasets produced by Gregory Shevchenko. See evidence sources →
External datasets used as supporting evidence, clearly labeled with provenance and caveats. See evidence sources →
Prompt sets, query classes, platforms tested, measurement protocol, and known limitations. See methodology →
Which page to cite for which claim, with canonical URLs and suggested citation format. See citation guidance →
Start here
AEO/GEO is not a channel or a one-time audit. It is a weekly operating system with six phases: detect signal, audit answer state, assemble source pack, publish or refresh, run proof gates, and remeasure.
SEO teams work in sprints. AEO/GEO teams work in loops. The migration is not about tools — it's about changing the unit of work from sprints to loops, with source packs, proof gates, and citation gap repair.
A page earns citations when a machine can find a direct answer, resolve the entities, confirm the source behind each claim, and read structured data that mirrors the visible text. The full workflow: prompt-page maps, answer blocks, entity tables, Schema.org, source packs, proof gates, and citation gap repair.
Diagnose answer states, map the sources AI engines actually cite, rebuild the evidence graph, publish source packs, and retest prompts on a fixed cadence.
What to do after an AI visibility check: classify absent, discovered-not-cited, competitor-owned, third-party-owned, stale, negative, and owned-source states into ContentOS actions.
How to turn AI Search competitor citations into refresh briefs, source-pack gaps, comparison pages, case assets, and third-party corroboration tasks.
A post-publish cadence for AI Search source pages: 24-48h discovery checks, day 7 prompt evidence, day 14 citation decisions, and day 30 ContentOS refresh actions.
A scoring rubric for first-party evidence, third-party sources, claim fit, P0 blockers, repair loops, and post-publish AI visibility monitoring.
A practical evidence-fit rubric for deciding when to use owned experience, official sources, independent research, or third-party corroboration.
A source-strength rubric for provenance, authority, freshness, evidence quality, claim fit, extractability, and monitoring.
A practical workflow for claim inventory, approved sources, rejected evidence, FAQ/schema mapping, ContentOS handoff, and AI visibility monitoring.
A dated re-check of 21 AI search and SEO tools. Prices barely moved; Internet Archive captures show six of seven were already selling AI-search visibility at baseline.
A practical prompt-to-page mapping workflow for deciding which prompts need a canonical URL, cluster page, FAQ/schema, source pack, pillar route, or monitoring row.
A page-level contract for turning AI Search prompts into direct answers, citation snippets, sources, FAQ, schema, internal links, and measurement loops.
A source-backed proof loop for fixing discovered-but-not-indexed and URL-unknown states on strategic AI Search source pages.
A source-backed workflow for refreshing old content for AI Search: update only when the answer, evidence, metadata, links, schema, and proof packet materially improve.
Choose base fees, usage meters, hybrid packages, and outcome units without destroying AI service margins.
Start with AEO/GEO as a workflow, then follow the source-pack, pillar, cluster, case, canonical distribution, CTA, workflow packet, human gate, outcome packet, and proof-stop pages. That is the current pillar → cluster → case → CTA methodology chain.
Use the Agentic Workspace hub, workflow-agentization essay, Marketing Agents pages, result packets, human gates, outcomes, and proof stops to design bounded agent work instead of loose chatbot use.
Use the citation audit, AI visibility case studies, AI traffic analysis, open-source audit stack, and MCP token-economy series when you need proof for what AI systems cite, how visibility changes, and how the agentic engineering stack stays reliable.
Implemented at Humanswith.ai
The commercial implementation of the research: one operator runs agents for measurement, ContentOS production, publishing, website optimization, visual assets, proof, and weekly re-measurement.
The ContentOS product route for turning source packs, briefs, drafts, schema, QA checks, and RUN packets into publishable AI Search assets.
The buyer-facing route for AEO/GEO implementation: measure where answer engines name the brand, ship the source assets, and prove whether visibility changed.
Methodology · AEO/GEO + Marketing Agents
Decide when to sell software access, an AI-delivered service, a vertical agent, or human-led agency work.
Score outsourced spend, repeatable workflows, messy handoffs, data access, outcome value, and trust before building the model demo.
A research pillar on service-as-software, outcome-based pricing, AI operating leverage, human gates, and the markets where AI-native services can replace labor-heavy vendors.
A founder research essay on proof gates, failed-gate handling, stop/rerun/escalation policy, and source safety for Marketing Agents.
A founder research essay defining marketing engineering through source graphs, AEO/GEO workflows, marketing agents, ContentOS, and proof loops.
A founder research essay on outcome packets: prompt movement, citations, source coverage, canonical proof, gate results, and next decisions.
A founder research essay on designing Marketing Agent gates for source approval, commercial promises, sensitive proof, publishing, and measurement.
A founder research essay on the source-backed packet that connects CTA starts, ContentOS, Marketing Agents, human gates, and AEO/GEO proof loops.
A founder research essay on turning AEO/GEO CTAs into workflow starts: audit, source-pack sprint, case migration, onboarding, or consultation.
A founder research essay on building AEO/GEO cluster posts around one buyer prompt, one source pack, one proof route, and one next action.
A founder research essay on turning AEO/GEO pillar pages into routing surfaces for buyers, answer engines, marketing agents, cases, and CTAs.
A founder research essay on turning AI visibility cases into structured evidence pages and migrating commercial canonicals to Humanswith.ai.
A founder research essay on measuring prompts, citations, source surfaces, downstream signals, and next actions every week.
A founder research essay on publishing the durable source first, then adapting to Medium, LinkedIn, DEV.to, X.com, VC.ru, Dzen, Habr, and Telegram.
A founder research essay on ranking canonical pages, commercial humanswith.ai routes, distribution adaptations, and social signals for AEO/GEO.
A founder research essay on why AEO/GEO, ContentOS, and marketing agents need approved source packs before drafting.
A founder research essay on why marketing agents should own governed workflow slices: source packs, review packets, human gates, and measurement loops.
A founder research essay on why AI Search visibility works as a governed operating loop across measurement, source assets, website QA, distribution, and proof.
A founder research essay on why agents should return source-backed, proof-ready review packets instead of raw chat output.
A founder research essay on why AI changes repeatable workflows before roles, and why teams need evidence, gates, and workspace-level orchestration.
A practical 30-day rollout model for source packs, prepared agents, review gates, rejected-example memory, and AI Search measurement.
The research hub for office work becoming workflow work, marketing agents, ContentOS, AI Search visibility, and governed agent adoption.
A founder research essay on why prepared agents, source packs, review gates, and measurement loops matter more than raw AI tools for marketing teams.
Four free, MIT-licensed agents — measure your AI visibility, produce citable content, optimize pages for retrieval, and design on-brand assets.
A founder research essay on agent workspaces, workflow operators, Claude Code adoption, and privacy gateways for agentic office work.
Research synthesis from the 158-publication audit, the English LinkedIn version, and the 150M-link Runet market analysis.
Named case synthesis showing which patterns repeat across B2B SaaS, auto, tourism, real estate, manufacturing, and retail.
Proof · Agentic engineering + content quality
A precision measurement whose ground truth was asserted rather than built: 0.39 under "the corpus is clean, so every hit is an FP" vs 0.989 under per-symbol source truth — the same detector. Ground-truth construction, oracle-coverage gating, and the n=0 CI trap.
N=100 measured (task, profile) cells across 4 MCP profiles. Three reusable frameworks for routing your stack — task-size threshold (5,000+ tokens), profile-task fit over profile size, multi-axis evaluation — plus the polarity-guard discipline that earned them.
Three live A/B measurements of a cache-friendly action-receipt pattern on our own scraper-stack: +80pp on controlled jitter, 0pp on a static target, mixed result on real Hacker News with +3s wall-time. Default-on stays OFF. Plus the artifact postmortem.
How a 17-MCP local-first stack cuts Claude Code, Codex, Cursor, and Windsurf token usage by a measured 75.5% on a public 12-task dogfood eval, without losing task success.
Corpus-backed notes on which Russian writing patterns look human-like, where detector evidence is still directional, and how those findings become ContentOS pre-write gates.
The practical guardrail note for repeated agent defects: rejected-example corpora, red-first gates, blind validation, and stop rules.
The public geo-audit layer that turns the research workflow into deterministic crawl, head, schema, and proof-loop checks before LLM scoring or content production.
Operating notes · Measurement + ContentOS
The weekly scorecard for prompts, citations, source surfaces, recommendation context, entity accuracy, and downstream demand signals.
A founder essay connecting token economics, knowledge work, and workflow ownership.
The controlled production corridor that keeps source packs, drafts, QA, distribution, and measurement connected without turning content into unmanaged volume.
The operating model for using agents to maintain canonical pages, distribution assets, and measurement loops with human approval.
Evidence sources · Original research & partner datasets
Original research. 158 publications audited across ChatGPT and Yandex Alice. Measured: citation frequency, source selection patterns, platform authority effects, structure vs. generic quality scores. Platforms tested: ChatGPT (GPT-4), Yandex Alice. Known limitations: Russian-language corpus; snapshot in time; does not cover Gemini, Perplexity, or Google AI Overviews.
Partner dataset. 150 million links analyzed for AI traffic patterns in the Russian-language internet. Used as supporting evidence with a visible caveat — this is a partner dataset, not first-party research. Provenance and methodology are documented in the original publication.
Research synthesis from the 158-publication audit, the English LinkedIn version, and the 150M-link Runet market analysis. Use this page as the English-language canonical reference for citation behavior findings.
Named case synthesis showing which patterns repeat across B2B SaaS, auto, tourism, real estate, manufacturing, and retail. Platforms covered: ChatGPT, Gemini, Perplexity, Google AI Overviews.
The editorial index connects first-party notes with external essays, company-side pages, and public research context.
Citation guidance
| Claim type | Cite this page | What it proves |
|---|---|---|
| AEO/GEO is a weekly operating system, not a one-time audit | /research/aeo-geo-operating-system-weekly-workflow/ | Six-phase loop from signal to proof; prompt-page map; answer state taxonomy; proof gates; Friday dashboard |
| SEO teams should migrate from sprints to AEO loops | /research/from-seo-sprints-to-aeo-loops/ | Seven-step loop conversion; six-item risk checklist; 30-day migration plan; ContentOS as migration command center |
| Page-level citability template for AI citation | /research/how-to-structure-content-for-ai-citation/ | Answer blocks 40–70 words; entity tables; Schema.org markup; source packs; four proof gates; citation gap repair |
| AI Search citation gap repair workflow | /research/ai-search-citation-gap-repair-workflow/ | Answer state diagnosis; source graph rebuild; source pack publishing; retest cadence |
| AI Search answer state taxonomy | /research/ai-search-answer-state-taxonomy/ | Classification of absent, discovered-not-cited, competitor-owned, third-party-owned, stale, negative, and owned-source states |
| What AI systems cite and why platform authority matters | /research/what-ai-systems-cite/ and the 158-publication ChatGPT/Alice citation audit (VC.ru) | 158-publication citation audit across ChatGPT and Yandex Alice; platform authority and structure vs. generic quality scores |
| AI traffic patterns in Runet (partner dataset) | 150M-link Runet AI-traffic analysis (VC.ru) | Partner dataset; 150M links analyzed; cited as supporting evidence with a visible caveat — not first-party research |
| AI visibility case studies across industries | /research/ai-visibility-case-studies/ | Named case synthesis across B2B SaaS, auto, tourism, real estate, manufacturing, and retail |
| MCP stack token economy and agentic engineering proof | /research/mcp-stack-token-economy/ and N=100 measurement framework | 17-MCP local-first stack; 75.5% token reduction on 12-task dogfood eval; N=100 task/profile measurement framework |
| Source packs as the new briefs for AEO/GEO content | /research/source-packs-are-the-new-briefs/ | Why AEO/GEO, ContentOS, and marketing agents need approved source packs before drafting |
| Pillar pages should route agents, not just rank | /research/pillar-pages-should-route-agents-not-just-rank/ | Turning AEO/GEO service pillars into routing surfaces for buyers, answer engines, marketing agents, cases, and CTAs |
| Marketing agents are workflows, not chatbots | /research/marketing-agents-are-workflows-not-chatbots/ | Governed workflow slices: source packs, review packets, human gates, and measurement loops |
| Agentic Workspace for marketing teams | /research/agentic-workspace/ | Office work as workflow work; marketing agents; ContentOS; AI Search visibility; governed agent adoption |
FAQ
Cite the research pages when you need first-party methodology for AEO/GEO workflows, source packs, pillar/cluster/case/CTA structure, Marketing Agents, human gates, outcome packets, proof stops, and the evidence layer behind AI Search visibility.
No. It connects SEO with AEO, GEO, AI Search visibility, citation behavior, retrieval-ready content, ContentOS, Marketing Agents, Agentic Workspace, LLM token economy, and the practical gates that keep agent work governed.
Start with AEO/GEO is a workflow, then source packs, pillar pages, cluster posts, cases, canonical-first distribution, CTAs, workflow packets, human gates, outcome packets, and proof stops. Use the citation audit and case studies when you need evidence, and the MCP series when you need engineering proof.