# Post-freeze reference search pass — 24 August 2026

## 1. Scope and denominator rule

This is an ordinary web-search continuation requested by the user; it is not a
research-assistant workflow. The supplied 42-record Zotero CSV remains the
canonical **Paper Atlas** denominator, and the existing 86-record controlled
**Reference Universe** remains the frozen course/Notes denominator for this
release. Later search results are recorded here so that discovery never
silently changes either denominator.

An item can be:

- **Already integrated** — its bounded idea or record is already present in a
  week, lab, Notes chapter, or controlled reference;
- **Mapped, not executed** — a credible implementation or dataset has a named
  future use, but has not been independently reproduced;
- **Queued for full-text review** — discovered after the controlled freeze and
  not yet allowed to carry a scientific claim;
- **Context-only / screened out** — useful for terminology, market context, or
  source discovery, but not accepted as evidence of mechanism, effect, law, or
  standard status.

## 2. High-value international additions and status changes

| Queue ID | Source | Type and verified fact | Evidence ceiling | Project disposition | Where the idea is already represented |
|---|---|---|---|---|---|
| Q-A01 | [Optimizing Visibility in Generative Engines: A Critical Survey of GEO (2023–2026)](https://arxiv.org/abs/2607.14035) | arXiv survey dated 15 July 2026; reports a 45-study review and explicitly separates discoverability, citation, absorption, and downstream outcomes | Preprint survey; secondary synthesis until study selection and every reused claim are checked against primary papers | **Queued for full-text review; not added to the 42-paper atlas and not yet cited as claim evidence** | The multistage visibility chain, evidence hierarchy, repeated measurement, and causal boundary were already designed independently in W01–W08 and the Notes; similarity is corroborative, not proof of copied or completed integration |
| Q-G01 | [SafeGEO repository](https://github.com/QianfengWen/SafeGEO) | Author repository for PAPER-39; on 24 August 2026 it reports EMNLP 2026 acceptance, Apache-2.0 code, CC BY 4.0 data, an offline sample path, and fixed-source recommendation experiments | Implementation and dataset for the named study only; repository claims still require paper/method verification | **Mapped, not executed**; use only a safe offline sample for an independent L07 reproduction after dependency and data review | W14 and L07 already teach synthetic attack cases, containment, evidence preservation, rollback, and release blocking; no live abuse procedure is imported |
| Q-V01 | [The 2026 State of GEO open datasets](https://github.com/Broadcastwell/state-of-geo-2026) | Vendor research repository with a cited Zenodo record and repeated-run/cross-engine datasets | Observational industry data; no causal mechanism or population generalization without an independent methods audit | **Context-only candidate** for W07/W13 sensitivity exercises; not absorbed into a result claim | W07, W13, and L06 already separate within-block variation, drift, engine differences, missingness, and sampling-frame limits |
| Q-C01 | [Cornell CS4740 Spring 2026 schedule](https://www.cs.cornell.edu/courses/cs4740/2026sp/schedule.html) | Official university course schedule including retrieval-augmented LLM material and linked retrieval notes | Curriculum/pacing comparator only | **Candidate teaching reference**; no scientific claim | The course already supplies prerequisite bridges, retrieval foundations, and a staged path from language-model concepts to RAG/GEO |
| Q-C02 | [UIC CS594: Vector Databases, Fall 2026](https://www.cs.uic.edu/~asudeh/teaching/archive/cs594fall26/index.html) | Official university syllabus centered on vector data systems, retrieval, RAG, indexing, and projects | Curriculum comparator; future-course information can change | **Candidate teaching reference** for the W04 technical bridge; not a current requirement | W04 already distinguishes sparse, dense, and hybrid retrieval with an executable deterministic explorer and low-compute lab route |

## 3. Existing references rechecked in this pass

| Existing ID | Source | Recheck result | Absorption boundary |
|---|---|---|---|
| S11 | [Project 20252040-Z-469 — AI: General Technical Requirements of Retrieval-Augmented Generation](https://std.samr.gov.cn/gb/search/gbDetailed?id=37FC03D2E1436322E06397BE0A0AA17F) | Official SAMR record shows **“正在批准”** and identifies it as a national standardization guidance-project registration, not a published standard | Remains **Context-only / N-watch**. It may support a standards watchlist and status-reading exercise; it cannot be presented as an in-force national standard or requirement |
| R01 | [The GEO Community resources, courses, and tutorials map](https://thegeocommunity.com/generative-engine-optimization-resources-courses-tutorials/) | Broad practitioner learning map; useful for finding papers, courses, videos, tools, and role-based paths | Already catalogued as a practitioner source. It informs resource discovery and sequencing, while primary papers/official documents carry scientific or platform claims |
| R03 | [The GEO Community explanation of the original GEO paper](https://thegeocommunity.com/blogs/generative-engine-optimization/geo-princeton-paper-original-study/) | Practitioner explanation of PAPER-16 and its simulator/visibility measures | Already catalogued; usable for teaching-language comparison only. PAPER-16 remains the primary source, and neither source establishes universal cross-platform gains |
| R07 | [Visibility Has Two Axes: Sampling Variance vs Drift](https://thegeocommunity.com/blogs/generative-engine-optimization/visibility-two-axes-repeated-sampling-variance-drift/) | Practitioner synthesis emphasizing the distinction between repeated-run variance and longitudinal drift | Already catalogued and pedagogically aligned with W07/W13. Quantitative claims must be checked against the cited primary studies before use |
| PLAT-01 | [Google’s guide to generative-AI features in Search](https://developers.google.com/search/docs/fundamentals/ai-optimization-guide) | First-party guidance for the named Google Search surface | Supports only documented Google guidance; it is not proof of ranking benefit, cross-engine behavior, indexing, or citation |

## 4. Domestic white papers and “standards”: screening rule

The search surfaced several 2026 Chinese GEO “white papers,” red books,
service-provider rankings, proposed occupation standards, and self-described
industry standards. They are **not** silently promoted into the standards
family. Examples include industry-media or vendor publications and open draft
standards without a national-standard record. Their permitted uses are limited
to vocabulary, market claims to test, stakeholder mapping, and future source
discovery.

Before any such item enters the controlled Reference Universe, it must disclose
at least the issuing body, named authors/editors, publication date and version,
method/data, conflicts of interest, stable download location, reuse licence, and
the exact legal/standardization status. “White paper,” “industry standard,”
“group standard,” “open draft,” “national standard project,” and a published
`GB`/`GB/T` record are never interchangeable.

## 5. Search-to-design decisions

1. **No new paper was inserted into the supplied 42-record atlas.** Q-A01 is a
   discovery candidate, not a back-door denominator change.
2. **No online claim was allowed to overwrite a frozen local record.** The
   SafeGEO acceptance and repository licence update are logged here pending a
   catalog-version refresh.
3. **The course design now exposes the distinctions repeatedly found in the
   search pass as executable pedagogy:** pipeline-stage boundaries,
   sparse/dense/hybrid retrieval, claim-level entailment, repeated runs versus
   time blocks, intervention equivalence, adversarial release control, and
   source/evidence-ceiling filtering.
4. **Practitioner tutorials remain valuable but bounded.** They can improve
   sequencing, examples, misconceptions, and lab prompts; primary papers,
   official platform documents, and status-verified standards carry the
   corresponding research, product, or compliance claim.
5. **Next review gate:** full-text human audit of Q-A01, version refresh for
   PAPER-39, dependency/licence audit of Q-G01, and independent reproduction of
   one safe offline fixture before any “reproduced” label appears.