Research

Sara Allawati +4 · SIGIR 2026 · Microsoft Research

When an AI answer occupies the top of the page, brand visibility inside that answer can matter before a user ever evaluates blue links.

Anar Yeginbergen +3 · arXiv

AI visibility should measure whether a cited page actually supports the answer, because URL-level citation alone can hide weak grounding.

Perplexity Research · Perplexity

Brand visibility is moving beyond blue-link rankings: AI assistants may find sources through many hidden sub-searches before a user sees the final answer.

Peibo Zhang +2 · arXiv

AI search can push attention toward cited communities, but interface changes can quickly rewrite the benefit.

Pablo Landeta-Lopez +3 · Computing

AI crawling is becoming semantic and agentic, which changes what crawlable content means for AI visibility.

Haofei Xu +2 · arXiv

Google AI visibility needs citation tracking and claim checks, not just rank tracking, because cited pages and supported claims are separate outcomes.

Devi Prasad Bal & Subhashree Puhan · arXiv

Query expansion can add noise; retrieval strategy should be tested before assuming more searches improve AI answers.

Zebin Guo +2 · arXiv

For brands with comparison tables, specs, pricing, or datasets, AI retrieval needs structured facts at the cell and entity level.

Riley Grossman +5 · SIGIR 2026

Brands cannot assume Google rankings, Gemini answers, and AI Overview citations are the same visibility surface.

Zhang Kai +2 · arXiv

Citation count alone can mislead; a cited page only matters if its evidence actually shapes the answer.

Li Ju +2 · arXiv

AI engines can cite a page but still answer from memory; visibility work needs context-faithfulness tests.

Trakkr Research

AI visibility decays fast. Getting cited once isn’t the finish line; staying cited is the real work — which is exactly what continuous tracking is for.

Beining Wu +9 · ACL 2026 Findings

This is useful for Trakkr because it treats AI visibility and citation accuracy as a joint measurement problem.

Shijia Xu +5 · arXiv

Citation quality depends on both what evidence enters the context and whether the answer is checked against it.

Saharsh Agarwal & Ananya Sen · SSRN

This is one of the most useful study shapes for Trakkr because AI visibility only matters commercially when it changes downstream attention.

ProGEO.ai Research · ProGEO.ai

Large brands are experimenting with GEO signals, but most still have shallow AI-specific technical coverage.

Hangeol Chang +4 · arXiv

AI engines often answer from hidden rewritten searches, so the phrasing of your content needs to satisfy multiple evidence paths.

Andrea Volpini +3 · arXiv

Structured data may matter most when it helps agents navigate entities, not when it is treated as a standalone ranking trick.

Zhihua Tian +4 · arXiv

The practical takeaway is sharp: diagnose the citation failure before rewriting the page.

Eeham Khan +2 · Canadian AI 2026

The useful pattern is not biomedical-specific: grounded sub-claims make citation failures easier to diagnose.

Jiaqi Yuan +5 · arXiv

GEO is moving from fixed page-writing recipes toward engine-aware, page-specific optimization.

Maango · Maango Research

Most of the web has not made a clear machine-readable choice about AI crawling, which makes crawler monitoring a real governance issue.

Sinan Aral +2 · arXiv

AI search may concentrate attention, so brands in the long tail need visibility measurement before they assume search demand still reaches them.

Faye Zhang +5 · arXiv

Visual and marketplace brands should think in AI-retrievable collections, not only individual asset captions.

Yilun Hua +4 · arXiv

Not every relevant page is useful to every model; content utility should be measured model by model.

Shuguang Jiao +6 · arXiv

For AI visibility, being retrieved is only the first gate; the model still has to use the evidence.

Roy Xie +6 · EACL 2026

More retrieval is not always better; AI visibility systems should measure retrieval quality, negative evidence, and abstention behavior.

Shayan Alipour +2 · arXiv

AI answers may reward existing attention, so brand visibility work needs to track third-party voices, not just owned pages.

Pranjal Aggarwal +5 · KDD 2024 · Princeton · IIT Delhi · Georgia Tech

This is the founding paper of the field Trakkr is built around. It shows the things you put on a page measurably change whether AI quotes you — visibility is optimisable, not luck.

Nelson F. Liu +6 · TACL 2024 · Stanford · UC Berkeley · Samaya AI

Being retrieved isn’t enough — position inside what the model reads matters. Being the clear, early, self-contained answer beats being one mention buried among many.

Nelson F. Liu +2 · EMNLP 2023 (Findings) · Stanford

AI cites loosely. If you’re the source, being the unambiguous support for a clear, checkable claim is how you earn a citation that holds up — not just a passing mention.

Patrick Lewis +11 · NeurIPS 2020 · Facebook AI · UCL · NYU

This is the architecture behind ChatGPT search, Perplexity and Google’s AI answers. It’s why “what’s retrievable about you” quietly decides what AI is able to say about you at all.