What the Absence Index can compare with the literature
Only 2,608 of 8,085 named vendor-answer pairs carry a qualifying own-domain citation (Broadcastwell, 2026, September release and registered Study 2).
The September record contains 8,085 named vendor-answer pairs, of which 2,608 include a qualifying citation to the vendor’s own domain: 32.3%, descriptive 95% Wilson interval 31.2% to 33.3%. Naming and own-source citation therefore need separate labels. Absence Index release 2026-09; registered Study 2, vendor-pair summary.
That distinction is also where many comparisons with the SEO literature begin to fail. A Google result URL, a scholarly article’s citation count, a website audit score and a vendor name in an answer are different observations. Similar-sounding language about visibility does not make their denominators interchangeable.
We place the Index beside the literature to identify comparisons worth making, not to claim that our record validates every earlier finding. The result is deliberately uneven. Some parallels concern measurement practice; others stop at a shared question because the outcomes differ too much.
Which comparisons survive their definitions?
| Outside evidence | What it measured | Relevant Index observation | Boundary that remains |
|---|---|---|---|
| Lewandowski, Sünkler and Yagci, 2021 | Observable optimization signals among selected German Google result URLs | A separately registered page-signal study can describe signals on URLs referenced in archived AI answers | Different dates, source selection and classification implementation prevent a causal comparison |
| Rovira and colleagues, 2019 | Scholarly relevance ordering compared with citation-based rank summaries | Questions, engine configuration and collection dates must remain attached to a visibility table | Scholarly citations are not source links in B2B answers; the deposited median-range discrepancy must stay disclosed |
| Ziakis and colleagues, 2019 | Exploratory page-feature correlations in a small Google sample | An answer record can describe who appeared without inferring which feature caused it | Neither the small correlations nor the Index supplies a universal optimization recipe |
| Giannakoulopoulos and colleagues, 2019 | University standing beside researcher-defined website-quality and SEO rankings | Recognizable vendors can have low naming rates for selected category questions | Academic prestige, vendor recognition and product quality are not measured by one common scale |
| Jill Whalen’s 2009 practitioner interview | Historical testimony about personalized and geographically varied search results | September repeats allow a bounded disagreement calculation | Testimony is not an experiment, and repeat disagreement is not a measured user-personalization effect |
| Google’s original-research guidance | Advice for publishers | Public evidence can be organized with clear units, dates and inspectable records | Guidance does not provide a guaranteed citation or commercial return |
| The restaurant-review chapters | Selected review corpora and a designed synthesis application | The citation archive records review-platform hosts beside vendor and other sources | Restaurant source selection does not establish the objectivity of all B2B reviews |
Primary locations: Lewandowski, §5 and Figure 1; Rovira, Table 2 and §4; Ziakis, Table 2; university study, Table 9; Whalen interview, printed pp. 447 to 449; Google helpful-content guidance. The restaurant distinction is documented in the review-platform analysis.
Start with an observation, then ask what could explain it
The pair result at the top of this page is reproducible from published material. The study uses the release’s checked vendor universe, public aliases and documented matching rules. It reconciles every vendor’s named count, qualifying own-domain cited count and eligible base to the deposit. It does not invent a private definition that changes when a vendor’s result looks surprising.
The remaining named pairs may cite third parties. Some answers may mention a product without attaching a source that meets the own-domain rule. Those possibilities are quite different from a claim that the engine knows nothing about the vendor. They also do not reveal which source caused the name to appear.
If you are investigating your own company, the saved answer is the starting evidence. Read the wording and its references. A missing citation is a defined observation; an explanation about why the engine omitted it is a hypothesis unless another design tests that explanation.
Repeat disagreement supplies a measurement parallel
Across 14,192 complete vendor-question-engine groups, 1,259 are mixed across the valid repeats: 8.9%, descriptive 95% Wilson interval 8.4% to 9.4%. A group is mixed when the checked vendor is named in some repeats and absent in others. Incomplete groups are omitted from that particular measure and counted in its audit. September engine_presence counts; repeat-disagreement CSV.
Whalen called ranking “an impossible measurement” in the context of personalized and geographically targeted results. Her statement is dated testimony, not a number we have validated experimentally. The September calculation contributes something different: a precisely defined disagreement rate for the captured questions, checked vendors and engines. [Ledford, 2009, Whalen interview, printed pp. 447 to 449].
The two meet at the need to specify a measurement. They do not meet at a common estimated effect. Repeats in our archive do not establish how much personalization matters, how a different account would behave or how another location would alter the answer. Uniform absence also counts as agreement, which keeps the statistic from becoming an engine-quality league table.
A dated pair of releases is not evidence of persistence
The August and September records share vendor identities in CRM and project management, but cover different engine configurations and valid-answer bases. The target question texts for those overlapping categories match; engine coverage, control scope and the archived collection conditions still need to remain visible. August release README; September release README.
Rovira’s historical Web of Science observation has a related caution. The paper proposes a possible operator test but calls that explanation “only an inference.” We preserve the observed dates and the proposed explanation separately. Likewise, our side-by-side vendor counts do not become a longitudinal effect merely because the product labels match. [Rovira et al., 2019, §4].
The registered persistence protocol waits for the required additional comparable release and stable configuration. It publishes dated observations while holding the persistence claim. There is a real cost to that restraint: you do not get an immediate persistence headline. The available records cannot supply one honestly.
Familiar names do not make the samples comparable
Our named low-presence examples include Microsoft Dynamics 365, 36/150 valid CRM answers, 24.0%, descriptive 95% Wilson interval 17.9% to 31.4%. That is a dated category result. It is not a measurement of Microsoft’s commercial strength, product quality or general awareness. September CRM vendor table.
The university study similarly should not be converted into a general law that prestige cannot influence discovery. Its SEO-performance measure is a researcher-defined website composite, not observed AI inclusion. Moreover, its deposited rank correlation and significance do not reproduce the printed sign and test value. The source correction preserves those discrepancies rather than selecting the version that best fits our argument.
Putting these records beside one another can help you ask whether a familiar name is being substituted for an actual observation. It cannot establish that universities and software suppliers experience the same mechanism. Their populations, definitions and outcomes remain separate.
What the registered citation analysis adds
The citation study makes a new descriptive table from the explicit source arrays. It retains empty answers in answer-level bases, reports URL occurrences separately and records host-purpose uncertainty. The large other group is conspicuous; that is preferable to presenting an incomplete publisher inventory as a complete trust map.
Its frozen primary classification and later corrected sensitivity are both public. A classification correction can change a review-platform share without changing a single source URL. The study page supplies the timing amendment, exact rules, complete engine/category tables and reproducibility files.
The answer record also cannot establish what an engine’s training data contain. A reference is displayed output, not a record of model development. The crawler log adds a separate public-request observation, with its own window and claimed-identity caveat. A request is not a citation, and neither event alone is a buyer outcome.
Which question is still open?
A software company can use the literature to choose a better research question without turning every association into a tactic. Does the proposed change concern technical accessibility, retrieval, the facts available about the product or the way an answer cites its sources? Those are different places to look for evidence.
The Index supplies dated answer observations and the registered analyses supply defined secondary measurements. What remains unresolved is how much a particular publishing change would alter future answers under a stable comparison, and whether that difference would matter to buyers. None of the parallels in this table makes those missing outcomes appear.
What our own data shows
Study 2 reproduces 8,085 named vendor-answer pairs among 42,792 eligible pairs, with the matching dictionary and valid-answer bases preserved. These are counted archive observations, not a measure of causal effectiveness. The pair, repeat and named-vendor figures above come from September version DOI 10.5281/zenodo.22907695, with the exact units stated beside each count. The registered citation analysis has published reproducibility files; the persistence plan has no persistence result. The separately registered page-signal study has its publication state recorded on the research hub; this comparison does not use its result. We have no causal estimate of how changing a company’s page will alter its visibility.
Sources
- Broadcastwell. (2026). Absence Index, release 2026-09 [Data set]. Zenodo. Release JSON, answer JSONL and README. https://doi.org/10.5281/zenodo.22907695 Verified 30 October 2026.
- Broadcastwell. (2026). Absence Index, release 2026-08.1 [Data set]. Zenodo. Release files and README. https://doi.org/10.5281/zenodo.22721962 Verified 30 October 2026.
- Lewandowski, D., Sünkler, S., & Yagci, N. (2021). The influence of search engine optimization on Google's results: A multi-dimensional approach for detecting SEO. In WebSci '21 (pp. 12 to 20). ACM. doi:10.1145/3447535.3462479 . Accessible original Verified 30 October 2026.
- Rovira, C., Codina, L., Guerrero-Solé, F., & Lopezosa, C. (2019). Ranking by relevance and citation counts, a comparative study: Google Scholar, Microsoft Academic, WoS and Scopus. Future Internet, 11(9), 202. doi:10.3390/fi11090202 . Accessible original Verified 30 October 2026.
- Ziakis, C., Vlachopoulou, M., Kyrkoudis, T., & Karagkiozidou, M. (2019). Important factors for improving Google search rank. Future Internet, 11(2), 32. doi:10.3390/fi11020032 . Accessible original Verified 30 October 2026.
- Giannakoulopoulos, A., Konstantinou, N., Koutsompolis, D., Pergantis, M., & Varlamis, I. (2019). Academic excellence, website quality, SEO performance: Is there a correlation? Future Internet, 11(11), 242. doi:10.3390/fi11110242 . Accessible original Verified 30 October 2026.
- Ledford, J. L. (2009). SEO: Search engine optimization bible (2nd ed.). Wiley. Jill Whalen interview, printed pp. 447 to 449 (PDF pp. 469 to 471). ISBN 978-0-470-45264-6. https://books.google.com/books?vid=ISBN9780470452646 Verified 30 October 2026.
- Google Search Central. (2026). Creating helpful, reliable, people-first content. https://developers.google.com/search/docs/fundamentals/creating-helpful-content Verified 30 October 2026.
- Broadcastwell. (2026). Who do AI engines cite for B2B software questions? [Registration]. OSF. Registered 30 October 2026, 01:32:03 UTC. https://doi.org/10.17605/OSF.IO/VCZKT Verified 30 October 2026.
- Broadcastwell. (2026). Does absence persist from one release to the next? [Registration]. OSF. Registered 30 October 2026, 01:12:44 UTC. https://doi.org/10.17605/OSF.IO/MKJ9V Verified 30 October 2026.
Cite this page
Broadcastwell research team. (2026, October 30). What the Absence Index can compare with the literature. Broadcastwell. https://docs.broadcastwell.com/research/insights/index-against-the-literature/ Published 30 October 2026; updated 30 October 2026.
@online{broadcastwell2026research50,
author = {{Broadcastwell research team}},
title = {What the Absence Index can compare with the literature},
year = {2026},
date = {2026-10-30},
url = {https://docs.broadcastwell.com/research/insights/index-against-the-literature/},
urldate = {2026-10-30},
note = {Published 2026-10-30; updated 2026-10-30}
}
Prepared by the Broadcastwell research team. Lead: Vikram Patel, AI architect and specialist.
Sources checked 30 October 2026. Independent check: 30 October 2026, Broadcastwell verification desk. This page has not been peer reviewed.
Last updated: 30 October 2026. Next review: 30 January 2027.