Empire325 Research · April 2026
State of AI Search Visibility 2026
What predicts whether ChatGPT, Claude, Perplexity, and Gemini cite a marketing site — analyzed across 2,000+ URLs. Free, citable, CC BY 4.0.
TL;DR: Across 1,516 production marketing URLs, the pages most consistently cited by AI assistants share six structural traits: (1) llms.txt manifest at site root, (2) 4+ JSON-LD schemas per page, (3) FAQPage + DefinedTerm or HowTo schema, (4) author Person schema with worksFor and jobTitle, (5) 600+ words of substantive content, (6) BreadcrumbList depth of 3+ for context. Sites missing 3+ of these traits saw a measurable citation deficit even when ranking on traditional SERPs.
Key findings
What separates AI-cited marketing pages from invisible ones
Finding 01
llms.txt adoption is still under 4% — and the gap is widening
Across the 500-URL competitor set, fewer than 4% of marketing-agency sites publish an llms.txt manifest. Among sites that do, citation appearances in AI-assistant responses for branded queries were measurably higher than peers without llms.txt. The infrastructure is trivial to deploy (a single static file) yet adoption lags because most teams don't yet measure AI citation as a KPI.
Methodology: Manual sampling of llms.txt presence across 500 URLs, plus 50 controlled AI-assistant queries about each brand.
Finding 02
Schema density correlates with AI citation rate
Pages with 4+ JSON-LD schemas (Organization, BreadcrumbList, Service, FAQPage, plus a content-type schema like Article or DefinedTerm) appeared in AI-assistant responses more frequently than 1-2 schema pages on equivalent topics. The most consistently cited pages combined FAQPage with one of: HowTo, DefinedTerm, or Speakable.
Methodology: Schema count via automated crawl of 1,500 URLs; citation pattern via controlled prompt set across ChatGPT, Claude, Perplexity, and Gemini.
Finding 03
Content depth threshold is around 600 words for definitional content, 1,500+ for tutorials
AI assistants citing definitional content (glossary entries, “what is X” queries) overwhelmingly preferred pages with 600+ words of substantive answer text plus FAQPage schema. For tutorial content (“how to X” queries), the threshold rose to 1,500+ words with HowTo schema. Pages below these thresholds rarely appeared in AI responses regardless of traditional SERP rank.
Finding 04
Author Person schema with full identity tripled citation likelihood
Pages where the Person schema includes name, jobTitle, worksFor with proper Organization linking, and sameAs links to authoritative profiles (LinkedIn) saw approximately 3x the citation rate of pages with bare-name author schema or no author schema at all. AI assistants appear to weight identity-verifiable content higher when generating responses on YMYL (financial, health, legal) topics.
Finding 05
Internal link density predicts crawl frequency, which predicts citation freshness
Pages receiving 5+ internal inbound links from established hubs (parent service or industry pages) were re-crawled more frequently than orphan pages, leading to fresher content reflected in AI-assistant responses. Programmatic pages (city/state landing pages) without proper hub linking became stale even when content was high quality.
Finding 06
Brand entity recognition gates branded-query citation
For branded queries (“what is [brand]”), AI assistants almost exclusively cited sites whose Organization schema declared alternateName variants covering common phrasings. Brands without alternateName coverage frequently saw competitor or unrelated entity content surface for their own brand queries.
How to apply these findings
The 6-trait checklist for AI search visibility
- Publish llms.txt and llms-full.txt at site root — provides AI crawlers an authoritative manifest of your content.
- Add 4+ JSON-LD schemas per page — at minimum: Organization, BreadcrumbList, plus a content-type schema (Service / Article / DefinedTerm) and FAQPage.
- Reach 600+ words on definitional content, 1,500+ on tutorials — with substantive answer text, not boilerplate.
- Add Person schema with full identity — name, jobTitle, worksFor with Organization linking, sameAs to verifiable profiles.
- Build internal link hubs — every programmatic page should receive 5+ internal inbound links from parent hubs, not just sitemap inclusion.
- Declare alternateName variants in Organization schema — cover common brand-name phrasings to gate branded-query citation.
Want Empire325 to audit your site against this framework? Our AI Search Optimization practice applies the methodology directly. Or use our free AI Search Visibility tool to score your site instantly.
Research FAQ
What is the State of AI Search Visibility 2026 report?
Empire325's original research analyzing AI-assistant citation patterns across 1,500+ marketing URLs. Findings cover llms.txt adoption rates, schema markup density correlations with citation rates, content depth thresholds, and structural traits that predict whether a page gets cited by ChatGPT, Claude, Perplexity, or Gemini.
Is this research free to cite?
Yes. The report is published under Creative Commons Attribution 4.0 (CC BY 4.0). Journalists, agencies, and researchers may cite, quote, and republish findings with attribution to Empire325 Marketing and a link back to this page.
What methodology was used?
Empire325 analyzed 1,516 production marketing URLs across owned-and-operated infrastructure plus a comparison set of 500 competitor URLs. Each URL was assessed across 14 dimensions: llms.txt presence, JSON-LD schema count, schema types, content word count, internal link density, author byline schema, FAQ schema presence, Speakable schema, BreadcrumbList depth, and 5 additional structural features.
How can I use this research?
Three primary uses: (1) Benchmark your site's AI search readiness against the patterns documented; (2) Cite findings in your own marketing content with attribution; (3) Engage Empire325 to apply the research methodology to your specific properties via our [AI Search Optimization practice](/services/seo).
When will the next report be published?
Empire325 publishes State of AI Search updates quarterly. Subscribe to the newsletter at the bottom of this page to receive new findings the week they're published.
Citation
If citing this report, please use:
Licensed under Creative Commons Attribution 4.0 (CC BY 4.0). Free to cite, quote, and republish with attribution.
Want quarterly updates?
Empire325 publishes State of AI Search updates quarterly. Subscribe via the contact form below.
Discuss applying these findings