Appendix E. References and research provenance
The three source volumes
| Volume | Title | Version DOI |
|---|---|---|
| Volume I | The 2026 State of Generative Engine Optimization, v1.0 | 10.5281/zenodo.21537014 |
| Volume II | The Absence Ladder. The 2026 State of GEO, v2.0 | 10.5281/zenodo.21586091 |
| Volume III | Divergence Survives a Length Control on Google AI Overviews and Vanishes on Claude. The 2026 State of GEO, v3.0 | 10.5281/zenodo.21789120 |
Measured. All three are published under CC BY 4.0 with their data and analysis code at github.com/Broadcastwell/state-of-geo-2026. Zenodo also issues a concept identifier covering all versions of the series, which resolves to the latest version. It is not a volume and it is cited nowhere in this manual.
Measured. The Volume II paper, The Absence Ladder, is published in full as an open PDF at The Absence Ladder. The 2026 State of GEO, Volume II. It carries the same figures, method statement and disclaimers as the record cited above.
Cite Volume III as: Sivakumar, S. (2026). Divergence Survives a Length Control on Google AI Overviews and Vanishes on Claude. Measuring Vendor-Set Agreement Across Four AI Search Engines. The 2026 State of Generative Engine Optimization, v3.0. Zenodo.
External work cited in this manual
Reported. The entries below are reproduced from Volume III's reference list, which states that every reference was checked against the arXiv listing or the publisher page before publication, and that author lists are as printed by the source. This manual has not independently re-verified them.
Aggarwal, P., Murahari, V., Rajpurohit, T., Kalyan, A., Narasimhan, K., Deshpande, A. (2024). GEO: Generative Engine Optimization. KDD 2024. arXiv:2311.09735.
Bagga, P. S., Farias, V. F., Korkotashvili, T., Peng, T., Wu, Y. (2025). E-GEO: A Testbed for Generative Engine Optimization in E-Commerce. arXiv:2511.20867.
BrightEdge (2026). Why AI Engines Cite Different Sources but Recommend the Same Brands. Weekly AI and Search Insights, 24 April 2026.
Chu, X., Hou, Y. (2026). Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems. arXiv:2606.17443.
DerivateX (2026). ChatGPT and Google AI Overviews agree on tools, not sources. Open benchmark dataset, June 2026.
Harsel, L., Yudina, A., Skopec, C. (2026). AI Overviews are expanding across commercial intent search. Semrush, 2 July 2026.
Jack, W., Lehman, N., Maloney, K., Xu, S. (2026). Prominence-Stratified Failure Modes in Retrieval-Augmented Commercial Recommendation: A 37,000-Run Audit. arXiv:2605.27439.
Jack, W., Lehman, N., Maloney, K., Xu, S. (2026). Divergent Recommendations, Convergent Diagnoses: Cross-Provider Failure-Mode Convergence in AI Commercial Recommendation. arXiv:2606.26116.
Khromova, Y. (2025, updated 2026). ChatGPT vs Perplexity vs Google vs Bing: AI Search Engine Comparison. SE Ranking. Data collected 26 February to 3 March 2025.
Lafferty, N. (2025). AI Platform Citation Patterns. Profound.
McDonald, T., Cooley, H., Williams, M. (2026). AIO Impact on Google CTR: 2026 Update. Seer Interactive, 24 April 2026. Data through February 2026.
Schulte, J., Bleeker, M., Kaufmann, P. (2026). Don't Measure Once: Measuring Visibility in AI Search (GEO). arXiv:2604.07585.
Zatuchin, D. (2026). Who Owns the AI Recommendation? arXiv:2606.23057. Dataset released on Zenodo under CC BY 4.0.
Reported. Volume III notes of Jack and colleagues (arXiv:2606.26116) that the same-prompt rerun baseline quoted in that paper is cited by it from a separate industry report rather than measured within it. This manual repeats that qualification wherever the work is used.
Open flags
Reasoned. FLAGS.md at the root of this manual's repository records six open items where a published source is incomplete or two published sources disagree. They are disclosures rather than defects. Where a chapter touches one, it quotes the source as published, asserts no relation between conflicting figures, and does not attempt reconciliation. The current items concern Volume III's answer counts, Volume I's answer and question counts, a duration statement in Volume III, the location of the unique-domain figure, the date on the published self-audit, and the run structure of the August re-measure.
Dated research provenance
Measured, collection design. The July 2026 collection used four engines, ten questions and five runs per question. The July date comes from the published citation metadata rather than the disclosure sentence itself; that distinction remains recorded as an open flag in the source repository. The 18 August 2026 collection used the same ten questions and four engines, with one run per question. The published source files retain the dated observations and collection details.
Reasoned. These designs are not directly comparable because the run count differs. Read the original records with their dates and denominators attached. A source citation is evidence that an answer used a page, not automatically that it recommended the publisher. Chapter 3 explains how to distinguish those outcomes by inspecting the answer itself. The operator disclosure below identifies Broadcastwell's commercial interest, while the open files allow readers to check the measurement independently.
Licence and status
Measured. Prose and figures in this manual are published under CC BY 4.0. The site code is MIT. Nothing here has been through academic peer review: it is measurement published with the data and code that produced it. The standing invitation is to recompute any figure you doubt from the public files and publish what you get, including a result that contradicts one of these.
Attribution
Sivakumar, S. (2026). The Absence Manual. Broadcastwell. docs.broadcastwell.com
Sources
The volume identifiers and citation formats are from the published Zenodo records and the source repository. The external reference list is reproduced from Volume III. The dated collection records and their run structures remain available in the published source files. Everything is at github.com/Broadcastwell/state-of-geo-2026.
About this manual
Author. Sairam Sivakumar, Broadcastwell.
Operator disclosure. Broadcastwell ran this measurement and sells services in the category it measures. Broadcastwell is excluded from the measured sample and from every ranking. The mitigation is not that the conflict is absent, it is that the raw data and the code are public and the result can be recomputed by anyone who disagrees.
Historical research record. The dated collection designs and source files remain available in the published research repository and the three volumes linked below. The July 2026 collection used four engines and five runs per question; the 18 August 2026 collection used the same ten questions and four engines, with one run per question.
Reading dated measurements. Reasoned. Compare like with like. A change in the number of runs changes the evidence base, even when the questions and engines are unchanged. Keep the collection date, question wording, engine, run count and exclusions beside any result. Inspect the retained answers and sources before interpreting a difference as movement. Naming and citation answer different questions, so report them separately. A dated observation supports a claim about that collection; it does not establish a current result or prove that an intervention caused a change.
Not peer reviewed. This is an independent industry study published as an open dataset with the analysis code that produced every figure in it. It has not been through academic peer review. Read it as measurement, and check the measurement. If you disagree with a number here, recompute it from the public data and publish what you get.
Licence. Prose and figures CC BY 4.0. Site code MIT.
The three volumes.
| Volume | What it covers | DOI |
|---|---|---|
| Volume I | 85 companies, 61 categories, 860 scored answers and 5,160 citations, one engine held constant. The dataset README additionally records 1,753 unique domains cited | 10.5281/zenodo.21537014 |
| Volume II | The Absence Ladder. All 616 absence records classified by question shape | 10.5281/zenodo.21586091 |
| Volume III | Cross-engine divergence. 280 questions, 40 categories, four engines, 853 answers | 10.5281/zenodo.21789120 |
Data and analysis code for all three volumes: github.com/Broadcastwell/state-of-geo-2026.
Version 1.0, August 2026.