Skip to content
Broadcastwell Fix one answer, $490
Menu

GEO Claim Registry: trace the number to its source

Our 83 source checks distinguish a printed result from a general GEO promise; each entry states its source, setting and verification limit (November 2, 2026).

Observed answers on stated dates, not a ranking.

GEO (generative engine optimization), AEO and AI search visibility often travel with the same headline numbers. A repeated number may lose its original denominator. We keep the circulating claim beside the original metric and the correction. An unresolved source remains unresolved.

Download claims.csv or claims.json. The two files contain the same 83 entries checked on November 2, 2026. This is a bounded claim audit, not a representative survey of research quality or an estimate of how often misinformation circulates. Entry headings state the claim evaluated. If no independently located republication is named, we do not imply a measured circulation history or attribute that exact wording to the original author.

How do the verdicts work?

  • Supported: the stated source and setting support the wording.
  • Supported with conditions: the result needs its original sample, definition or other limit.
  • Overstated: the wording extends beyond what the source measured.
  • Unsupported: the inspected evidence does not establish the claim.
  • Unverifiable: the original, denominator or calculation needed for the claim is not available or resolved.
  • Laundered: a purported citation traces to no identifiable real source for the asserted finding. This describes the citation chain, not an author’s intent.

What our own data shows

Our registered analyses describe the September 2026 answer archive and later page observations. Their denominators differ from the supplied-document experiments and vendor cohorts below. The published report contains the two completed descriptive analyses, a separate waiting protocol and the initial operational crawler snapshot. Registration is not peer review.

Supports: the bounded source checks and dated observations specified in each entry. Suggests: which source, definition or missing original a reader should inspect next. Does not support: a general citation guarantee, a product-quality verdict, an author-intent allegation or a causal effect outside a study’s design.

Find a claim

GEO increases visibility by 40%

Verdict: overstated. Checked 2026-11-02.

Circulation: Unqualified retellings of the GEO headline.

Primary source: Aggarwal et al. (2024), GEO, KDD 2024, DOI https://doi.org/10.1145/3637528.3671900; author manuscript https://arxiv.org/html/2311.09735v3. Location: Tables 1 and 6; sections 3.1 to 3.4.

What the source establishes: Table 1 gives position-adjusted word count 19.3 to 27.2 for quotation addition, about 40.9% relative. The five-seed Table 6 gives 19.8 to 27.1, about 36.9%. Neither is organic citation probability.

Base, setting and date: 1,000-query test split; five supplied source documents per query; 2024 publication.

Correction you may quote: The GEO benchmark measured a gain in a supplied-source visibility score; it did not establish a general citation uplift.

The ICLR draft headline is 37%

Verdict: unverifiable. Checked 2026-11-02.

Circulation: Aggarwal et al., GEO.

Primary source: Aggarwal et al., GEO; retrieved arXiv versions 1 and 2, https://arxiv.org/abs/2311.09735v1 and https://arxiv.org/abs/2311.09735v2. Location: Abstracts; exact ICLR submission not recovered.

What the source establishes: Both retrieved arXiv abstracts print up to 40%. We have not recovered the exact ICLR submission needed to check the separate 37% headline. The versions cannot be silently substituted.

Base, setting and date: Version audit, not a new effect estimate.

Correction you may quote: The 37% version comparison is unverified until its exact submission is identified.

Keyword stuffing makes you 15% less likely to be cited

Verdict: unsupported. Checked 2026-11-02.

Circulation: Unqualified marketing retelling.

Primary source: Aggarwal et al. (2024), GEO, KDD 2024, DOI https://doi.org/10.1145/3637528.3671900; author manuscript https://arxiv.org/html/2311.09735v3. Location: Table 7; Appendix C.1.

What the source establishes: In the 200-example supplied-file Perplexity test, judged impression is 28.1 for stuffing versus 24.7 baseline, about 13.8% higher. Source addition is 19.0, about 23.1% lower. Position-adjusted word count is a different outcome and does decline for stuffing.

Base, setting and date: 200 test examples supplied as files; 2024 paper; judged impression and position-adjusted word count must remain separate.

Correction you may quote: The GEO paper reports metric-specific results, not a universal penalty in citation probability.

Structured data adds 40% visibility

Verdict: unsupported. Checked 2026-11-02.

Circulation: Abu Rayhan report and derivative GEO claims.

Primary source: Aggarwal et al. (2024), GEO, KDD 2024, DOI https://doi.org/10.1145/3637528.3671900; author manuscript https://arxiv.org/html/2311.09735v3. Location: Tested methods in section 2.2.2 and Table 1.

What the source establishes: The original GEO experiment does not contain a structured-data treatment. Its quotation-addition result cannot be relabeled as a schema effect.

Base, setting and date: See the source-specific setting in the finding.

Correction you may quote: The original GEO benchmark did not test a structured-data uplift.

The International Chocolate Consumption Research Group is a research source

Verdict: laundered. Checked 2026-11-02.

Circulation: Aggarwal et al. (2024), GEO, KDD 2024, DOI 10.1145/3637528.3671900.

Primary source: Aggarwal et al. (2024), GEO, KDD 2024, DOI https://doi.org/10.1145/3637528.3671900; author manuscript https://arxiv.org/html/2311.09735v3. Location: Table 4, PDF page 8.

What the source establishes: The organization appears inside an experimental source rewrite. The paper does not provide an identifiable original study for the inserted chocolate claim. A later year or reference number cannot supply that missing primary report.

Base, setting and date: One illustrated source rewrite; this audit does not infer an underlying consumption measurement.

Correction you may quote: The chocolate reference is an experimental rewrite example, not a verified research report.

Source citations outperform quotations in the original GEO table

Verdict: unsupported. Checked 2026-11-02.

Circulation: Aggarwal et al. (2024), GEO, KDD 2024, DOI 10.1145/3637528.3671900.

Primary source: Aggarwal et al. (2024), GEO, KDD 2024, DOI https://doi.org/10.1145/3637528.3671900; author manuscript https://arxiv.org/html/2311.09735v3. Location: Table 1, position-adjusted word-count overall column.

What the source establishes: Quotation addition is 27.2, statistics addition 25.2 and source addition 24.6 against 19.3 baseline: about 40.9%, 30.6% and 27.5% relative.

Base, setting and date: 1,000-query supplied-source benchmark; 2024 publication.

Correction you may quote: In GEO Table 1, quotation addition exceeds statistics addition, which exceeds source addition, on the specified score.

Authoritative tone gives no significant improvement

Verdict: supported with conditions. Checked 2026-11-02.

Circulation: GEO results text.

Primary source: Aggarwal et al. (2024), GEO, KDD 2024, DOI https://doi.org/10.1145/3637528.3671900; author manuscript https://arxiv.org/html/2311.09735v3. Location: Section 4 text; Tables 1 and 6.

What the source establishes: The text says there is no significant improvement. Table 1 subjective impression rises from 19.3 to 22.9, about 18.7%; Table 6 has different values. A positive point estimate is not itself a significance test.

Base, setting and date: See the source-specific setting in the finding.

Correction you may quote: The paper pairs a no-significant-improvement statement with positive table estimates; quote both and do not invent a significance result.

The main GEO result and its repeated-seed appendix are the same estimate

Verdict: unsupported. Checked 2026-11-02.

Circulation: Retellings that combine GEO tables.

Primary source: Aggarwal et al. (2024), GEO, KDD 2024, DOI https://doi.org/10.1145/3637528.3671900; author manuscript https://arxiv.org/html/2311.09735v3. Location: Tables 1 and 6.

What the source establishes: The printed values differ throughout the method rows. The manuscript identifies Table 6 as its five-seed results. Retain table identities rather than combining values into a new estimate.

Base, setting and date: See the source-specific setting in the finding.

Correction you may quote: GEO Tables 1 and 6 report different values; a quoted number needs its table identity.

The 70% robot statistic is verified evidence

Verdict: unverifiable. Checked 2026-11-02.

Circulation: GEO representative Statistics Addition example.

Primary source: Aggarwal et al. (2024), GEO, KDD 2024, DOI https://doi.org/10.1145/3637528.3671900; author manuscript https://arxiv.org/html/2311.09735v3. Location: Table 4, PDF page 8.

What the source establishes: The percentage is inserted in an illustrative rewrite without a checkable primary measurement. It must not become a workforce statistic.

Base, setting and date: See the source-specific setting in the finding.

Correction you may quote: An inserted example statistic is not a verified measurement of robotic involvement.

AutoGEO improves performance up to 50.99%

Verdict: overstated. Checked 2026-11-02.

Circulation: Wu, Zhong, Kim and Xiong (2025), AutoGEO, version 1 preprint, https://arxiv.org/abs/2510.11438v1.

Primary source: Wu, Zhong, Kim and Xiong (2025), AutoGEO, version 1 preprint, https://arxiv.org/abs/2510.11438v1. Location: Table 1 and sections 4 to 5.

What the source establishes: The three Overall gains over Fluency are approximately 48.1%, 47.2% and 57.7%; their mean is about 51.0%. The largest of those gains is about 57.7%, rather than 51.0%.

Base, setting and date: Three 2025 supplied-document benchmarks on one specified experimental engine; v 1 tables.

Correction you may quote: The AutoGEO headline corresponds to an average across three settings, not the maximum.

AutoGEO has an average gain of 35.99% across engines

Verdict: overstated. Checked 2026-11-02.

Circulation: Retelling recorded in the October findings brief.

Primary source: Wu, Zhong, Kim and Xiong (2025), AutoGEO, version 1 preprint, https://arxiv.org/abs/2510.11438v1. Location: Introduction and Table 1, PDF pages 2 and 7.

What the source establishes: Using the Overall columns of Table 1, the larger rewrite variant gains about 50.99% on average over Fluency across three datasets; the smaller variant gains about 20.99%. The mean of those two averages is about 35.99%. All six comparisons use the single named experimental engine in that table.

Base, setting and date: Table 1 Overall score; three datasets and two rewrite variants; version 1 preprint.

Correction you may quote: The 35.99% figure averages two rewrite variants across three benchmark settings on one experimental engine; it is not a cross-engine field effect.

The AutoGEO GEO-Bench score has one consistent printed value

Verdict: supported with conditions. Checked 2026-11-02.

Circulation: AutoGEO v 1 summary comparisons.

Primary source: Wu, Zhong, Kim and Xiong (2025), AutoGEO, version 1 preprint, https://arxiv.org/abs/2510.11438v1. Location: Tables 1 and 2 versus Table 13, PDF page 29.

What the source establishes: The main GEO-Bench result prints 34.92; the comparable row in Table 13 prints 34.81. A third attributed value, 35.30, was not found in the retrieved version and is not asserted here.

Base, setting and date: AutoGEO preprint version 1; repeated reporting of a supplied-document benchmark result.

Correction you may quote: The main and appendix tables print different scores; the third attributed value remains unverified.

AutoGEO preserves utility in every setting

Verdict: overstated. Checked 2026-11-02.

Circulation: AutoGEO v 1 section 5.2.

Primary source: Wu, Zhong, Kim and Xiong (2025), AutoGEO, version 1 preprint, https://arxiv.org/abs/2510.11438v1. Location: Table 2, Researchy-GEO, Claude precision.

What the source establishes: The displayed precision falls from 96.51 to 84.98 under the larger rewrite method, a decline of 11.53 percentage points. Other utility measures differ.

Base, setting and date: One supplied-document benchmark/engine setting; 2025 preprint version 1.

Correction you may quote: AutoGEO v 1 does not preserve every utility measure in every displayed setting.

AutoGEO rule overlap is exactly 40.00% in the stated cross-domain setting

Verdict: unsupported. Checked 2026-11-02.

Circulation: AutoGEO v 1 Figure 2.

Primary source: Wu, Zhong, Kim and Xiong (2025), AutoGEO, version 1 preprint, https://arxiv.org/abs/2510.11438v1. Location: Figure 2 b, section 5.3 and Tables 7 to 8, PDF pages 15 to 16.

What the source establishes: The printed keyword labels give 17 Researchy-GEO, 15 GEO-Bench and 14 e-commerce labels. Their first two Jaccard overlaps reproduce 88.24% and 34.78%. GEO-Bench versus e-commerce shares 7 labels in a union of 22, giving 31.82%, rather than the printed 40.00%. This is a label-set reconstruction, not a reconstruction of latent rule meaning.

Base, setting and date: Three printed keyword sets in AutoGEO preprint version 1.

Correction you may quote: The printed keyword sets reproduce two overlaps but give 31.82%, not 40.00%, for the remaining pair.

AutoGEO isolates the effect of learned rules from rewriting alone

Verdict: unsupported. Checked 2026-11-02.

Circulation: Methodological reading of AutoGEO v 1.

Primary source: Wu, Zhong, Kim and Xiong (2025), AutoGEO, version 1 preprint, https://arxiv.org/abs/2510.11438v1. Location: Sections 3 to 5 and ablations.

What the source establishes: A rewrite using the same instruction without extracted rules is not supplied as the required matched control. Its absence limits a component-specific interpretation.

Base, setting and date: See the source-specific setting in the finding.

Correction you may quote: The experiment does not isolate learned rules from an otherwise identical rewrite without those rules.

KPR and KPC are consistently defined across the reported datasets

Verdict: unsupported. Checked 2026-11-02.

Circulation: Wu, Zhong, Kim and Xiong (2025), AutoGEO, version 1 preprint, https://arxiv.org/abs/2510.11438v1.

Primary source: Wu, Zhong, Kim and Xiong (2025), AutoGEO, version 1 preprint, https://arxiv.org/abs/2510.11438v1. Location: Section 4, PDF page 6; appendix G, PDF page 25; Table 11.

What the source establishes: Section 4 says the reference-dependent measures apply only to Researchy-GEO; appendix G says only GEO-Bench. Table 11 supplies them for Researchy-GEO and marks GEO-Bench and E-commerce unavailable. The appendix sentence conflicts with the methods and table.

Base, setting and date: See the source-specific setting in the finding.

Correction you may quote: AutoGEO v 1 has a contradictory dataset-eligibility sentence; the methods and result table use Researchy-GEO.

Sentiment is 6.7 times noisier than mentions

Verdict: overstated. Checked 2026-11-02.

Circulation: Kumar 2026 abstract and discussion.

Primary source: Kumar (2026), Generative Engine Optimization at Scale, vendor preprint, https://arxiv.org/abs/2606.20065v1. Location: Sections 6.4 and 6.6, mention bands and sentiment eligibility.

What the source establishes: Section 6.6 divides 45.5% positive/negative sentiment flips by 6.8% mention cells within the 30%-to-70% naming band. These are different flip definitions and eligibility rules. The broader sometimes-named share is 22.5%, giving about 2.0 when divided into 45.5, but that arithmetic does not repair the eligibility difference.

Base, setting and date: 102-brand vendor archive; March to May 2026; no released underlying data.

Correction you may quote: The 6.7-times ratio divides differently defined and differently eligible stability summaries; it is not a calibrated noise ratio.

The GEO effectiveness experiment used 10,000 test queries

Verdict: unsupported. Checked 2026-11-02.

Circulation: Kumar 2026 description of Aggarwal.

Primary source: Aggarwal et al. (2024), GEO, KDD 2024, DOI https://doi.org/10.1145/3637528.3671900; author manuscript https://arxiv.org/html/2311.09735v3. Location: Section 3.2 and 3.3.

What the source establishes: The complete benchmark has 10,000 queries partitioned into 8,000 training, 1,000 validation and 1,000 test. The experimental test base is 1,000.

Base, setting and date: See the source-specific setting in the finding.

Correction you may quote: Ten thousand is the GEO benchmark size, not the experiment test split.

Kumar establishes causal gains from schema, bylines or cross-engine optimization

Verdict: unsupported. Checked 2026-11-02.

Circulation: Kumar 2026 discussion claims.

Primary source: Kumar (2026), Generative Engine Optimization at Scale, vendor preprint, https://arxiv.org/abs/2606.20065v1. Location: Sections 3.5 and 6.7, protocols P1 to P7.

What the source establishes: The paper describes technical audit dimensions and separately lists the prospective controlled protocol slate as not yet reported. Those proposed interventions are not completed causal experiments.

Base, setting and date: See the source-specific setting in the finding.

Correction you may quote: The paper separates observed source patterns from interventions it has not yet reported.

FID is the current Core Web Vitals interaction metric

Verdict: unsupported. Checked 2026-11-02.

Circulation: Kumar 2026 technical checklist.

Primary source: Kumar (2026), Generative Engine Optimization at Scale, vendor preprint, https://arxiv.org/abs/2606.20065v1; Google web.dev, https://web.dev/articles/vitals. Location: Kumar section 3.5; current Google Core Web Vitals page, opened 2 November 2026.

What the source establishes: The paper lists LCP, FID and CLS. The current operator page identifies LCP, INP and CLS as the Core Web Vitals metrics.

Base, setting and date: See the source-specific setting in the finding.

Correction you may quote: The current Core Web Vitals interaction metric is INP; a dated checklist listing FID needs updating.

The location audit fully reports a test proving declared coordinates have no effect on the Google engine

Verdict: overstated. Checked 2026-11-02.

Circulation: Chen and Koudas 2026 abstract and methods.

Primary source: Chen and Koudas (2026), Location, Location, Generation, preprint, https://arxiv.org/abs/2610.24081v1. Location: Section 3, E2; PDF page 2.

What the source establishes: The declared-coordinate conclusion refers to a preliminary comparison without a reported preliminary comparison design or result table. The main network experiment is reported separately.

Base, setting and date: 31,200 main answer records; 300 authored English consumer questions; October 2026.

Correction you may quote: The paper reports the network experiment; its declared-coordinate preliminary comparison is not documented at the same level.

A country change replaces most sources in every non-local question

Verdict: overstated. Checked 2026-11-02.

Circulation: Chen and Koudas 2026 wording.

Primary source: Chen and Koudas (2026), Location, Location, Generation, preprint, https://arxiv.org/abs/2610.24081v1. Location: Sections 4 and 5.

What the source establishes: The non-local source localization index is 0.23 with a stated interval 0.20 to 0.27. This index is twin overlap minus between-setting overlap, not a fraction of sources replaced.

Base, setting and date: See the source-specific setting in the finding.

Correction you may quote: The reported localization index is a difference in overlap, not the percentage of sources replaced.

Only 51.5% of answer sentences are fully supported

Verdict: supported with conditions. Checked 2026-11-02.

Circulation: Liu et al. 2023 result reused as a current universal rate.

Primary source: Liu, Zhang and Liang (2023), Evaluating Verifiability in Generative Search Engines, EMNLP Findings, DOI https://doi.org/10.18653/v1/2023.findings-emnlp.467. Location: Section 4.2; Appendix G, Table 7, printed page 7023.

What the source establishes: The average citation-recall score across four historical engines is 51.5%. It is not a 2026 audit of every answer product.

Base, setting and date: 1,450 queries per engine across 12 distributions; 2023 Bing Chat, NeevaAI, Perplexity and YouChat; one annotation per query-response pair, with 250 pairs triple-annotated for agreement. Collection late February to late March 2023. Reported averages are across engine scores, not a pooled all-sentences accuracy rate..

Correction you may quote: Liu and colleagues reported 51.5% citation recall across four engines in their 2023 evaluation.

74.5% of citations support their sentence

Verdict: supported with conditions. Checked 2026-11-02.

Circulation: Liu et al. 2023 finding.

Primary source: Liu, Zhang and Liang (2023), Evaluating Verifiability in Generative Search Engines, EMNLP Findings, DOI https://doi.org/10.18653/v1/2023.findings-emnlp.467. Location: Section 4.2; Appendix G, Table 8, printed page 7024.

What the source establishes: The paper reports 74.5% average citation precision across the four evaluated engines, a separate measure from citation recall.

Base, setting and date: 1,450 queries per engine across 12 distributions; 2023 Bing Chat, NeevaAI, Perplexity and YouChat; one annotation per query-response pair, with 250 pairs triple-annotated for agreement. Collection late February to late March 2023. Reported averages are across engine scores, not a pooled all-sentences accuracy rate..

Correction you may quote: The 2023 evaluation reports 74.5% average citation precision; it is not the share of all answers that are correct.

Affiliate sites exceed news and research in the cited-page chart

Verdict: unsupported. Checked 2026-11-02.

Circulation: Hudgens chapter 1, prose on page 28.

Primary source: Hudgens (2026), Generative Engine Optimization: The Definitive Guide to AI SEO, Wiley, ISBN 9781394439805; publisher chapter excerpt https://catalogimages.wiley.com/images/db/pdf/9781394439805.excerpt.pdf. Location: Chapter 1 page 28 prose and page 30 Figure 1.26, both read including the rendered chart.

What the source establishes: The prose puts research behind affiliate pages. Figure 1.26 prints product 45.4%, news and research 24.8%, affiliate 15.6%, blog 7.0%, user reviews 6.6%, press releases 0.6%. Thus the figure places research above affiliate.

Base, setting and date: Chart attributed to HubSpot AEO 2025; underlying sample base is not printed in this excerpt.

Correction you may quote: The chapter chart places news and research above affiliate pages; the prose reverses their order.

The displayed Reddit audience is daily active users

Verdict: unverifiable. Checked 2026-11-02.

Circulation: Hudgens chapter 1, audience chart.

Primary source: Hudgens (2026), Generative Engine Optimization: The Definitive Guide to AI SEO, Wiley, ISBN 9781394439805; publisher chapter excerpt https://catalogimages.wiley.com/images/db/pdf/9781394439805.excerpt.pdf. Location: Chapter 1 page 7, Figure 1.2, rendered chart.

What the source establishes: The chart is labeled daily active user growth and its 2025 bar is approximately 530 million. Its citation is only Reddit, 2026. No original investor metric or denominator is supplied here; this audit does not relabel it as monthly users.

Base, setting and date: See the source-specific setting in the finding.

Correction you may quote: The displayed audience chart lacks a recoverable underlying metric source; do not silently relabel its users.

A domain usage percentage is 124%

Verdict: unverifiable. Checked 2026-11-02.

Circulation: Hudgens chapter 1, domain chart.

Primary source: Hudgens (2026), Generative Engine Optimization: The Definitive Guide to AI SEO, Wiley, ISBN 9781394439805; publisher chapter excerpt https://catalogimages.wiley.com/images/db/pdf/9781394439805.excerpt.pdf. Location: Figure 1.14, publisher excerpt PDF page 19.

What the source establishes: The rendered chart prints FSC.org as Used 124%. The excerpt does not identify a denominator or sufficiently define Used. A value above 100% cannot be repaired into a valid domain share by guessing a different metric.

Base, setting and date: One chart label in the publisher chapter excerpt, 2026.

Correction you may quote: The chart prints 124%, but its undefined base prevents us from treating it as a share.

The chart tracks conversions rather than conversations

Verdict: unsupported. Checked 2026-11-02.

Circulation: Hudgens chapter 1, language-use chart.

Primary source: Hudgens (2026), Generative Engine Optimization: The Definitive Guide to AI SEO, Wiley, ISBN 9781394439805; publisher chapter excerpt https://catalogimages.wiley.com/images/db/pdf/9781394439805.excerpt.pdf. Location: Figure 1.21, publisher excerpt PDF page 25.

What the source establishes: The chart heading says 37.5% of ChatGPT Conversions are Generative. Its subtitle and intent categories concern prompts, not completed commercial conversions. The wording does not establish a conversion rate.

Base, setting and date: One prompt-intent chart reproduced in the 2026 publisher excerpt.

Correction you may quote: The intent chart does not measure commercial conversions.

The relationship is not casual

Verdict: supported with conditions. Checked 2026-11-02.

Circulation: Hudgens chapter 1 page 39.

Primary source: Hudgens (2026), Generative Engine Optimization: The Definitive Guide to AI SEO, Wiley, ISBN 9781394439805; publisher chapter excerpt https://catalogimages.wiley.com/images/db/pdf/9781394439805.excerpt.pdf. Location: Page 39 prose.

What the source establishes: The supplied page uses casual where the surrounding correlation discussion requires causal. This is a wording defect, not a new empirical result.

Base, setting and date: See the source-specific setting in the finding.

Correction you may quote: The chapter is discussing causality, not informality.

The worked review scores reproduce from the stated normalized formula

Verdict: unverifiable. Checked 2026-11-02.

Circulation: Roumeliotis et al. 2026 Table 1.

Primary source: Roumeliotis et al. (2026), The Evolution of SEO in the Era of Generative AI, Electronics 15, 3541, DOI https://doi.org/10.3390/electronics15163541. Location: Equations 1 and 2, Table 1, pages 6 to 7; independent upper-bound arithmetic.

What the source establishes: Under the stated six normalized weighted criteria and the visible raw maxima, the example upper bounds are approximately 0.792, 0.650 and 0.587, below printed 0.89, 0.84 and 0.61. The text also mentions a DOI bonus without showing how it enters the normalized equation. The published recipe therefore does not reproduce its worked scores.

Base, setting and date: Three worked examples; weights 0.40/0.25/0.15/0.10/0.05/0.05; additional DOI-bonus handling unspecified.

Correction you may quote: The review needs a reproducible normalization and bonus calculation before its worked scores can be verified.

Uniformly changing every score weight by 10% changes the ranking

Verdict: unsupported. Checked 2026-11-02.

Circulation: Roumeliotis et al. 2026 sensitivity analysis.

Primary source: Roumeliotis et al. (2026), The Evolution of SEO in the Era of Generative AI, Electronics 15, 3541, DOI https://doi.org/10.3390/electronics15163541. Location: Section 2.5, page 8.

What the source establishes: Multiplying every fixed weight by the same positive constant preserves every pairwise score order. The reported 54.3% to 80.0% overlaps do not give per-scenario results reconciling the uniform cases or additional manual changes.

Base, setting and date: See the source-specific setting in the finding.

Correction you may quote: Uniform positive rescaling of fixed weights cannot change the score ranking; the stated sensitivity analysis needs further explanation.

The reported venue shares use the stated 70-paper base

Verdict: unverifiable. Checked 2026-11-02.

Circulation: Roumeliotis et al. 2026 corpus description.

Primary source: Roumeliotis et al. (2026), The Evolution of SEO in the Era of Generative AI, Electronics 15, 3541, DOI https://doi.org/10.3390/electronics15163541. Location: PDF page 7, venue percentage passage.

What the source establishes: Several printed venue shares are compatible with integer counts out of 59 and incompatible with conventional one-decimal rounding of integer counts out of 70. The passage does not supply the venue counts. We do not replace its stated corpus size with an inferred 59-paper denominator.

Base, setting and date: Venue-share arithmetic in the 2026 review; raw venue counts unavailable.

Correction you may quote: The printed venue shares need their actual counts and denominator; 59 is an inference, not a recovered corpus.

Verdict: supported with conditions. Checked 2026-11-02.

Circulation: Xu et al. 2026, 55,393 trending queries over 40 days.

Primary source: Xu, Iqbal and Montgomery (2026), Measuring Google AI Overviews, IMC 2026, DOI https://doi.org/10.1145/3777912.3839818; arXiv 2605.14021 v 2. Location: Section 4.1, Table 3, PDF pages 6 to 7.

What the source establishes: 7,583 of 55,393 trending-query executions returned an AI Overview, about 13.7%, in March 13 through April 21, 2026. The executions span 32,608 unique normalized queries; a query-balanced calculation is separate.

Base, setting and date: 55,393 executions, 19 topical categories, 40-day 2026 window; not a representative rate for every search.

Correction you may quote: Xu and colleagues observed AI Overviews in 7,583 of 55,393 trending-query executions during their 40-day 2026 window.

AI Overviews appear for 64.7% of searches

Verdict: supported with conditions. Checked 2026-11-02.

Circulation: Xu et al. 2026, question-form subset.

Primary source: Xu, Iqbal and Montgomery (2026), Measuring Google AI Overviews, IMC 2026, DOI https://doi.org/10.1145/3777912.3839818; arXiv 2605.14021 v 2. Location: Section 4.1 and Table 4, PDF page 7.

What the source establishes: 2,716 of 4,198 question-form query executions returned an AI Overview, about 64.7%. Question form is defined by a leading interrogative from the printed list, not every conversational query.

Base, setting and date: Question-form subset of the same March 13 through April 21, 2026 trending-query executions.

Correction you may quote: The 64.7% figure applies to the paper's 4,198 question-form executions, not all searches.

AI Overviews appear for 51.5% of searches

Verdict: supported with conditions. Checked 2026-11-02.

Circulation: Grossman et al. 2026 representative query sample.

Primary source: Grossman et al. (2026), How Generative AI Disrupts Search, SIGIR 2026; https://arxiv.org/abs/2604.27790v1. Location: Table 1 and section 4.1, PDF pages 3 to 4.

What the source establishes: The 51.5% activation figure is for the 5,000-query ORCAS subset. The entire mixed 11,500-query benchmark has a different 65.6% figure. Collection was December 7 to 8, 2025, under the stated mobile Newark setting.

Base, setting and date: 5,000 sampled ORCAS real-user queries, rather than all 11,500 benchmark questions.

Correction you may quote: Grossman and colleagues report 51.5% for the 5,000-query ORCAS subset, not their entire mixed benchmark.

AI Overviews appear for about 30% of searches

Verdict: unverifiable. Checked 2026-11-02.

Circulation: Authoritas figure repeated by Otterly.

Primary source: Exact original vendor measurement not recovered in this audit. Location: Vendor-origin record and denominator unverified.

What the source establishes: We have not recovered the original vendor table and sampling base needed to verify about 30% overall activation. Secondary repetition cannot establish the target population, dates or measurement unit. This entry records an unresolved source, not a verified rate.

Base, setting and date: Unresolved vendor-origin claim; no general search-population inference permitted.

Correction you may quote: This vendor figure remains unverified without its original sample, dates and metric.

AI Overviews appear for 74 to 75% of searches

Verdict: unverifiable. Checked 2026-11-02.

Circulation: Authoritas problem-solving subset repeated by Otterly.

Primary source: Exact original vendor measurement not recovered in this audit. Location: Vendor-origin record and denominator unverified.

What the source establishes: We have not recovered the original vendor table and sampling base needed to verify 74 to 75% activation in a problem-solving subset. Secondary repetition cannot establish the target population, dates or measurement unit. This entry records an unresolved source, not a verified rate.

Base, setting and date: Unresolved vendor-origin claim; no general search-population inference permitted.

Correction you may quote: This vendor figure remains unverified without its original sample, dates and metric.

AI Overviews appear for about 18% of searches

Verdict: supported with conditions. Checked 2026-11-02.

Circulation: Pew 2025 observed search visits.

Primary source: Pew Research Center (2025), Google users are less likely to click on links when an AI summary appears; https://www.pewresearch.org/short-reads/2025/07/22/google-users-are-less-likely-to-click-on-links-when-an-ai-summary-appears-in-the-results/. Location: How we did this and overall activation paragraph, opened 2 November 2026.

What the source establishes: 12,593 of 68,879 unique Google searches generated a summary in the study. The searches were drawn from March 2025 browsing by 900 US adults and their results were collected April 7 to 17, 2025. The page rounds the rate to 18%.

Base, setting and date: 900 US adults; 68,879 unique Google searches; March histories with April result collection.

Correction you may quote: Pew's rounded 18% concerns the study's March 2025 search histories with results collected in April, not all present searches.

Recognition of 99.4% means generic discovery of 99.4%

Verdict: unsupported. Checked 2026-11-02.

Circulation: Sharma 2026, 112-new products preprint.

Primary source: Amit Prakash Sharma (2026), The Discovery Gap, single-author preprint, https://arxiv.org/abs/2601.00912. Location: Sections 3.1 to 3.4 and Table 4.1, PDF pages 6 to 8.

What the source establishes: The paper reports ChatGPT direct-query success 334/336, 99.4%, and discovery success 26/784, 3.32%, for 112 selected Product Hunt products. Perplexity figures are 317/336 and 65/784. Collection was December 15 to 20, 2025; the two named configurations differ in search availability.

Base, setting and date: 112 selected products from the top 500 of the 2025 Product Hunt leaderboard; three direct and seven discovery questions per product and engine.

Correction you may quote: In Sharma's single preprint, product-name recognition and generic discovery use different question sets and yield different rates; recognition is not discovery.

ChatGPT sent traffic to 900,000 domains in one month

Verdict: unverifiable. Checked 2026-11-02.

Circulation: Vendor figure dated November 2024.

Primary source: Exact original vendor measurement not recovered in this audit. Location: Vendor-origin record and denominator unverified.

What the source establishes: We have not recovered the original vendor table and sampling base needed to verify 900,000 referring destinations in November 2024. Secondary repetition cannot establish the target population, dates or measurement unit. This entry records an unresolved source, not a verified rate.

Base, setting and date: Unresolved vendor-origin claim; no general search-population inference permitted.

Correction you may quote: This vendor figure remains unverified without its original sample, dates and metric.

Verdict: unsupported. Checked 2026-11-02.

Circulation: Circulating snippet headline.

Primary source: Ahrefs, We Analyzed 2 Million Featured Snippets, https://ahrefs.com/blog/featured-snippets-study/. Location: Click-distribution comparison in the primary vendor article.

What the source establishes: The recovered Ahrefs article reports about 8.6% for the featured snippet and 19.6% for the following result in its comparison. It does not establish a universal 35% share for snippets. The exact origin and sampling base of the 35% retelling remain unrecovered.

Base, setting and date: Historical vendor study; its figure cannot be generalized to all searches or AI answers.

Correction you may quote: The recovered vendor study does not support a universal 35% snippet click share.

Triet and Thanh simulated outcomes presented as field evidence

Verdict: unsupported. Checked 2026-11-02.

Circulation: Nguyen Minh Triet and Ta Van Thanh (2025), Smart Search Optimization: A Theoretical Framework Integrating SEO, AEO and GEO in the Era of Genetic AI, European Modern Studies Journal 9(6),288 to 297.

Primary source: Nguyen Minh Triet and Ta Van Thanh (2025), Smart Search Optimization: A Theoretical Framework Integrating SEO, AEO and GEO in the Era of Genetic AI, European Modern Studies Journal 9(6),288 to 297; DOI https://doi.org/10.59573/emsj.9(6).2025.20. Location: PDF page 8, printed 295, results and discussion.

What the source establishes: The discussion calls the figures derived from simulation, and the leads row is labeled Synthetic. The row moves from 10 to 26, a 160% arithmetic increase. This is a constructed example, not a measured field effect.

Base, setting and date: One simulated worked example in a 2025 theoretical chapter.

Correction you may quote: The stated gain is simulated and cannot be quoted as a measured field result.

Cite Sources has the 31.4% combination average

Verdict: unsupported. Checked 2026-11-02.

Circulation: GEO section 5.3 prose.

Primary source: Aggarwal et al. (2024), GEO, KDD 2024, DOI https://doi.org/10.1145/3637528.3671900; author manuscript https://arxiv.org/html/2311.09735v3. Location: Combination discussion and Figure 4.

What the source establishes: The prose attributes 31.4 to source citation; Figure 4 places 31.4 on the fluency row. Keep the table labels and the prose discrepancy visible.

Base, setting and date: Separate 200-example combination analysis, not the full test split.

Correction you may quote: The 31.4 value belongs to the plotted fluency row, not the source-citation row.

Fluency has the highest average across combinations

Verdict: unsupported. Checked 2026-11-02.

Circulation: GEO section 5.3 prose.

Primary source: Aggarwal et al. (2024), GEO, KDD 2024, DOI https://doi.org/10.1145/3637528.3671900; author manuscript https://arxiv.org/html/2311.09735v3. Location: Figure 4.

What the source establishes: The plotted statistics row has 32.1 against 31.4 for fluency. This is a benchmark combination average, not a field citation probability.

Base, setting and date: Separate 200-example combination analysis, not the full test split.

Correction you may quote: The plotted statistics row has the higher combination average.

The location study fully identifies its 15 US cities

Verdict: unverifiable. Checked 2026-11-02.

Circulation: Chen and Koudas 2026 release statement.

Primary source: Chen and Koudas (2026), Location, Location, Generation, preprint, https://arxiv.org/abs/2610.24081v1. Location: Methods and materials statement.

What the source establishes: The paper states a 15-city US component but does not enumerate those cities in the retrieved report. That limits exact replication from the paper alone.

Base, setting and date: See the source-specific setting in the finding.

Correction you may quote: The US city component is not fully reproducible without its city list.

Verdict: unverifiable. Checked 2026-11-02.

Circulation: Chen and Koudas 2026 release statement.

Primary source: Chen and Koudas (2026), Location, Location, Generation, preprint, https://arxiv.org/abs/2610.24081v1. Location: Materials release statement.

What the source establishes: The retrieved paper says materials are released but supplies no usable repository link. We have not established that the materials are publicly accessible elsewhere.

Base, setting and date: See the source-specific setting in the finding.

Correction you may quote: A release statement is not a usable materials link.

Negative rejection means failing to ignore irrelevant retrieved passages

Verdict: unsupported. Checked 2026-11-02.

Circulation: Roumeliotis et al. 2026 review.

Primary source: Roumeliotis et al. (2026), The Evolution of SEO in the Era of Generative AI, Electronics 15, 3541, DOI https://doi.org/10.3390/electronics15163541; Chen et al., RGB, https://arxiv.org/abs/2309.01431, sections 2 to 3. Location: Review PDF p. 17; RGB primary sections 2 to 3.

What the source establishes: The review uses this gloss for RGB. In the primary benchmark, negative rejection is declining to answer when none of the retrieved documents contains the required knowledge. Ignoring noise when relevant evidence is present is a different ability.

Base, setting and date: Definition and scope audit of two cited passages, not a GEO efficacy experiment.

Correction you may quote: Negative rejection concerns missing supporting knowledge, not merely noisy retrieval.

Geographic information retrieval demonstrates generative optimization

Verdict: unsupported. Checked 2026-11-02.

Circulation: Roumeliotis et al. 2026 review.

Primary source: Roumeliotis et al. (2026), The Evolution of SEO in the Era of Generative AI, Electronics 15, 3541, DOI https://doi.org/10.3390/electronics15163541; Chen et al., RGB, https://arxiv.org/abs/2309.01431, sections 2 to 3. Location: Review PDF p. 20.

What the source establishes: The review connects a geographic-IR source to the GEO discussion. A place-oriented retrieval task does not, by its name or subject, establish an effect of rewriting pages for generated answers.

Base, setting and date: Definition and scope audit of two cited passages, not a GEO efficacy experiment.

Correction you may quote: Geographic retrieval and generative optimization require separate task definitions.

Repeated Pew citations are separate independent results

Verdict: unsupported. Checked 2026-11-02.

Circulation: Hidayet Karamuk, From SEO to Answer Engine Optimization (AEO): Generative AI and the Transformation of Search Visibility, chapter 14, printed pages 278 to 295.

Primary source: Hidayet Karamuk, From SEO to Answer Engine Optimization (AEO): Generative AI and the Transformation of Search Visibility, chapter 14, printed pages 278 to 295; Google Search Central, https://developers.google.com/search/blog/2023/08/howto-faq-changes. Location: Chapter measurement and answer-architecture discussions.

What the source establishes: The chapter cites the same Pew report in several discussions. Its 8% versus 15%, and about 1% summary-link figures, occur together in the measurement passage. Repeated references do not create independent evidence. We do not repeat the unverified claim that the numerical figure appears six times.

Base, setting and date: Audit of one 18-page chapter; no tactic-effect estimate.

Correction you may quote: Several references to one report still represent one evidence source.

The AEO chapter is evidence of comparative GEO performance

Verdict: overstated. Checked 2026-11-02.

Circulation: Hidayet Karamuk, From SEO to Answer Engine Optimization (AEO): Generative AI and the Transformation of Search Visibility, chapter 14, printed pages 278 to 295.

Primary source: Hidayet Karamuk, From SEO to Answer Engine Optimization (AEO): Generative AI and the Transformation of Search Visibility, chapter 14, printed pages 278 to 295; Google Search Central, https://developers.google.com/search/blog/2023/08/howto-faq-changes. Location: Chapter 14, printed 278 to 295.

What the source establishes: No occurrences of ChatGPT, Perplexity, Claude or Aggarwal were found in the retrieved chapter text. It is a conceptual discussion centered on search answer presentation, not a comparative test of the commercial engines.

Base, setting and date: Audit of one 18-page chapter; no tactic-effect estimate.

Correction you may quote: The chapter is not a comparative GEO experiment.

Correct FAQPage typing establishes Google rich-result eligibility

Verdict: unsupported. Checked 2026-11-02.

Circulation: Hidayet Karamuk, From SEO to Answer Engine Optimization (AEO): Generative AI and the Transformation of Search Visibility, chapter 14, printed pages 278 to 295.

Primary source: Hidayet Karamuk, From SEO to Answer Engine Optimization (AEO): Generative AI and the Transformation of Search Visibility, chapter 14, printed pages 278 to 295; Google Search Central, https://developers.google.com/search/blog/2023/08/howto-faq-changes; current Google documentation change https://developers.google.com/search/updates#removing-faq-rich-result. Location: Chapter structured-markup discussion; Google August 8, 2023 notice and current FAQ documentation-removal update, read November 2, 2026.

What the source establishes: The chapter distinguishes FAQPage and QAPage vocabulary but omits the 2023 Google restrictions. The August 8, 2023 operator notice restricted FAQ rich results to authoritative government and health sites. The current operator update states that FAQ rich results are no longer shown, following the May 2026 announcement. Valid vocabulary does not establish search-result eligibility.

Base, setting and date: Audit of one 18-page chapter; no tactic-effect estimate.

Correction you may quote: Correct schema typing does not establish a Google rich result; the current operator update says FAQ rich results are no longer shown.

The GEO experiment validates the manuscript content-format ratings

Verdict: overstated. Checked 2026-11-02.

Circulation: Akshat Singh Bisht, Generative Engine Optimisation (GEO) vs. Traditional SEO: Implications for Digital Marketing Strategy in the Age of AI Search, supplied manuscript.

Primary source: Akshat Singh Bisht, Generative Engine Optimisation (GEO) vs. Traditional SEO: Implications for Digital Marketing Strategy in the Age of AI Search, supplied manuscript; Aggarwal et al., KDD 2024, DOI https://doi.org/10.1145/3637528.3671900. Location: Manuscript paragraphs 195,212,250; GEO Table 1.

What the source establishes: The manuscript attributes ratings to the GEO experiment while also identifying the ratings as practitioner synthesis rather than a validated scale. The original experiment does not empirically validate every format and strategic framework in that synthesis.

Base, setting and date: Bibliographic and attribution audit of the retrieved manuscript, not an assessment of author intent.

Correction you may quote: The format ratings are synthesis, not a scale validated by the GEO experiment.

The retrieved manuscript proves Reddit ranks ahead of Quora

Verdict: unverifiable. Checked 2026-11-02.

Circulation: Akshat Singh Bisht, Generative Engine Optimisation (GEO) vs. Traditional SEO: Implications for Digital Marketing Strategy in the Age of AI Search, supplied manuscript.

Primary source: Akshat Singh Bisht, Generative Engine Optimisation (GEO) vs. Traditional SEO: Implications for Digital Marketing Strategy in the Age of AI Search, supplied manuscript; Aggarwal et al., KDD 2024, DOI https://doi.org/10.1145/3637528.3671900. Location: Manuscript domain discussion.

What the source establishes: The current text names Reddit and Quora together. We have not identified an exact passage in this version that ranks Reddit first, so the stronger attributed error is not asserted. A source-specific ordering needs the original vendor table.

Base, setting and date: Bibliographic and attribution audit of the retrieved manuscript, not an assessment of author intent.

Correction you may quote: A pair of domain names does not establish their order.

The Gartner search-volume forecast was issued in 2023

Verdict: unsupported. Checked 2026-11-02.

Circulation: Akshat Singh Bisht, Generative Engine Optimisation (GEO) vs. Traditional SEO: Implications for Digital Marketing Strategy in the Age of AI Search, supplied manuscript.

Primary source: Akshat Singh Bisht, Generative Engine Optimisation (GEO) vs. Traditional SEO: Implications for Digital Marketing Strategy in the Age of AI Search, supplied manuscript; Aggarwal et al., KDD 2024, DOI https://doi.org/10.1145/3637528.3671900; linked primary release https://www.gartner.com/en/newsroom/press-releases/2024-02-19-gartner-predicts-search-engine-volume-will-drop-25-percent-by-2026-due-to-ai-chatbots-and-other-virtual-agents. Location: Manuscript paragraph 389; Gartner release 2024-02-19.

What the source establishes: The manuscript reference dates the forecast February 19, 2023, while the exact linked primary-release address is dated February 19, 2024. The live release body returned an access challenge on November 2, 2026, so this entry establishes a bibliographic date mismatch, not a fresh review of the forecast body or its accuracy.

Base, setting and date: Bibliographic audit of the supplied manuscript and its exact primary-release address; live body unavailable at this check.

Correction you may quote: The search-volume forecast was issued February 19,2024.

The KDD GEO article occupies proceedings pages 22 to 32

Verdict: unsupported. Checked 2026-11-02.

Circulation: Akshat Singh Bisht, Generative Engine Optimisation (GEO) vs. Traditional SEO: Implications for Digital Marketing Strategy in the Age of AI Search, supplied manuscript.

Primary source: Akshat Singh Bisht, Generative Engine Optimisation (GEO) vs. Traditional SEO: Implications for Digital Marketing Strategy in the Age of AI Search, supplied manuscript; Aggarwal et al., KDD 2024, DOI https://doi.org/10.1145/3637528.3671900. Location: Manuscript reference 7; DOI10.1145/3637528.3671900.

What the source establishes: The manuscript prints 22 to 32. The KDD proceedings record gives 5 to 16. This is a bibliographic correction, independent of the experiment conclusions.

Base, setting and date: Bibliographic and attribution audit of the retrieved manuscript, not an assessment of author intent.

Correction you may quote: The KDD GEO proceedings pages are 5 to 16.

The manuscript establishes that AI summaries caused the quoted CTR decline

Verdict: overstated. Checked 2026-11-02.

Circulation: Akshat Singh Bisht, Generative Engine Optimisation (GEO) vs. Traditional SEO: Implications for Digital Marketing Strategy in the Age of AI Search, supplied manuscript.

Primary source: Akshat Singh Bisht, Generative Engine Optimisation (GEO) vs. Traditional SEO: Implications for Digital Marketing Strategy in the Age of AI Search, supplied manuscript; Aggarwal et al., KDD 2024, DOI https://doi.org/10.1145/3637528.3671900. Location: Abstract and limitations.

What the source establishes: The abstract uses attributable language for a secondary industry figure. The manuscript does not supply a randomized or otherwise identified causal design for that attribution, and its limitations acknowledge commercially interested secondary sources.

Base, setting and date: Bibliographic and attribution audit of the retrieved manuscript, not an assessment of author intent.

Correction you may quote: A quoted CTR association is not a causal estimate supplied by this manuscript.

No declared conflict establishes absence of commercial affiliation

Verdict: supported with conditions. Checked 2026-11-02.

Circulation: Akshat Singh Bisht, Generative Engine Optimisation (GEO) vs. Traditional SEO: Implications for Digital Marketing Strategy in the Age of AI Search, supplied manuscript.

Primary source: Akshat Singh Bisht, Generative Engine Optimisation (GEO) vs. Traditional SEO: Implications for Digital Marketing Strategy in the Age of AI Search, supplied manuscript; Aggarwal et al., KDD 2024, DOI https://doi.org/10.1145/3637528.3671900. Location: Author affiliation and conflict statement, paragraphs 5 and 345.

What the source establishes: The manuscript lists a commercial research-firm affiliation and separately declares no conflict. Both are visible disclosures. They do not by themselves establish whether a financial relationship affected a result, and we make no allegation about intent.

Base, setting and date: Bibliographic and attribution audit of the retrieved manuscript, not an assessment of author intent.

Correction you may quote: Read the affiliation and conflict declaration together; neither proves influence on a finding.

arXiv:2310.XXXX identifies the GEO paper

Verdict: laundered. Checked 2026-11-02.

Circulation: Nguyen Minh Triet and Ta Van Thanh (2025), Smart Search Optimization: A Theoretical Framework Integrating SEO, AEO and GEO in the Era of Genetic AI, European Modern Studies Journal 9(6),288 to 297.

Primary source: Nguyen Minh Triet and Ta Van Thanh (2025), Smart Search Optimization: A Theoretical Framework Integrating SEO, AEO and GEO in the Era of Genetic AI, European Modern Studies Journal 9(6),288 to 297; DOI https://doi.org/10.59573/emsj.9(6).2025.20. Location: Reference list, printed 297.

What the source establishes: The reference contains a literal placeholder, with a title and authors that do not identify the real GEO paper. The identifiable original is Aggarwal and colleagues, arXiv 2311.09735 and KDD 2024 DOI10.1145/3637528.3671900.

Base, setting and date: Title, bibliography and diagram audit of one 2025 chapter.

Correction you may quote: A placeholder identifier cannot serve as a primary research citation.

Genetic AI is the title’s term for this generative-search framework

Verdict: supported with conditions. Checked 2026-11-02.

Circulation: Nguyen Minh Triet and Ta Van Thanh (2025), Smart Search Optimization: A Theoretical Framework Integrating SEO, AEO and GEO in the Era of Genetic AI, European Modern Studies Journal 9(6),288 to 297.

Primary source: Nguyen Minh Triet and Ta Van Thanh (2025), Smart Search Optimization: A Theoretical Framework Integrating SEO, AEO and GEO in the Era of Genetic AI, European Modern Studies Journal 9(6),288 to 297; DOI https://doi.org/10.59573/emsj.9(6).2025.20. Location: Title, printed 288.

What the source establishes: The retrieved title prints Genetic AI. This is an editorial wording defect, not evidence about how the document was written or about the authors’ intentions.

Base, setting and date: Title, bibliography and diagram audit of one 2025 chapter.

Correction you may quote: The title prints Genetic AI; that wording is not a separate optimization method.

The schematic and prose specify the same spatial hierarchy

Verdict: supported with conditions. Checked 2026-11-02.

Circulation: Nguyen Minh Triet and Ta Van Thanh (2025), Smart Search Optimization: A Theoretical Framework Integrating SEO, AEO and GEO in the Era of Genetic AI, European Modern Studies Journal 9(6),288 to 297.

Primary source: Nguyen Minh Triet and Ta Van Thanh (2025), Smart Search Optimization: A Theoretical Framework Integrating SEO, AEO and GEO in the Era of Genetic AI, European Modern Studies Journal 9(6),288 to 297; DOI https://doi.org/10.59573/emsj.9(6).2025.20. Location: Figure 1 and prose, printed 290.

What the source establishes: The diagram runs downward from SEO through AEO to GEO. The prose calls SEO the base and GEO uppermost. The spatial descriptions are not aligned; neither is an empirical test of a necessary hierarchy.

Base, setting and date: Title, bibliography and diagram audit of one 2025 chapter.

Correction you may quote: The schematic is conceptual and its orientation differs from the prose.

The undated agency report’s method order is verified

Verdict: unverifiable. Checked 2026-11-02.

Circulation: Aggarwal et al., GEO, KDD 2024, DOI 10.1145/3637528.3671900, Tables 1 and 4.

Primary source: Aggarwal et al., GEO, KDD 2024, DOI https://doi.org/10.1145/3637528.3671900, Tables 1 and 4; exact undated agency original not recovered. Location: Agency original not recovered; comparison GEO Table 1.

What the source establishes: The original benchmark comparison gives quotation addition about 40.9%, statistics 30.6% and source citation 27.5% on position-adjusted word count. We have not recovered the exact agency original needed to attribute an inverted ordering to it.

Base, setting and date: Primary benchmark checked; agency-specific allegations remain unverified.

Correction you may quote: Use the primary benchmark order; the agency-specific attribution remains unverified.

The agency report prints the GEO formula upside down

Verdict: unverifiable. Checked 2026-11-02.

Circulation: Aggarwal et al., GEO, KDD 2024, DOI 10.1145/3637528.3671900, Tables 1 and 4.

Primary source: Aggarwal et al., GEO, KDD 2024, DOI https://doi.org/10.1145/3637528.3671900, Tables 1 and 4; exact undated agency original not recovered. Location: Agency original not recovered.

What the source establishes: We cannot verify the agency formula without its exact original. The GEO primary metric must be read in its own definition rather than reconstructed from an unavailable retelling.

Base, setting and date: Primary benchmark checked; agency-specific allegations remain unverified.

Correction you may quote: The alleged formula defect requires the exact agency original.

Three agency claims are present in Sharma and Dhiman

Verdict: unverifiable. Checked 2026-11-02.

Circulation: Aggarwal et al., GEO, KDD 2024, DOI 10.1145/3637528.3671900, Tables 1 and 4.

Primary source: Aggarwal et al., GEO, KDD 2024, DOI https://doi.org/10.1145/3637528.3671900, Tables 1 and 4; exact undated agency original not recovered. Location: Exact agency original not recovered; Sharma and Dhiman supplied article.

What the source establishes: The agency text needed to match the three attributed claims has not been recovered. Sharma and Dhiman is not used here as evidence of tactic effectiveness.

Base, setting and date: Primary benchmark checked; agency-specific allegations remain unverified.

Correction you may quote: The attributed claim chain remains unverified.

The agency reference list is a verified copy of another paper

Verdict: unverifiable. Checked 2026-11-02.

Circulation: Aggarwal et al., GEO, KDD 2024, DOI 10.1145/3637528.3671900, Tables 1 and 4.

Primary source: Aggarwal et al., GEO, KDD 2024, DOI https://doi.org/10.1145/3637528.3671900, Tables 1 and 4; exact undated agency original not recovered. Location: Exact agency original not recovered.

What the source establishes: The list-to-list comparison cannot be performed without the agency file. We make no copied-reference or intent finding from the supplied allegation alone.

Base, setting and date: Primary benchmark checked; agency-specific allegations remain unverified.

Correction you may quote: The copied-reference allegation is not established by this audit.

The reported survey has a consistent 300+ respondent base

Verdict: unsupported. Checked 2026-11-02.

Circulation: Apoorav Sharma and Prabhjot Dhiman, The Impact of AI-Powered Search on SEO: The Emergence of Answer Engine Optimization, supplied Chandigarh University article.

Primary source: Apoorav Sharma and Prabhjot Dhiman, The Impact of AI-Powered Search on SEO: The Emergence of Answer Engine Optimization, supplied Chandigarh University article; publication year and public identifier unconfirmed. Location: PDF p. 5 methods andp 7 to 10 tables.

What the source establishes: The methods claim 300+ respondents. Most result totals are 244, while the displayed age counts total 234. The paper does not reconcile these bases in the retrieved passages.

Base, setting and date: Internal sample and table consistency audit; no raw respondent records available.

Correction you may quote: The displayed sample bases are inconsistent:300+,244 and 234.

The two Q10 categories have correlation+0.12

Verdict: unsupported. Checked 2026-11-02.

Circulation: Apoorav Sharma and Prabhjot Dhiman, The Impact of AI-Powered Search on SEO: The Emergence of Answer Engine Optimization, supplied Chandigarh University article.

Primary source: Apoorav Sharma and Prabhjot Dhiman, The Impact of AI-Powered Search on SEO: The Emergence of Answer Engine Optimization, supplied Chandigarh University article; publication year and public identifier unconfirmed. Location: PDF p. 10, Tables 4 to 5.

What the source establishes: Table 5 reports+0.12 between declining CTR and algorithm transparency. Under the single-choice indicator interpretation implied by counts 112 and 49 out of 244, their mutually exclusive indicators have correlation−0.4617. Raw records are unavailable, so the interpretation and the missing-data limit remain explicit.

Base, setting and date: Internal sample and table consistency audit; no raw respondent records available.

Correction you may quote: The displayed single-choice categories do not reproduce the reported positive correlation.

All claimed research methods have reported results

Verdict: unverifiable. Checked 2026-11-02.

Circulation: Apoorav Sharma and Prabhjot Dhiman, The Impact of AI-Powered Search on SEO: The Emergence of Answer Engine Optimization, supplied Chandigarh University article.

Primary source: Apoorav Sharma and Prabhjot Dhiman, The Impact of AI-Powered Search on SEO: The Emergence of Answer Engine Optimization, supplied Chandigarh University article; publication year and public identifier unconfirmed. Location: Methods and results, PDF p. 5 to 12.

What the source establishes: The methods describe a broad program including interviews, cases, comparisons and regression. The retrieved results do not separately report outputs from that full program. A method named in a plan is not itself a completed result.

Base, setting and date: Internal sample and table consistency audit; no raw respondent records available.

Correction you may quote: The reported tables do not supply results for the full methods program.

The GEO experiment tested a structured-data uplift

Verdict: unsupported. Checked 2026-11-02.

Circulation: Abu Rayhan, Generative Engine Optimization (GEO): The Mechanics, Strategy, and Economic Impact of the Post-Search Era, supplied report.

Primary source: Abu Rayhan, Generative Engine Optimization (GEO): The Mechanics, Strategy, and Economic Impact of the Post-Search Era, supplied report; primary comparison Aggarwal et al. (2024), GEO, KDD, DOI https://doi.org/10.1145/3637528.3671900. Location: Report section 4.3, PDF p. 10; original GEO methods and Table 1.

What the source establishes: The report bundles structured data with statistics and quotations in its up-to 40% passage. The original GEO experiment does not contain a structured-data treatment. The supplied attribution cannot establish that additional effect.

Base, setting and date: Citation and treatment audit of the retrieved report; no evidence of a structured-data effect.

Correction you may quote: The original GEO benchmark did not test a structured-data uplift.

The cited Gartner mobile-app release supports the search-volume forecast

Verdict: unsupported. Checked 2026-11-02.

Circulation: Abu Rayhan, Generative Engine Optimization (GEO): The Mechanics, Strategy, and Economic Impact of the Post-Search Era, supplied report.

Primary source: Abu Rayhan, Generative Engine Optimization (GEO): The Mechanics, Strategy, and Economic Impact of the Post-Search Era, supplied report; primary comparison Aggarwal et al. (2024), GEO, KDD, DOI https://doi.org/10.1145/3637528.3671900. Location: Reportp 2 and references 5 to 6, p 19.

What the source establishes: The search-volume claim points to reference 5, whose listed title concerns mobile-app usage by 2027. The search-volume forecast for 2026 is the separate release listed as reference 6.

Base, setting and date: Citation and treatment audit of the retrieved report; no evidence of a structured-data effect.

Correction you may quote: The search forecast points to the wrong Gartner reference.

Exactly 23 of 69 references are uncited

Verdict: unverifiable. Checked 2026-11-02.

Circulation: Abu Rayhan, Generative Engine Optimization (GEO): The Mechanics, Strategy, and Economic Impact of the Post-Search Era, supplied report.

Primary source: Abu Rayhan, Generative Engine Optimization (GEO): The Mechanics, Strategy, and Economic Impact of the Post-Search Era, supplied report; primary comparison Aggarwal et al. (2024), GEO, KDD, DOI https://doi.org/10.1145/3637528.3671900. Location: Retrieved report reference list and in-text numbering.

What the source establishes: The report lists 69 references, but the asserted 23-uncited count has not been reconstructed from its in-text mapping. We do not present that count as a verified finding.

Base, setting and date: Citation and treatment audit of the retrieved report; no evidence of a structured-data effect.

Correction you may quote: The exact uncited-reference count remains unverified.

The printed per-run drift can be converted into a settled time forecast

Verdict: unverifiable. Checked 2026-11-02.

Circulation: Kumar 2026 abstract and discussion.

Primary source: Kumar (2026), Generative Engine Optimization at Scale, vendor preprint, https://arxiv.org/abs/2606.20065v1. Location: Section 6.4 drift table and interpretation.

What the source establishes: The paper labels the cohort band−1.34% to+0.03% per run and discusses slow erosion. Its analysis code and customer data are not public, and the printed units do not let us safely choose between a relative percentage change and percentage points for a time forecast. Prompt-set drift is itself acknowledged. We do not convert the slope into a fixed prediction.

Base, setting and date: Vendor observational cohort; variable repeated tracking and unresolved slope units.

Correction you may quote: The printed slope is not a settled forecast of future visibility.

88.5% of cited pages are optimized

Verdict: supported with conditions. Checked 2026-11-02.

Circulation: Potential misreading of our dated result; no independent republication history established by this audit.

Primary source: Broadcastwell (2026), AI search visibility: source evidence and two registered analyses, version DOIhttps://doi.org/10.5281/zenodo.23804758; public data https://zenodo.org/records/23804758. Location: Study 1 class-share CSV, target/all/all.

What the source establishes: Study 1 classifies 3,913 of 4,420 successfully observed target URLs as definitely optimized:88.5%. The eligible target archive contains 5,861 URLs; failures are outside the classification base.

Base, setting and date: Captured target pages,30 October 2026; September 2026 answer archive.

Correction you may quote: The 88.5% describes observed signatures in successfully retrieved cited pages, not a causal gain.

All eligible target pages were observable

Verdict: overstated. Checked 2026-11-02.

Circulation: Potential misreading of our dated result; no independent republication history established by this audit.

Primary source: Broadcastwell (2026), AI search visibility: source evidence and two registered analyses, version DOIhttps://doi.org/10.5281/zenodo.23804758; public data https://zenodo.org/records/23804758. Location: Study 1 fetch-outcome CSV.

What the source establishes: Complete HTML retrieval succeeded for 4,420 of 5,861 target URLs,75.4%. Missing pages do not receive a negative optimization class.

Base, setting and date: Target URLs cited in September 2026; later 30 October 2026 retrieval.

Correction you may quote: A failed retrieval is missing page evidence, not proof that the page lacks a feature.

A single answer is a stable brand rank

Verdict: overstated. Checked 2026-11-02.

Circulation: Potential misreading of our dated result; no independent republication history established by this audit.

Primary source: Broadcastwell (2026), AI search visibility: source evidence and two registered analyses, version DOIhttps://doi.org/10.5281/zenodo.23804758; public data https://zenodo.org/records/23804758. Location: repeat-disagreement CSV.

What the source establishes: Naming was mixed in 1,259 of 14,192 complete vendor-question-engine groups,8.9%. Each group has three valid repeats;108 incomplete groups are omitted.

Base, setting and date: September 2026 fixed question archive; descriptive dependent observations.

Correction you may quote: The archive measures repeated naming under stated conditions, not a permanent rank.

All valid answers supply a citation

Verdict: overstated. Checked 2026-11-02.

Circulation: Potential misreading of our dated result; no independent republication history established by this audit.

Primary source: Broadcastwell (2026), AI search visibility: source evidence and two registered analyses, version DOIhttps://doi.org/10.5281/zenodo.23804758; public data https://zenodo.org/records/23804758. Location: answer-citation-bases CSV.

What the source establishes: 39 of 2,095 valid target answers have empty citation arrays,1.9%. Missing arrays remain in the answer denominator and contribute no citation entry.

Base, setting and date: September 2026 target answers; controls separate.

Correction you may quote: Citation-entry counts and answer counts require different denominators.

The top-ten citation hosts are the engines’ trusted sources

Verdict: overstated. Checked 2026-11-02.

Circulation: Potential misreading of our dated result; no independent republication history established by this audit.

Primary source: Broadcastwell (2026), AI search visibility: source evidence and two registered analyses, version DOIhttps://doi.org/10.5281/zenodo.23804758; public data https://zenodo.org/records/23804758. Location: citation-concentration CSV.

What the source establishes: The leading ten exact hosts supply 3,192 of 18,893 target citation entries,16.9%. This records source occurrence, not a measurement of trust or source accuracy.

Base, setting and date: September 2026 target citation arrays; registered Study 2.

Correction you may quote: A citation concentration table does not measure an engine’s trust.

A brand mention means its own site was cited

Verdict: overstated. Checked 2026-11-02.

Circulation: Potential misreading of our dated result; no independent republication history established by this audit.

Primary source: Broadcastwell (2026), AI search visibility: source evidence and two registered analyses, version DOIhttps://doi.org/10.5281/zenodo.23804758; public data https://zenodo.org/records/23804758. Location: vendor-pair-summary CSV.

What the source establishes: Of 8,085 named vendor-answer pairs,2,608 have a qualifying own-domain citation,32.3%. A third-party reference does not count as the vendor’s own site.

Base, setting and date: September 2026; one checked vendor and one eligible answer per pair.

Correction you may quote: Naming and a qualifying own-domain citation are separate outcomes.

G2 accounts for 3.1% of software answers

Verdict: overstated. Checked 2026-11-02.

Circulation: Potential misreading of our dated result; no independent republication history established by this audit.

Primary source: Broadcastwell (2026), AI search visibility: source evidence and two registered analyses, version DOIhttps://doi.org/10.5281/zenodo.23804758; public data https://zenodo.org/records/23804758. Location: review-platform-occurrences CSV.

What the source establishes: The exact host g 2.com appears in 580 of 18,893 target citation entries,3.1%. The base is citation entries, not 2,095 answers or unique readers.

Base, setting and date: September 2026 target citation arrays.

Correction you may quote: The 3.1% is a citation-entry share, not an answer share.

The review-platform share has one undisputed value

Verdict: overstated. Checked 2026-11-02.

Circulation: Potential misreading of our dated result; no independent republication history established by this audit.

Primary source: Broadcastwell (2026), AI search visibility: source evidence and two registered analyses, version DOIhttps://doi.org/10.5281/zenodo.23804758; public data https://zenodo.org/records/23804758. Location: Frozen and post-audit class-share CSVs.

What the source establishes: The frozen class contains 1,663 of 18,893 target entries,8.8%. The dated post-audit sensitivity has 1,706,9.0%, after five host-role corrections. The original result remains published beside the correction.

Base, setting and date: September 2026 archive;30 October 2026 host-purpose audit.

Correction you may quote: The original class count and dated sensitivity must remain distinguishable.

Zero crawler matches prove AI services never visit

Verdict: overstated. Checked 2026-11-02.

Circulation: Potential misreading of our dated result; no independent republication history established by this audit.

Primary source: Broadcastwell (2026), AI search visibility: source evidence and two registered analyses, version DOIhttps://doi.org/10.5281/zenodo.23804758; public data https://zenodo.org/records/23804758. Location: ai-crawler-log.csv.

What the source establishes: The initial 30 October 2026 window has zero declared AI-token matches among 83 requests after excluding one own check; the descriptive Wilson 95% interval is 0.0% to 4.4%. It is a short partial-window request log, not an engine-use census.

Base, setting and date: UTC2026-10-30 T05:35:44.478 Z to 05:38:00.771 Z.

Correction you may quote: A short zero-match log does not establish absence of AI-service use.

The September release contains 2,250 target answers

Verdict: overstated. Checked 2026-11-02.

Circulation: Potential misreading of our dated result; no independent republication history established by this audit.

Primary source: Broadcastwell (2026), The Absence Index, release 2026-09, version DOI https://doi.org/10.5281/zenodo.22907695; public benchmark https://index.broadcastwell.com/downloads#benchmark; README https://index.broadcastwell.com/data/benchmark/README.md. Location: September long-format benchmark, release-2026-09.jsonl; README and schema; direct downloads checked November 2, 2026 Eastern.

What the source establishes: The public benchmark has 2,100 target-attempt rows: 2,095 valid answers and five failed outcomes. The separate 150 controls are outside this file and the target-answer base. Failed outcomes cannot be silently folded into a vendor-absence rate.

Base, setting and date: September 22 and 23, 2026; 14 categories, 286 checked vendors, 140 fixed target questions, five products and three scheduled repeats. Repeats may be dependent; the checked set is not a probability sample..

Correction you may quote: State whether the denominator is scheduled attempts, valid target answers or controls.

Zero retained August attempt rows mean zero August observations

Verdict: overstated. Checked 2026-11-02.

Circulation: Potential misreading of our dated result; no independent republication history established by this audit.

Primary source: Broadcastwell (2026), historical August aggregate release, version DOI https://doi.org/10.5281/zenodo.22721962; benchmark README https://index.broadcastwell.com/data/benchmark/README.md; unchanged aggregate download https://index.broadcastwell.com/data/benchmark/release-2026-08.1-aggregates.json. Location: Benchmark README, Files and evidence boundary; historical aggregate file with its erratum.

What the source establishes: August supplies historical aggregate observations. Original answer text and per-engine attempt identities were not retained. Zero supplied clean-format attempt rows mean unavailable evidence, not zero observations. The aggregate file cannot be expanded into fabricated attempts.

Base, setting and date: Historical August 2026 release; public benchmark and aggregate downloads checked November 2, 2026 Eastern. No comparable consecutive pair or Study 3 persistence result is established..

Correction you may quote: Unavailable historical answer rows are not zero observations.

Wikipedia occupies 27% of all first positions

Verdict: overstated. Checked 2026-11-02.

Circulation: Potential misreading of our dated result; no independent republication history established by this audit.

Primary source: Lewandowski, Sünkler and Yagci (2021), WebSci, DOIhttps://doi.org/10.1145/3447535.3462479. Location: Lewandowski et al. (2021), section 5, PDF page 7, printed page 18; DOI 10.1145/3447535.3462479.

What the source establishes: The paper reports 27% of its 1,974 Wikipedia results at the first position. The denominator is Wikipedia results, not every first-position slot.

Base, setting and date: Selected German Google queries collected 2020;2021 paper.

Correction you may quote: The 27% is a position share conditional on a Wikipedia result.

Corrections and disputes

A correction keeps the original claim, the new evidence, the date and our resulting verdict together beside its entry. Disputed findings are not silently removed. An unavailable original is a reason to withhold a conclusion, rather than infer a motive. Our current public exports retain the checked date and the unresolved-source limits.

What does an unverifiable verdict mean?

The original, denominator or calculation needed for the claim is not available or resolved.

Does a rejected claim prove author intent?

No. A source or inference defect does not by itself establish an author’s intent.

This page belongs to evidence. Read unverified numbers, how to inspect a statistic, academic access, the library’s citation guidance, and the Index method.

Sources

The author, venue or publication record, exact location and verification limit for each source are printed beside its entry and preserved in both downloadable files. The source report is AI search visibility: source evidence and two registered analyses, November 2, 2026. Its underlying September release remains separately versioned.

Cite this page

Broadcastwell research team. (2026, November 2). GEO Claim Registry: trace the number to its source. Broadcastwell. https://docs.broadcastwell.com/research/claims/

@online{broadcastwell2026researchclaims,
  author = {{Broadcastwell research team}},
  title = {GEO Claim Registry: trace the number to its source},
  year = {2026},
  date = {2026-11-02},
  url = {https://docs.broadcastwell.com/research/claims/}
}

Prepared by the Broadcastwell research team. Lead: Vikram Patel.

Sources checked November 2, 2026. Not peer reviewed.

Continue

Broadcastwell research. Sources and dates are stated on each page. Prose and derived tables CC BY 4.0, site code MIT. External sources retain their own licenses. Not peer reviewed. Evidence standard.