The AI Citation Gap: How Brand Demand, Authority, and Knowledge Panels Decide Which of 1,000 US Personal Injury Law Firms AI Assistants Cite
Description
When a US consumer asks an AI assistant a legal question, the assistant answers with a synthesized paragraph and a short list of named entities and cited pages. Which law firms appear in those answers, and whose web pages host the citations, is rapidly becoming the discoverability question of the decade for the personal injury (PI) vertical, where a single signed case can be worth six or seven figures and a single head-term click transacts for hundreds of dollars. This paper is the eighth in a series on entity-completeness signals in the top-ranking US PI vertical. It presents, to our knowledge, the first large-N per-assistant AI citation audit of the vertical: for the same 1,000 top-ranking US personal injury law firm sample used throughout the series, we join per-firm mention and cited-page counts across four AI answer surfaces (ChatGPT, Gemini, Google AI Overview, and Google AI Mode) to the brand-search-demand, Person Entity Panel, Google Business Panel, domain-authority, and schema-completeness signals established in the prior papers. Three findings define what we name the AI Citation Gap. First, AI visibility is extremely concentrated: the median firm receives just 22 AI mentions across all four surfaces combined, the top 100 firms absorb 51.6% of all mentions, the bottom 500 firms share 9.6%, and the Gini coefficient of the mention distribution is 0.65. Second, the four assistants agree on who matters (pairwise Spearman rho = 0.70 to 0.88) but not symmetrically: 126 firms that are visible on Google's AI surfaces are completely absent from ChatGPT, while only 14 firms show the reverse pattern. Third, the citations are mostly not on the firm's own website. We introduce the Citation Intermediation Ratio (CIR), the share of a firm's AI citation footprint hosted by domains other than its own, and report a median CIR of 0.72: for the median firm, roughly 72% of its AI citations are hosted by third parties, with legal directories alone hosting 23%, and 99% of firms fall below 50% own-site share. Domain authority is the dominant correlate of AI mentions (rho = 0.80), brand search demand is the strongest firm-controllable second lever (firm brand rho = 0.44, significant in regression net of authority), the Person Entity Panel is citation-neutral, and the Google Business Panel is again associated with a citation deficit (median 19 vs 36 mentions, p < 0.001; OLS coefficient negative and significant), extending the zero-click cannibalization pattern documented in the Brand Demand Flywheel paper into the AI answer layer. An OLS model on log mentions reaches R-squared = 0.64. The paper closes with Citation Surface Optimization (CSO), a practitioner framework for managing owned versus rented AI citation surfaces.
Files
AI_Citation_Gap_by_Behzad_Hussain.pdf
Additional details
Identifiers
Dates
- Copyrighted
-
2026-09-20
References
- Panda, N., & Ofitserov, V. Ranking search results (US-8682892-B1). Google Inc.
- Lehman, A. R., & Panda, N. Site quality score (US-9031929-B1). Google Inc.
- Henry, J. W. Providing knowledge panels with search results (US-11836177-B2). Google LLC.
- Trew, B. K., Swerdlow, A., & Lai, S.-W. Personal knowledge panel interface (US-9311362-B1). Google Inc.
- Betz, J. T. Corroborating facts extracted from multiple sources (US-8682913-B1). Google Inc.
- Czuba, K., & Gabrilovich, E. Predicting accuracy of submitted data (US-2023113420-A1). Google LLC.
- Guha, R. V. Search result ranking based on trust (US-8352467-B1). Google Inc.
- Chechik, G., Leviathan, Y., Tzur, Y., et al. Grouping related entities (US-9411857-B1). Google Inc.
- Iter, D., Yu, X., & Li, F. Identifying entity attribute relations (US-11263400-B2). Google LLC.
- Thakur, K. M., & Jha, M. Detection of related local entities (US-9858291-B1). Google Inc.
- Blair-Goldensohn, S., Hannan, K., McDonald, R., et al. Sentiment detection as a ranking signal for reviewable entities (US-9317559-B1). Google Inc.
- Hawkins, J. A., & Stancu-Mara, C. Identifying local experts for local search (US-9792330-B1). Google Inc.
- Rofouei, M., Shukla, A., Wei, Q., et al. Search with stateful chat (US-2024289407-A1). Google LLC.
- Bousquet, O. J. A., Sandberg, O., Gelly, S., & Brown, R. G. Associating a search query with an entity (US-2016224621-A1). Google Inc.
- Gray, M. K., Blitzer, J., Herrick, C., Venkatachary, S., Madhavan, J., Oates, S., Parakh, P., Shah, A., Rofouei, M., & Badr, I. Generative summaries for search results (US-11900068-B1). Google LLC.
- Alakuijala, J., Buck, C., Bulian, J., Ciaramita, M., Gajewski, W., Gesmundo, A., Houlsby, N., & Wang, W. Generating query variants using a trained generative model (US-11663201-B2). Google LLC.
- Broder, A. (2002). A taxonomy of web search. SIGIR Forum, 36(2), 3–10.
- Singhal, A. (2012, May 16). Introducing the Knowledge Graph: things, not strings. Official Google Blog.
- Bollacker, K., Evans, C., Paritosh, P., Sturge, T., & Taylor, J. (2008). Freebase: a collaboratively created graph database for structuring human knowledge. SIGMOD.
- Dong, X., Gabrilovich, E., Heitz, G., Horn, W., Lao, N., Murphy, K., Strohmann, T., Sun, S., & Zhang, W. (2014). Knowledge Vault: a web-scale approach to probabilistic knowledge fusion. KDD.
- Guha, R. V., Brickley, D., & Macbeth, S. (2016). Schema.org: evolution of structured data on the web. Communications of the ACM, 59(2), 44–51.
- Guu, K., Lee, K., Tung, Z., Pasupat, P., & Chang, M.-W. (2020). REALM: Retrieval-augmented language model pre-training. ICML. arXiv:2002.08909.
- Lewis, P., Perez, E., Piktus, A., Petroni, F., Karpukhin, V., Goyal, N., et al. (2020). Retrieval-augmented generation for knowledge-intensive NLP tasks. NeurIPS.
- Nogueira, R., & Cho, K. (2019). Passage re-ranking with BERT. arXiv:1901.04085.
- Metzler, D., Tay, Y., Bahri, D., & Najork, M. (2021). Rethinking search: making domain experts out of dilettantes. SIGIR Forum, 55(1). arXiv:2105.02274.
- Bohnet, B., et al. (2022). Attributed question answering: evaluation and modeling for attributed large language models. Google Research. arXiv:2212.08037.
- Aggarwal, P., Murahari, V., Rajpurohit, T., Kalyan, A., Narasimhan, K., & Deshpande, A. (2024). GEO: Generative Engine Optimization. KDD. arXiv:2311.09735.
- AI answer engine citation behavior: an empirical analysis of the GEO-16 framework (2025). arXiv:2509.10762.
- What gets cited: competitive GEO in AI answer engines (2026). arXiv:2605.25517.
- Generative engine optimization at scale: measuring brand visibility across AI search engines (2026). arXiv:2606.20065.
- Don't measure once: measuring visibility in AI search (2026). arXiv:2604.07585.
- Google. (2024). Generative AI in Search: AI Overviews. Official Google Blog.
- Google. (2025). Expanding AI Overviews and introducing AI Mode. Official Google Blog.
- OpenAI. (2024). Introducing ChatGPT search. OpenAI Blog.
- Microsoft. (2023, February). Building the new Bing. Microsoft Bing Blogs.
- Google Knowledge Panel Help. About knowledge panels. Google.
- Google Search Quality Rater Guidelines (2024 revision). Google.
- Fishkin, R. (2019, 2023). Zero-click search analyses. SparkToro.
- Semrush. AI visibility and AI Overviews research (US market, 2025-2026).
- SE Ranking, AirOps, and Authoritas. Empirical analyses of ChatGPT citation drivers: referring-domain thresholds, review-platform presence, and content freshness (2025).
- Seer Interactive. (2025). AI Overview cited-brand click-through analyses.
- Hussain, B. (2026). Schema Markup Adoption in Personal Injury Law Firm Websites: A Systematic Analysis of Structured Data Implementation Across North American Legal Services. International Journal of Law and Society, 9(3), 361-369. https://doi.org/10.11648/j.ijls.20260903.16, http://dx.doi.org/10.2139/ssrn.6551638
- Hussain, Behzad. Schema Markup Adoption in Top-Ranking Personal Injury Law Firm Websites: A Structured Data Audit of 1,005 Google Page-1 Sites Across 50 US States (July 19, 2026). http://dx.doi.org/10.2139/ssrn.7157958, https://doi.org/10.13140/RG.2.2.14954.27842
- Hussain, Behzad. The Google Business Panel Completeness Gap: A 9-Attribute Audit of 1,005 US Personal Injury Law Firms (August 08, 2026). http://dx.doi.org/10.2139/ssrn.7249838
- Hussain, Behzad. Knowledge Panels and Person Entity Confidence in the AI Search Era: A Signal, Traffic, and AI-visibility Audit of 1,005 US Personal Injury Attorneys and Law Firms. http://dx.doi.org/10.2139/ssrn.7247178, https://doi.org/10.13140/RG.2.2.31864.81924
- Hussain, Behzad. The Brand Demand Flywheel: How Brand Search Demand and Knowledge Panels of Firms and Attorneys Impact Organic Traffic in 1,000 US Personal Injury Law Firms (August 10, 2026). https://doi.org/10.13140/RG.2.2.28601.12646, http://dx.doi.org/10.2139/ssrn.7261478
- Hussain, Behzad. (2026). The Person Entity Confidence Gap: A Knowledge Panel Taxonomy Across 1,005 US Personal Injury Attorneys and Law Firms in the AI Search Era. https://doi.org/10.13140/RG.2.2.36059.12325, https://doi.org/10.6084/m9.figshare.33187500
- Hussain, B. (2026). The 2026 PI SERP Authority Report: An 11-Question Structural Audit of 1,005 US Personal Injury Law Firm Websites Against the Personal Injury Organic Authority Engine (PIOAE) Framework. J Hum Res Sus Org Stud, 1(2), 01-11. https://dx.doi.org/10.33140/JHRSOS.01.02.04, also available on SSRN: https://dx.doi.org/10.2139/ssrn.7284998