AI Search SEO: Citation Strategies skill

AI Search SEO: Citation Strategies

Use GEO and AEO as legacy labels for AI citation readiness. For Google, the Google Search Central AI features guidance is explicit: optimization for AI Overviews and AI Mode is SEO, not a separate discipline.

Core AI Search SEO Research

Princeton GEO Paper (KDD 2024)

The Princeton GEO paper reports that content changes can boost AI visibility by up to 40% in test conditions. Treat the result as research on AI search surfaces, not as a separate Google ranking discipline.

Technique	Improvement
Citing authoritative sources	+115.1% visibility (5th-ranked sites, main experiment)
Quotation addition	+28% (main experiment); +37% (Perplexity.ai validation, Table 7)
Statistics addition	+41% (main experiment); +22% (Perplexity.ai validation, Table 7)
FAQPage entity markup	May aid AI citation as an entity signal for visible Q&A; no Google rich result; impact unverified

Traditional keyword stuffing performs worse than baseline in generative engines.

Cross-Platform Citation Divergence

• Only 11% of domains are cited by both ChatGPT and Perplexity (Digital Bloom, 2025; domain-level, not URL-level; AI Overviews not included in that study)
• 80% of LLM citations don't rank in Google's top 100 (Ahrefs, Aug 2025) - classic organic rankings alone are a poor predictor of AI citation
• Brands are 6.5x more likely to be cited through third-party sources than their own domains (AirOps, Oct 2025) - earned media dominates AI visibility

Kevin Indig's AI Search Pipeline (Jan 5, 2026)

Three critical stages:

1. Retrieval: Which pages enter the candidate set
   • Server response time under 200ms TTFB
   • Metadata relevance
   • Content must be in HTML (not behind JS)
2. Citation: Which sources get mentioned
   • Content freshness dominates (70%+ cited pages updated within 12 months)
   • Content within 3 months performs best
3. Trust: Which citations users click
   • Brand recognition
   • Source authority

Content Format Impact on Citations

Format	Impact	Source
Listicles	Often over-index in vendor datasets; exact shares vary and are directional
Tables/structured data	May improve extractability; specific 2.5x vendor claim is unverified
Long-form (2,000+ words)	Often over-indexes in vendor studies; treat multipliers as directional
FAQPage entity markup	May aid AI citation as an entity signal for visible Q&A; no Google rich result; impact unverified
Content with statistics	Original, sourced statistics improve citeability; exact lifts vary by study
Sections of 120-180 words between headings	Self-contained passages likely help extraction; exact lift is unverified
Comparison tables with <thead>	May improve extraction; attributed SEL figure is unverified

Passage-Level Extractability (2026)

Google's AI systems fragment pages and evaluate self-contained answer passages, not just whole documents. Start each H2 with an approximately 50-word direct-answer sentence, then build a 120-180 word passage that can stand alone if quoted or summarized. Support it with named entities, dates, source attribution, and a specific example.

Entity density and demonstrated first-hand Experience break ties. A single clean passage with original testing, named tools, and verifiable evidence can earn an AI Overview citation even when the full page is not the strongest organic result.

Platform-Specific Citation Patterns

Each AI platform has distinct content preferences:

Platform	Favored Content Type	Key Bias
ChatGPT	"Best X" listicles	43.8% of citations are list-format content
Perplexity	Reddit discussions	6.6% of all citations come from Reddit
AI Overviews	Google properties	23% of citations favor Google-owned sources

2026 wrinkle: AI Overviews now highlight links from a user's subscribed publications, so publisher subscriptions can influence which sources users see inside the AI answer (Nieman Lab, 2026-05).

Perplexity content decay: Citation relevance begins declining 2-3 days post-publication - Perplexity heavily weights recency, making it the most freshness-dependent platform. Content older than 1 week sees sharp citation drops.

Content Freshness Requirements

• 76.4% of ChatGPT's most-cited pages updated within 30 days (Ahrefs, ~17M citations)
• URLs cited in AI results are 25.7% fresher than traditional search
• Content < 3 months old is 3x more likely to get cited
• Action: Update critical content when facts, screenshots, pricing, methods, source availability, or SERP intent have materially changed

Off-Site Signals (Dominate AI Visibility)

Ahrefs Study (Dec 2025, 75,000 brands)

Factor	Correlation with AI Visibility
YouTube mentions	0.737 (strongest)
Branded web mentions	0.656-0.709
Domain Rating	0.266-0.326
Backlinks	0.218 (dramatically weaker than expected)

Platform-Specific Citation Rates

YouTube:

• Citations in AI Overviews up 414% (Q1 2025, NP Digital, 10K+ AIO analysis)
• How-to videos up 651%
• Visual demos up 592%
• 200x more cited than any other video platform
• Optimization: keywords in titles/transcripts, Q&A-style, 10+ min, public transcripts

Reddit:

• Citations surged 1.30% → 7.15% (450% growth)
• Google's $60M annual API deal
• 2.2-21% of AI Overview citations by query type
• Strategy: Authentic participation in 3-5 subreddits BEFORE any promotional content

Review Platforms (B2B):

• G2 accounts for 22-23% of review-platform citations (Radix via G2's own blog; self-reported - Hall.com's independent analysis found G2 at only 8.25% of B2B software citations in ChatGPT)
• 33% of review citations come from G2 (Profound via G2's blog; treat as directional)
• Multi-platform presence: 4.6-6.3 citations vs 1.8 without (2.6-3.5x multiplier)

Wikipedia/Wikidata:

• 7.8% of all ChatGPT citations (Profound)
• Used as "credibility tiebreaker" when sources conflict

Budget Allocation

Recommended: 40% owned content / 60% earned media (Most companies allocate 90/10 - this is wrong for AI search SEO)

88-92% of AI citations come from off-site signals in vendor-reported datasets. Treat this as directional, not a universal law.

AI Crawler Technical Requirements

Crawler	JavaScript Rendering
GPTBot (OpenAI)	No
ChatGPT-User	No
ClaudeBot	No
PerplexityBot	No
Googlebot	Yes

Critical: Content behind JavaScript is invisible to ChatGPT, Claude, Perplexity. Use SSR, SSG, or ISR. Test by disabling JS and reloading.

Google's Official Gen-AI Guidance

Google's stance holds: optimization for AI Overviews and AI Mode is SEO. There is no special schema for gen-AI features, and Google does not need llms.txt. Use standard crawlable HTML, Article schema with author and Organization entities, helpful content, clear source attribution, and fast server responses.

AI Crawler Traffic Growth

• Cloudflare AI crawling rose 32% YoY across all monitored sites
• GPTBot traffic grew 305% YoY
• PerplexityBot traffic grew 157,490% YoY (from near-zero baseline)
• 65% of AI bot hits target content published within the past year (Seer Interactive)
  • freshness is a retrieval signal, not just a citation signal

Performance Requirements for AI Retrieval

• Server response time under 200ms TTFB (Kevin Indig pipeline)
• TTFB above 600ms may reduce crawl or extraction reliability
• Some crawlers and retrieval systems use short practical timeouts; verify per crawler
• Core Web Vitals are a constraint, not a growth lever - good CWV doesn't reliably outperform, but severe LCP failure creates disadvantage (Search Engine Land, 107,352 pages)
• Top 10 domains capture 46% of all ChatGPT citations per topic (Growth Memo, Mar 2026)
• Slow pages may miss crawl, fetch, or extraction opportunities
• Vercel analysis of 500+ million GPTBot fetches found zero evidence of JS execution

robots.txt for AI Visibility

User-agent: GPTBot
Allow: /
User-agent: ChatGPT-User
Allow: /
User-agent: ClaudeBot
Allow: /
User-agent: PerplexityBot
Allow: /

llms.txt Standard

Google does not need llms.txt for AI Overviews or AI Mode. Treat the file as an optional site inventory for non-Google tools, not a ranking or citation lever. Do not spend AI search SEO budget on llms.txt before crawlability, passage extraction, Article schema, source quality, and entity consistency.

Attribution Gaps

Perplexity visits ~10 pages per query but cites only 3-4. Not all AI responses include citations - optimizing for retrieval is critical. Content must enter the candidate set before citation is possible.

AI Search Case Study Results

These examples are illustrative, vendor-reported, and not independently verified. Do not reuse the numbers as factual benchmarks without primary confirmation.

Company	Results	Timeframe
Go Fish Digital	+43% AI traffic, +83% conversions, 25x conversion rate	3 months
Netpeak USA	+120% revenue, +693% AI visits	Ongoing
Nine Peaks Media	36% visibility improvement, first ChatGPT citations	Ongoing
ABM Agency/Chemours	82% ChatGPT mention rate, $90M+ pipeline	Ongoing
Smart Rent	32% SQL increase, 40% faster pipeline	Ongoing

Entity-First SEO

Every page should unambiguously represent ONE canonical entity. Google Knowledge Graph: 800B facts about 8B entities.

Entity building timeline (3-6 months):

1. Create entity map with Wikidata Q-IDs
2. Establish Wikipedia/Wikidata presence only when the entity meets independent notability requirements; disclose conflicts of interest and follow platform policy
3. Build entity consistency across all platforms (exact same name)
4. Practice "controlled co-occurrence" via third-party mentions
5. Earn external citations from recognized publications

Readability and AI Search Connection

Readability can support AI citation rates, but the lift is not independently verified. Use Flesch 60-75 as a clarity heuristic, not as a citation guarantee.

Flesch Score & AI Citation Rates

• Commercial platform reports associate Flesch 60-75 with more AI citations; this is vendor-reported internal data with no independent verification.
• Teams improving Flesch from 52→68 saw parallel citation lifts within two crawl windows
• Content that is too complex (Flesch <50) or too simple (Flesch >80) gets fewer citations - AI systems prefer fluent, authoritative writing

Citation Position Bias

• 44.2% of all LLM citations come from the first 30% of text is a single-source Growth Memo finding (Feb 2026, Kevin Indig). Treat it as directional support for answer-first formatting, not a settled benchmark.
• Direct answers in the first 1-2 sentences of each section maximize extractability for AI systems

AI Search Tactic Combinations

Princeton GEO paper (KDD 2024) findings on readability-related tactics:

• Fluency optimization = 15-30% visibility boost
• Statistics addition = up to 41% visibility boost
• Fluency + Statistics combined outperforms any single tactic by 5.5%
• Keyword stuffing performs -10% WORSE than baseline

FLOW evidence triple is mandatory for AI-citation readiness. AI assistants extract claims that have year anchor in prose, inline publisher + title, and URL with retrieval date. Stats without the triple are less likely to surface in citations. See flow-alignment.md.

Schema & Structure for AI Citation

• Comparison tables with proper HTML (<thead>, <tbody>) = 47% higher AI citation rates (attributed to SEL; primary source unlocatable - treat as directional)
• Structured data helps machine understanding, but do not claim all major AI platforms use schema during citation selection without current primary evidence

Platform-Specific Citation Behaviors

Platform	Key Behavior	Readability Preference
ChatGPT	Wikipedia = 7.8% of citations; SearchGPT: 87% match Bing top 10	Prefers well-structured, fluent content
Perplexity	Reddit = 46.7% of top-10 sources; strongest depth correlation (0.191)	2-3 day content decay; heavily weights recency
AI Overviews	93.67% from top-10 organic; avg 10.2 links per response	Prefers established authority + clear answers

Only 11% of domains are cited by both ChatGPT and Perplexity (Digital Bloom). Only 12% of URLs cited by ChatGPT, Perplexity, and Copilot rank in Google's top 10 (Ahrefs).

Content Freshness for AI Citation

• 65% of AI bot hits target content published within the past year (Seer Interactive)
• 85% of AI Overview citations come from content <2 years old
• 44% of AI Overview citations come from 2025 content specifically
• 50% of Perplexity citations come from 2025 alone
• Content older than 3 months sees 3x fewer citations