Measurement & Analytics

A 2026 Ahrefs analysis (via AuthorityTech) of 75K brands found brand web mentions had a 0.664 correlation with AI citations vs 0.218 for backlinks — vendor data, not a causal study.

Averi and CiteMetrix both tracked 680M AI citations and found ChatGPT and Perplexity share only ~11% of cited domains. Reddit dominates Perplexity top-10 (46.7%); Google's own surfaces share just

Muck Rack's December 2025 analysis of 1M+ citations found only 2% overlap between journalists PR teams pitch and those AI cites. Press release citations grew 5x in 6 months.

AnswerManiac.ai tested 100 B2B companies across HR tech, fintech, and SaaS. 53 were never mentioned by ChatGPT or Perplexity. Only 22 were cited by both engines.

Semrush + Kevin Indig study (3,981 domain appearances, 115 prompts, 14 countries, 4 AI engines): 62% of citations are ghost citations where the brand name never appears in the AI answer text.

AllAboutAI survey of 500+ participants: 65.9% say citations increase trust even without clicking. Only 27% 'often' click citations. 43.9% click to double-check facts.

AllAboutAI survey of 500+ participants: 65.9% say citations increase trust even without clicking. Only 27% 'often' click citations. 43.9% click to double-check facts.

Foglift's 2026 report (29,910 answers, Mar-Aug 2026): 79.3% include citations (avg 5.83). Per engine: Perplexity 9.1/answer, AI Overviews 7.25, Gemini 7.17, ChatGPT 2.98, Claude 2.32.

Manual audit of 1,700 businesses across 32 industries and 3 countries found 88.1% invisible in ChatGPT — 77% of Google page 1 businesses also absent.

AuthorityTech's 8-signal AEO framework: FAQ Quality 20%, Answer Clarity 19%, Statistical Density 16%, Heading Structure 16% — top 4 signals = 71% of citation likelihood.

Grossman's SIGIR 2026 research found AEO and SEO source pools overlap only ~18%. A ranking page needs a separate page-level answer-readiness check to get cited by AI.

Gartner 2026: 34% of complex purchase decisions involve AI agents completing preliminary research. Agent-first content structure gets 67% more AI agent citations.

Gartner 2026: 34% of complex purchase decisions involve AI agents completing preliminary research. Google SAGE paper (Jan 2026) shows AI agents pull from top-3 ranked pages.

ChatGPT strips query params, Claude/Gemini look like direct. Use custom channel groups, log analysis, assisted-conversion models.

Gumlet/DerivateX May 2026: AI referred 27% of new signups but GA tracked only 0.5%. AI traffic lacks referrer headers — dedicated attribution is essential for GEO ROI.

Nature Communications (Wu et al.) found 50–90% of LLM citations don't fully support claims. AI crawlers consume content 38,000x more than they refer traffic (Cloudflare). Optimize to survive

Profound (11.8B citations, 29 industries, 8 models): 57% of AI citations come from brand sites. Gemini cites brands 69% vs ChatGPT 47%. Pharma earned-media share 59%, SaaS just 11.4%.

Magna AI (May 2026): 150+ clients, 14 industries, rolling 12-month data. Wikipedia leads at 47% density on factual queries. Community platforms (Reddit/Quora) = 24%. .gov/.edu = 11.75x more likely cit

Profound analysis of 240M citations (2026): median cited-source half-life ~4.5 weeks. 40-60% of domains cited for a query rotate month-to-month. Citations are not permanent assets.

Cloudflare data: AI crawlers consume content at 38,000x the rate they send referral traffic. Just 1.08% of website traffic comes from AI referrals. ChatGPT drives 87.4% of that.

Across 22.7M citations, 79.6% of cited sites appear on only one of five AI engines; exact-page agreement is 6.8%, brand agreement 30.3%. Cross-engine overlap rising.

Erlin 500+ brand study: unmonitored brands 67 days to detect AI citation errors vs 14 days monitored. Industry error rates: e-commerce 18%, SaaS 12%, finance 11.3%. AI Overviews 44% more critical

Similarweb: ChatGPT-Google user overlap held at 95% for 8 months while gen-AI visits hit 9.5B/month (+70% YoY). ChatGPT's share of gen-AI traffic slid from 76% to 53%.

Trakkr (797,644 comparisons, 8 models): perfect consensus just 4.0%, 60% of queries have <50% agreement. Comparison queries most stable (50.4%); 'best of' queries offer opportunity.

Content Intelligence Lab cross-platform citation analysis (April 2026): average AI Overview now cites 4.1 unique domains, down from 6.2 in early 2025 — fewer slots mean higher stakes per citation.

Serpstat study: 70% of AI Overview-triggering keywords show volatility. Only 4-8% of AIO citations overlap with organic top 20. Health leads at 54% AIO share.

Omnibound/BrightEdge: AIOs expanded from 6.49% (Jan 2025) to ~48-50% (Feb 2026). Healthcare 88%, Education 83% trigger rates. 2 billion monthly users.

2026 measurement study (arXiv:2605.14021): Google AI Overviews activate on 13.7% of trending queries overall and 64.7% of question-form queries — the high-intent informational class that once drove

OtterlyAI 1M+ citation study: AI Overviews show strongest brand preference. ChatGPT weak link citations. Perplexity balanced. Community dominates Perplexity.

AI Overviews increasingly cite sources outside Google's first page. Mid-2025: 76% of AIO citations from top-10 organic. Early 2026: dropped to 38% (Ahrefs) and as low as 17% (BrightEdge).

arXiv:2605.16428 diff-in-diff: AI Overviews raised comments +12.0% and commenting users +12.4% in indexed (SFW) Reddit communities, concentrated in experience-based posts; the arrival of AI Mode

Google AI Overviews prevalence surged from ~15% to 43% YoY (Similarweb); Semrush measures ~48%. AI Mode visits rose to 279M/month. Informational queries run higher.

BrightEdge data (Feb 2026): 48% of tracked queries trigger AI Overviews. Seer Interactive: organic CTR on queries with AIO dropped 61% (1.76%→0.61%). But cited brands earn 120% more organic clicks

ChatGPT's share of generative-AI web traffic fell from 76% (Jun 2025) to ~53% (May 2026); Gemini reached 27-28%, Claude ~9% (Similarweb). ChatGPT-only coverage misses ~half the audience.

ChatGPT ecommerce referrals convert 1.81% vs 1.39% organic (+31%) but AOV is 14.3% lower (SEL, 94 brands). Perplexity converts 10.5% vs 1.76% organic. AI traffic still ~1% of sessions.

Conductor 2026 benchmarks analyzed 3.3B sessions across 13K+ domains — AI referrals are still tiny but growing steadily, with ChatGPT dominating.

Adobe Analytics: AI-referred visits match or beat general traffic engagement. Visitors from AI stay 8% longer, view 12% more pages, 23% less likely to bounce but lower purchase intent.

Conductor 2026 Benchmarks (3.3B sessions, 13K+ domains): ChatGPT 15.9% conversion, Perplexity 10.5%, Claude 5.0%. B2B SaaS sees 23x uplift. AI-referred visitors spend 68% more time on site.

Machine Relations research (Apr 2026): Perplexity, ChatGPT, Gemini share zero cited domains on 35-40% of queries. Perplexity cites 2-3x more domains per query than parametric models.

LovedByAI natural experiment across 116 sites, 108,834 pages: optimized pages earned 173.5 AI referral visits per 1,000 pages over 90 days vs 15.8 for untouched — an 11x lift.

Loganix synthesis: AI traffic converts at 14.2% vs Google organic 2.8% (5.1x). Claude 16.8%, ChatGPT 14.2%, Perplexity 12.4%. ConvertMate 2026: 4.4x better.

Reuters Institute Digital News Report 2026: only 20% globally trust AI chatbot answers for news. UK trust is just 6%, creating both risk and opportunity for brands.

Superlines analyzed 34,234 AI responses across 10 platforms: brand visibility 2.8× higher in US vs non-US markets. Citation rate: US 10.31%, non-US 3.73-6.58%.

Microsoft Clarity (1,200+ publisher sites): 70.6% of AI-referred traffic arrives without a referrer header, classified as 'direct' in GA4. AI referral traffic accounts for <1% of sessions but

AI-referred visitors now convert 42% better than non-AI traffic (vs 38% worse 12 months prior). They also generate 37% more revenue per visit and spend 48% longer on-site.

Google AI Mode returns 91% different URLs on repeat same-day searches; 40–60% of cited sources churn monthly. Single snapshots mislead.

Superlines tracked brand visibility down 35.9%, citation rate -34.4%, share of voice -34.8% in five weeks (Jan-Feb 2026). Quarterly audits are insufficient; weekly monitoring is the new minimum.

AI Overview citations have decoupled from organic top-10 rankings: overlap fell from 76% to 38% in a year (Ahrefs, CiteMetrix 680M citations). Pages ranking for underlying fan-out queries see 161%

All 14 optimized articles already ranking in Google's top 10 earned an AI citation; positions 11-30 converted 48%, below 30 only 26%.

Cambridge Data & Policy (April 2026): ~14K conversations analyzed. Gemini provides 0 clickable citations in 92% of answers. Perplexity visits ~10 pages per query but cites only 3-4. Ecosystem

ChatGPT + Perplexity = 68% of all tracked citations Q1 2026. Gemini fastest-growing (doubled MoM). Sector breakdown from 250 engagements tracked Jan-Mar 2026.

Dupple 2026 analysis confirms AI search has become the primary starting point for B2B research. The shift from Google-first to AI-first B2B discovery is well underway.

AI Overviews appear on ~82% of B2B tech searches (BrightEdge/SEO Discovery) but only ~4% of product queries; brands cited in AI answers get ~120% more clicks per impression; ~68% of US searches are

BWT AI Performance dashboard (public preview Feb 10, 2026) reveals real citation events, cited URLs, and grounding queries from Copilot/Bing AI/ChatGPT-linked surfaces. Free and underused.

Capconvert analyzed 1,000 ChatGPT citations across 200 queries. Top-3 Bing ranking correlates with citation at 80%. Content >2,000 words = multi-citation. Named authors = 1.4x lift.

Ahrefs 75K brand study shows branded web mentions r=0.664 vs backlinks r=0.218. Top quartile averages 169 AI Overview mentions vs 14 for next tier.

Wellows 22.7M-citation study: 79.6% of sources appear on one engine only; engines name the same brand 30.3% of the time but the same page only 6.8%. Page overlap is 42x above random.

Ranqo preprint (arXiv 2606.20065, Jun 2026): 100K+ AI-search responses across 100+ brands. Brand sentiment flips 6.7× more often than whether a brand is mentioned at all.

Ranqo: 102K+ prompts, 102 brands, 5 engines. Tier 1 brands in 73% of answers, mid-market 44%, niche 11%. Listicles = 21% of citations. Sentiment flips 6.7x more than mention.

Ahrefs study of 75K brands found off-site brand mentions are the

Ahrefs 75K-brand study: web mentions correlate with AI Overview visibility at 0.664 vs backlinks 0.218, Domain Authority 0.18. Top 25% by mentions earn 10x more AI citations. Seer: 75x multiplier with

BrightEdge 50K-citation analysis: 71% of AI Overview citations from organic top-12. 29% outliers succeed via original research (34%), expert authorship (28%), unique case studies (19%).

BrightEdge research (July 2026) reveals Google's AI is drawing from social platforms at massive scale, with Facebook cited in 19.5 million answers, signaling a shift in AI source diversity.

BrightEdge survey of 750+ marketers (2026): 68% making strategic adjustments for AI search. 27% targeting both AIO + ChatGPT, 18% taking comprehensive multi-platform approach. 57% still 'wait and

C-SEO Bench (NeurIPS 2025, parameterlab) found most C-SEO document-modification methods are largely ineffective or negative on ranking, while traditional SEO is significantly more effective; gains

Ahrefs finding: 28.3% of ChatGPT's most-cited pages rank nowhere in Google organic. Fewer than 10% of AI-cited sources rank in Google top 10 for the same query.

Latest ChatGPT model as of July 2026 dramatically shifts citation behavior again — the site: operator is back, brand citations and freshness increased.

OpenAI disclosures + Similarweb: ChatGPT hit 800-900M weekly users in 2026 (4x from 200M late 2024). ~17% of global digital queries. AI-platform sessions 45B/month.

ChatGPT citation accuracy rose from 68% (2023) to 87.3% (2026). Multi-source validation requires corroboration across 2+ sources, reducing errors by 43%.

seoClarity tracked ChatGPT citations across 5 markets Feb-May 2026: -94% at trough then rebounded. April update added double compression: fewer responses cite and fewer sources per cited response.

Everything-PR Research: ChatGPT's top 50 cited domains. Wikipedia still #1 but under 20% of prompts. Reddit collapsed from ~60% to 10% after September 2025 change.

ChatGPT kept citing after first hit in only ~44% of a prompt's polls vs ~70% for Gemini. One-off 'cited in ChatGPT' screenshots are near-meaningless.

AirOps (548,534 retrieved pages, 15,000 prompts): ChatGPT cites only 15% of retrieved pages; 32.9% of cited pages surface only via internal fan-out queries invisible to keyword tools.

Superlines' 34,234-response study: Grok has the highest citation rate (27.01%) and brand visibility (8.47%) vs ChatGPT 0.59%/0.14% — a ~615x citation-volume gap across platforms.

Previsible analyzed 6.77M LLM sessions across 166 GA4 properties: ChatGPT 92.4% share, Claude 64x growth, Perplexity dropped 61% from peak.

Statcounter (Apr 2026): ChatGPT = 76.85% of AI referral traffic (down from 84.21% YoY). Gemini surged from 2.31% to 9.0%. Perplexity = 7.73%. Claude = 2.66%.

Statcounter (Apr 2026): ChatGPT = 76.85% of AI referral traffic (down from 84.21% YoY). Gemini surged from 2.31% to 9.0%. Perplexity = 7.73%. Claude = 2.66%.

Foglift study of 1,430 classified citations across 5 engines. ChatGPT skips video, community discussion, personal blogs entirely. Only 55.4% classifiable — significant long-tail diversity.

Shashko May 2026 study (153K citations, 6 platforms, 5K queries): only 4.2% of ChatGPT-cited URLs also rank in Google top-10. Classical SEO does not transfer to ChatGPT.

CiteDash benchmark of ~3,400 citations found ChatGPT hallucinates ~38% of academic references. Claude ~24%, Gemini ~19%, Perplexity ~8%.

Semrush found only 25.6% of cited domains overlap between ChatGPT minimal and high reasoning modes; high reasoning runs 4.6x more internal searches and doubles citations per response.

After ChatGPT's May 2026 in-answer brand-link update, Similarweb measured referral traffic up 157.7% week-over-week; ~60% of referrals land on homepages and ChatGPT traffic converts at 7.1%.

Across 680M citations, ChatGPT and Perplexity share only 11% of cited domains. AI Overviews and AI Mode share 13.7%. Each engine reads a different corpus.

AirOps analyzed 16,851 queries: top retrieval result cited ~58% of time, position 10 only ~14%. Narrow focused pages beat broad ultimate guides.

After Google's Sept 2025 num=100 parameter removal, ChatGPT's top sources shifted dramatically: Wikipedia from 55% to <20%, Reddit from ~60% to ~10%, Forbes doubled, LinkedIn climbed to #5.

Trakkr analyzed 575K+ AI crawler visits, 1.3M citations, 11,500+ query rewrites. Only 0.17% of prompts reach ChatGPT search as typed. ChatGPT agrees with other models on

Source citations in US ChatGPT prompts rose from 1.6% (Jun 2025) to ~6.8% (May 2026). Travel cites ~23% of prompts, Automotive ~20%, Professional Services <4% (Similarweb).

First Page Sage July 2026: ChatGPT Gen AI market share dropped to 60.5% (from 74.8% June 2025). Gemini rose from 5.7% to 21.5%. Multi-engine optimization essential.

Talkory.ai (May 2026) tested 50 factual queries: Perplexity 85% citation accuracy, ChatGPT+Browse 76%, base ChatGPT 52%. Base model fabricates 30-40% of citations.

RanketAI (50 brands, Jan-Mar 2026). ChatGPT: direct-answer paragraphs + Bing. Claude: source attribution + verifiability. Gemini: E-E-A-T + freshness. FAQPage: 89% Gemini-cited vs 58% Claude.

Being listed as a source ≠ shaping the answer. Measure both selection rate (appearing) and absorption score (how deeply your language lands in the final answer).

RankScope 2026: top 5 domains = 38%, top 10 = 54%, top 20 = 66% of AI Overview citations. 26% of brands have zero AI mentions. Only 11% cited by both ChatGPT and Perplexity.

cloro's ~13,000-answer study: Google AI Mode averages 17.7 citations per answer vs Perplexity 1.3 (0.0 for dining). AI Overviews trigger on 98% of commerce queries but 3% of restaurant queries.

Digital Applied Jun 2026: Citation Reach Rate (CRR) averages 33%. 60% of AI-cited URLs never appear in Google Search Console or standard SEO rank trackers.

AI citation compounds across all channels. Cited brands see +35% organic CTR, +91% paid CTR. AI-referred visitors convert 4.4x higher than traditional search.

Seer Interactive (3,119 queries, 42 orgs, 25.1M impressions): brands cited in AI Overviews get +35% organic and +91% paid clicks vs uncited. Overall, AI Overviews cut organic CTR 61%.

CiteLens benchmark (320 buyer queries, July 2026). Google AI Mode cites top-10 organic at 93%, Perplexity 89%, Claude follows brand search demand (53%), ChatGPT ignores both (30% overlap).

OtterlyAI's own data reveals Claude drives 10.6% of signups (self-reported) vs 0.1% in GA4 last-touch. ChatGPT: 11% survey vs 7% GA4.

Claude and ChatGPT cited sources overlap just 8% for same prompts. Claude aligns with Google rankings 64%, ChatGPT does not. Each engine reads a different web.

Oltre AI analyzed 2,170 Claude-cited URLs: zero from Forbes, TechCrunch, NYT, WSJ, Bloomberg. Zero from Reddit, LinkedIn, YouTube. Claude favors niche SaaS blogs, documentation, practitioner content.

Meltwater May 2026: Claude's strongest model-source pairing is Statista at 21,620 citations. Claude prefers structured, authority-led, data-rich sources.

Presenc AI analyzed 3,800 Claude responses (Q1 2026): enterprise buyers face higher source-quality bar. Peer-reviewed sources weighted 8x more than marketing content in Claude's citation algorithm.

Only 11% of Claude buyer-research responses name a single brand. The dominant pattern (54%) is shortlist-without-ranking, shifting the optimization target from being

Metronyx AI synthesis: Anthropic holds 32% enterprise LLM market share vs OpenAI 27%. Claude dominates coding at 54%. Claude users convert at 16.8% vs ChatGPT 14.2%.

OtterlyAI analysis of 379,321 Claude citations: 0 Reddit citations, 64% brand domains, only 13% URL-level overlap with ChatGPT. Unique among all AI engines.

OtterlyAI 1M+ citation study: community platforms dominate with 52.5% of citations vs 47.5% brand domains. News sites 20.3%, community forums 5.9%.

Erlin tracked high-traffic prompts: 23% monthly citation churn, median recovery time 45 days. Competitive displacement causes 80% of lost citations. Monitoring cuts error discovery from 67 to 14 days.

Fractl surveyed 1,008 consumers and 150 marketers: trust fell sharply. 80% feel confident AI gives unbiased info, but 85% still double-check answers elsewhere.

Profound 240M-citation longitudinal tracking: content citation half-life ranges 3.4-5.7 weeks depending on platform. AI Overviews highest drift at 59.3% monthly. GEO is a continuity discipline.

ConvertMate's 2026 GEO benchmark: content updated within prior 30 days has 3.2x citation multiplier. Ahrefs corroborates: AI-cited URLs are 25.7% fresher than non-cited URLs.

From AI training crawl to click: 65.3M crawls narrowed to 6.8M answer fetches and 45,776 referral visits. ChatGPT delivered 86.8% of AI referral traffic.

Burson and Profound June 2026 study of 85 companies (55K+ believability forecasts) found being cited in AI answers does not guarantee being believed.

Same brand's citation volume differs up to 615x across platforms. Perplexity cites Reddit in 90%+ answers, Gemini in as few as 7%. Multi-platform tracking essential.

153,425 citations across 5,000 queries on 6 AI platforms (May 2026). ChatGPT cited-URL overlap with Google top-10: just 4.2%. Median cited sentence: 10 words. Freshness collapsed to 298 days.

DataDome network of 400+ companies: 17.7B AI agent requests in Q2 2026, up 45% from 12.2B in Q1. Meta's RAG crawler grew 163% QoQ. AI traffic now 57.4% of web requests.

arXiv (May 2026): 14 LLMs benchmarked. Frontier models: link validity 94%+, relevance 80%+, but factual accuracy only 39-77%. More retrieval (2→150 tool calls) drops accuracy by 42%.

Across 680M citations, directory listings are 54.53% of all distinct URLs engines cite, but a brand's own website is revisited 4.31x per URL vs 2.46x for listings (TLSOLS synthesis). Owned pages are

Muck Rack analyzed 1M+ AI citations: 82% from earned media, 94% from non-paid. Stacker/Scrunch controlled study: 3rd-party news distribution lifts AI visibility 239%.

Muck Rack's March 2026 research finds earned media = 25% of all LLM citations. Branded web mentions 3x stronger than backlinks for AI citation decisions. Most agency GEO retainers lack PR placements.

Earned media (journalism, editorial) drives 84% of AI citations across ChatGPT, Claude, Gemini. Paid/advertorial content = 0.3%. Only 2.9% of citations point to brand-owned domains.

5W Public Relations (July 2026, 40 queries, 6 categories on Claude): hard-paywalled publishers received 0% of AI citations. Open-web content captured 91.3%. Earned media drives 84% of AI citations.

Graphite's methodology for measuring entity position in AI responses: 10 responses gives ~1 position error. Position is probabilistic but measurable.

arXiv (April 2026): 9 models, 108K references tested. Author names fail most. Reasoning distillation degrades recall. Sparse FH-neurons identified. Suppressing them improves accuracy 6pp.

Nature Communications study (Wu et al., 2025) of 366,000+ citations finds most AI citations don't back the claims they're attached to. SourceCheckup achieves 88.7% expert agreement.

5W PR annual report (May 2026): 10K+ prompts, 15+ categories, 5 engines, year of tracking. Small number of brands capture majority of citations. Gap between leaders and challengers widening.

FogTrail analyzed 1,122 citation URLs across 100 engine-query pairs: ChatGPT links to brand sites in 24% of citations vs Grok 1.9%. Engines disagree on the #1 recommendation in 50% of queries.

Writesonic tracked 23M cited sources across AI platforms (Apr–Jun 2026): 44% of cited pages appeared in exactly one answer before vanishing. Median citation lifespan: 11-15 days.

Track AI visits in GA4 (custom channel for chatgpt.com/perplexity.ai/gemini) and run monthly 20–30 query × 3-platform prompt audit recording cited/mentioned/position. AI traffic converts ~14.2% vs

Post-Gemini 3 only 38% of AI Overview citations come from top-10 organic. 62% from deeper SERP. Fan-out sub-query mapping is essential.

SE Ranking analysis: Gemini 3 (Jan 2026) replaced 42% of cited domains and generated 32% more source URLs per response, creating massive citation volatility.

Google's Gemini app reached 750M monthly users as of early 2026. Combined with ChatGPT's 900M weekly users, AI search platforms reach audiences comparable to major social networks.

Omnia citation database (May-June 2026): 90.1% of Gemini citations go to third-party editorial, owned domains only 6.7%. Average 4.28 domains cited per answer.

MaxAEO tracked 1,200 prompts across 60 B2B brands (Apr-Jun 2026): brands cited in AI Overviews appeared in Gemini only 38% of the time. Separate systems, different picks.

Seer Interactive tracked 82,000 Gemini responses across 20 brands. Citation rate dropped from 99% to 76% between February and March 2026 — not from a model switch but a quieter change.

CiteMetrix 2026: Gemini cites Reddit in only ~5% of responses versus 44% for Google AI Overviews. Perplexity leads at 46.7%. A single Reddit strategy cannot serve all platforms.

Massive execution gap: only <15% of teams have formal GEO programs but 67% of Fortune 500 CMOs rank it top-3. GEO services market projected $850M to $17-20B by 2034.

A July 2026 arXiv survey of 45 GEO studies finds topical relevance and context position are the only reproducible levers; foundational gains conditional on fixed context.

Authority Tech's GEO Measurement Framework (Jun 2026): citation share, AI SoV, source attribution per page, prompt coverage rate, pipeline-influenced conversions. How to connect AI visibility to reven

Discovered Labs benchmarks: B2B SaaS citation rate starts at 5-15%. With systematic GEO, 40-50% within 6 months. Track Share of Model and citation share across 50-100 prospect queries.

OtterlyAI experiment tracked 15 deleted comparison pages across 7 AI engines. 35% of citations came after deletion. Claude most persistent (78%).

arXiv (2026): 56,381 papers, 2.2M citations analyzed. 1.07% contain invalid citations, +80.9% in 2025. LLM hallucination rates 14-95% across 13 models. Error propagation across papers documented.

WP SEO AI study: same site — Google AI 85% English, Bing AI (ChatGPT/Copilot) 56% German. A German 'best of' page got 776 Bing citations vs 2 Google AI impressions.

Google I/O 2026: AI Mode crossed 1B MAU. Query fan-out issues 8-12 parallel sub-queries per question. Between Apr-May 2026 AI Mode slashed cited URLs per response by 59% — fewer slots, higher competit

Google AI Mode now displays citation count + favicon below answers (June 2026). Zero-clicks hit 68% of US searches. Only 27.6% of clicks reach open web. Citation count becomes visible brand metric.

AI Mode grew 4x in 2 months to 75M daily users. Zero-click rate is 92-94% vs 35-46% for traditional Google. Only 1-3 sources cited per response.

OrganikPI May 2026 study: AI Mode #:~:text= fragment coverage fell from 70.9% (Mar 2026) to 0% across 88,392 citations. Gemini doubled down: 51.8% to 84.1%.

Ranking

Overlap between AI Overview citations and organic top 10 fell from 76% (mid-2025) to 38% (early 2026, Ahrefs) — as low as 17% per BrightEdge.

BrightEdge Feb 2026: Healthcare 88%, Education 83%, B2B Tech 82% AIO penetration. Seer: even queries without AIOs saw 41% CTR decline. Being cited = primary CTR preservation strategy.

A randomized field experiment on 1,065 US users (68,089 searches) found AI Overviews caused a 39.8% drop in outbound clicks and 33% more zero-clicks. Removing them did not change user satisfaction.

Google updated core update documentation July 9, 2026 confirming continuous smaller updates without public announcements. Semrush/Sistrix volatility data shows elevated movement June–July outside

July 2026: Google began surfacing AI visibility impressions in Search Console and Bing Webmaster Tools added an AI Citation Share metric - the first native windows into AI citations.

BeVisibleIQ Mar 2026 analysis of 2,391 citations across 4 models: GPT-5.4 cites vendor content at 74.6% vs ~79% third-party for older models. Trend toward own-site rewards.

Naser audited 10 LLMs generating 69,557 citations. GPT-5-mini hallucinates at 11.4% (down from GPT-4o-mini 45.3%). Multi-model consensus = 95.6% accuracy.

ChatGPT's free-tier default model upgrade caused brand website citations to halve, with Reddit surging to become the most-cited domain by 3x over any other source.

Goodie's analysis found Grok's algorithm updates every 8-10 weeks. Citation patterns shift with each update. Monthly monitoring is minimum cadence for active Grok optimization.

Natural experiment: when Grok API went offline, total Reddit AI citations collapsed 91% (35→3). Grok drives 60-77% of all Reddit citations across AI search.

Ahrefs June 2026 Grok citation ranking: Reddit 16.3%, YouTube 15.1%, Facebook 9.8%. Brand domains outside top 10. Grok used by 117M monthly users per SpaceX S-1.

Indexably analyzed 18,129 cited pages and 4,622 controls across 5 AI platforms. Domain authority (DataForSEO Rank d=1.075) is 7x stronger than any single page signal.

First quantified SEO vs GEO factor comparison (SSRN March 2026). Brand entity mentions (NIS 0.918) and evidence-bearing content (NIS 0.747) dominate GEO. Authority persists (DI +0.136).

SE Ranking analysis of ~300K domains: no statistically significant correlation between llms.txt and AI citation frequency. Removing the variable from the model improved prediction accuracy.

GrowthPro AI 2026 survey found 88% of local businesses have no active Generative Engine Optimization strategy. Early adopters gain disproportionate AI citation share in local search.

Visiby June 2026 benchmark (2,443 prompt-runs, 7 brands): median brand citation rate 11% ChatGPT, 9% Perplexity, 4% AIO, with up to 24-point swings between engines.

Meltwater's June 2026 AI Visibility Report (8M+ citations, 8 LLMs). Each model prefers different sources: Gemini institutional, Perplexity research, Claude data-centric sources.

AirOps research: brands that earned both a mention and a citation were 40% more likely to reappear across consecutive AI answers.

Seer Interactive's 6-week controlled experiment: altering meta descriptions produced −1.3% effect (95% CI: −5.9% to +3.3%) — a statistical zero on AI crawls and citations.

Graphite's research on n8n: GA4 captures 1% of AEO conversions, survey shows 9%. 90% of AEO-influenced users go direct — invisible to analytics.

Organic Labs synthesis of 19 studies + 6 case studies: FAQs/schema r=0.81, comprehensive content r=0.79, brand mentions r=0.76. Word count and backlinks show near-zero correlation.

19,556 queries tested across ChatGPT, Claude, Perplexity, Gemini. Jaccard similarity = 0.014. Each platform has dramatically different citation profiles. Multi-platform GEO is essential.

Analysis of 680M citations (5WPR/Profound, 2026): only 11% of domains are cited by both ChatGPT and Perplexity. Each platform uses fundamentally different citation logic, source preferences, and

Across 100,000 identical prompts, only 11% of cited domains appear in both ChatGPT and Perplexity (Profound). ChatGPT leans Wikipedia, Perplexity Reddit; AI-cited URLs are 25.7% fresher than organic.

19,556 queries tested across ChatGPT, Claude, Perplexity, Gemini. Jaccard similarity = 0.014. Each platform has dramatically different citation profiles. Multi-platform GEO is essential.

AgentVisibility.ai's 90-day study found just 16% of brands appear in ChatGPT, Gemini, Claude, Perplexity, and Copilot simultaneously — multi-platform presence is rare.

WhatsMyGeoScore analyzed 8,500 AI Overviews across Google AIO, ChatGPT, Perplexity, Claude. Only 41-54% from top 10 positions. Ranking

12,000 AI responses, ~200 brands, 221K URLs. Only 23.1% of mentions include citation. ChatGPT cites most (28.3%). Category queries: 7.2% mention-to-citation. Healthcare leads at 34.5%.

Ahrefs (863K keywords, 4M URLs): AI Overview citation overlap with organic top-10 dropped from 76% (July 2025) to 38% (Q1 2026). Query fan-out into sub-SERPs now drives most source selection.

Perea.ai's GEO 2026 paper synthesizes 100+ sources — Princeton/KDD, Profound 680M-citation dataset, SE Ranking 300K-domain study. Core thesis: citations compound as an asset each time a model emits

Qwairy analyzed 118,000 AI responses (Jan-Mar 2026): Perplexity 21.87 citations/response, ChatGPT 7.92, Google AI Mode 8.34, Claude 5.67. Perplexity gives 32% of citations to second-tier/niche sources

Perplexity cites brands at 13.05% vs ChatGPT's 0.59% — a 22x gap making Perplexity the most brand-friendly AI surface for earned media.

Perplexity's top 50 most-cited domains show Reddit at 20–24% share, brand citation rate 22x higher than ChatGPT, highest citation density of any engine.

Across 316 tracked keywords, 35% held a Perplexity citation vs 17% on ChatGPT and 6% in Google AI Overviews. Perplexity cites ~6.4 domains per answer vs ChatGPT 3.1.

ZenoX Media's 543-answer buying-intent study: Perplexity cites at least one source in 60.6% of answers, ChatGPT 32.6%, Gemini 43.5%. 34% of citation losses came from no web search at all.

Perplexity delivers inline citations that convert at 1.66% for sign-ups vs 0.15% from traditional search — 11x advantage. Highest B2B conversion platform despite 15-20% referral volume.

Presenc AI analysis: Perplexity first citation position receives nearly 5x the click-through rate of the fifth citation. Average 5.8 sources per answer (up from 4.2 in 2024). First citation position i

Attrifast benchmark of 200 Stripe-connected sites: Claude converts at 3.4%, Perplexity at 3.1%, ChatGPT at 2.5%. Deep blog pages with inline CTA convert 3.5x higher.

Semrush 325,000-prompt cross-engine study found Perplexity cites LinkedIn Company Pages 59% vs individual posts 41%, opposite to ChatGPT where individual posts dominate 59%.

SimilarWeb data shows Perplexity Pro subscribers generate 2.1× more search queries than free-tier users, averaging 7 queries per session with deeper exploration behavior.

Conductor 2026 AEO Benchmarks Report (13,770 domains, 17M AI responses): AI referral traffic converts at 1.66% sign-up rate vs 0.15% organic. Perplexity visitors convert at 10.5% (some studies show

Perplexity reached 230M MAU in Q1 2026 (184% YoY), cites 8.2 sources per answer on average, and its referral traffic converts at 3.1x the rate of Google organic.

Searchless analysis: Perplexity allocates 35% of citations to niche authority content vs 24% for Claude, 28% for Google AI Mode. Niche experts and specialized blogs genuinely compete with major publis

When AI summary is present, only 8% of users click any result vs 15% without. Being inside the summary equals existence; absence equals invisibility.

Geo-citation-lab, 21,143 citations: Perplexity cites 13.8 sources per query (breadth), ChatGPT 7-8 (depth). 80% of Perplexity-cited content not in Google top 10.

AI traffic converts at 14.2% average vs Google 2.8%. Each platform has distinct citation profiles requiring separate approaches.

Meltwater May 2026: press release citation share fell from 0.4% to 0.2% MoM. Earned editorial remains at 37.6%, making it 188x more valuable for AI visibility.

Previsible third AI Traffic Study (6.77M sessions, 166 sites, Nov 2024–May 2026): ChatGPT 92.4% of standalone LLM referral traffic. E-commerce AI referral rose 37x. AI discovery inside Google > all

Discovered Labs 2M-citation study: prompt-content alignment (β=+0.37, q≈10⁻⁷³) dominates. Standard AEO checklist effects reverse with domain fixed effects (Simpson's paradox).

Trakkr (33.69M citations): social/UGC sources are ~1% of AI citations; Wikipedia leads at ~15%. Contradicts popular 'Reddit is most-cited domain' claims — mix is corpus-dependent.

SE Ranking analysis of 216K pages found referring domains are the strongest predictor of ChatGPT citations. Sites with 350K+ RDs average 8.4 citations; under 2,500 RDs average 1.6.

Mike King's framework: cross-functional team (engineer + content + UX + PR + SEO), 3-bucket measurement (input/channel/performance), $26M client value.

DataForSEO analysis of 100,249 fan-out queries: Wikipedia tops cited sources at 25.3% share while ranking only 106th in AI search results, separating discovery from citation-worthy content.

Brands with 50+ recent reviews on G2/Capterra cited at 3.2x rate of brands with <20 reviews. Single highest-impact GEO activity across case studies.

Am I Cited Jan 2026: review velocity correlates r=0.72 with AI citations vs r=0.31 for average star rating. 50+ reviews minimum threshold. Incentivize review flow, not just scores.

Semrush scaled from 2,500 to 126M AI search prompts. Only 36 'Universal 36' brands maintained visibility across all 4 platforms. Gemini shows just 30% overlap between mentioned and cited brands.

Semrush 2025 analysis: average AI search visitor is 4.4x as valuable by conversion rate vs traditional organic. Wix AI Search Lab: AI search visits grew 42.8% YoY to 27.4B in Q1 2026.

Semrush 2025 analysis: average AI search visitor is 4.4x as valuable by conversion rate vs traditional organic. Wix AI Search Lab: AI search visits grew 42.8% YoY to 27.4B in Q1 2026.

Semrush scaled from 2,500 to 126M US AI search prompts; only 9% of marketing leaders have tools to track all AI visibility metrics while 45% cannot measure at all.

Semrush + Kevin Indig study: 3,981 domain appearances across 14 countries, 4 engines. 62% ghost citations (link without brand name). ChatGPT: 87% cite rate, 20.7% mention rate.

Semrush's 126M-prompt AI Visibility Index: only YouTube, Google, Reddit, Amazon, Apple and 31 others maintain top-100 across ChatGPT, Gemini, AI Mode, AI Overviews.

Mixed-effects logistic regression on 114,034 observations. Schema markup strongest content predictor (OR=1.31). 75.4% of aggregate citations from outside top 30 but this is denominator artifact.

Machine Relations (71% single-engine), SurfacedBy (69.6%, 2.7% all-five) and Wellows 22.7M-citation study (79.6%, 0.31% all-five) all confirm AI sources are single-engine exclusive; optimize per

Siftly aggregated citation rate benchmarks: B2B SaaS 18–28%, E-commerce 12–22%, Local Services 15–30%, Healthcare 10–18%, Finance 8–16%, Travel 20–35%. Judge your rate relative to industry median, not

Chartbeat March 2026 data: small publishers (1K-10K daily views) lost 60% of search referral traffic over 2 years vs 22% for large publishers (100K+). AI referrals <1% of pageviews.

Brand mentions correlate 0.664 with AI citation vs 0.218 for backlinks. Store-bought backlinks and aggressive link building show near-zero effect on AI citations.

Presenc position analysis: relative CTR is 100 for the first ChatGPT citation, 68 for the second, 44 for the third and just 13 for the fifth and beyond.

Analyze AI study of 83,670 citations: 82.9% from third-party (reviews, news, Wikipedia). Capterra leads. ChatGPT cites own sites least (13.5%). Brands 6.5x more likely cited externally.

Eight Oh Two 500-user study: 37% start with AI tools. 62% want quick summaries. 60% say AI gives clearer answers. 85% double-check with traditional search.

Similarweb's three AI-presence KPIs: citation frequency (how often AI names your content), share of model, and branded search lift after AI exposure.

CITETRACE (May 2026): 11,200 queries, 112K responses, 761K citation pairs across 10 models. 30.6% citations distort source meaning. 96% of users encounter at least one misleading citation per

Webflow analyzed 2,000 websites: median company appears in 16% of relevant AI answers and in a citation with a link only 6% of the time. Large companies 2× higher mention rates.

WebFX analyzed 590K AI sessions across 2,500 URLs in 15+ industries. Homepages drove 31% of AI traffic. 92% landed on consideration/decision-stage content. ChatGPT 97.5% of referrals.

Wikipedia reached 83,191 citations in May 2026, growing 55.2% MoM with visibility score up 50% and citation rate improving 14.3% across major LLMs.

Geonimo's 50,000-response analysis: the top 10% of brands take 68% of AI mentions while the bottom 50% receive under 5% — AI search is concentrated.

Yext's 17.2M citation study found verified structured data accounts for 54.53% of distinct citation sources across ChatGPT, Perplexity, Gemini, and Claude.

YouTube generated 188,863 AI citations in May 2026, growing 56.4% MoM — the most-cited single source across all major LLMs per Meltwater.

Google AI Mode ends 93% of sessions with zero external clicks (Quantumrun). Overall 60% of Google searches end without a click; cited brands earn 35% more organic and 91% more paid clicks.

Zero-click rate rose from 60.45% (2024) to 68.01% (2026). AI Overviews ~83% zero-click, AI Mode ~93% zero-click. Traditional Google ~60%.

Randomized field study proves AI Overviews directly reduce organic click-through rates. Gartner projects 25% of organic search traffic will shift to AI chatbots by end 2026.

First meta-analysis scoring 23 AI citation factors by evidence strength across 54 experiments. URL Accessibility