Last quarter, an agency owner sent us a screenshot. Their client ranked #1 on Google for the category-defining query their CMO cared about most. The same query, run in ChatGPT 90 seconds later, returned three brand recommendations — and their client was not one of them. Different engine. Different answer. Same buyer.
That gap is not an outlier anymore. It is the new normal, and the data is now firm enough to settle the argument. Per Ahrefs' February 2026 analysis, only 38% of pages cited in Google AI Overviews rank in the Google top 10 for the same query. Another 31% rank between positions 11 and 100, and 31% rank beyond position 100 entirely. Per Profound, 28.3% of ChatGPT's most-cited pages have zero organic Google visibility. The two systems are no longer measuring the same thing.
This post lays out what the published research now shows about Google rank vs AI citations, why the systems decoupled, and what agencies should do this quarter to keep their clients on the buyer's Day-One AI shortlist — the place where, per the 6sense 2025 Buyer Experience Report, 95% of deals are won.
Start your 14-day free trial
Growth plan free for 14 days. Five AI engines. Full agency dashboard.
Start free trialThe decoupling, in one chart's worth of numbers
Three numbers, each from a published source, sit on top of every conversation worth having about modern client reporting:
That third number matters more than agencies have priced in. Per Ahrefs' December 2025 update, AI Overview trigger rate climbed from 31% to 48% in twelve months, and position-1 organic CTR drops 58% when an Overview appears. The page that ranks first when no Overview is present collapses in click-through the moment the Overview lands above it. Being first in Google is not the same outcome it was two years ago, and the trajectory is steepening.
The flipside is that being cited inside the Overview is disproportionately valuable. Per Seer Interactive's September 2025 analysis of 3,119 informational queries, citation in an AI Overview is worth 35% more organic clicks and 91% more paid clicks than not being cited. AI-sourced visitors also stick around: per Conductor's 2026 benchmark, AI search visitors spend 68% more time on the website than organic search visitors. Smaller volume, higher intent, harder to win, and worth more per visit.
The buyer-side signal is even sharper. Per the 6sense 2025 Buyer Experience Report, 94% of B2B buyers use LLMs during their purchasing journey, 95% buy from the Day-One shortlist, and 83% of the journey happens before any sales contact. Per Loamly's February 2026 analysis of 2,089 brands, 77% are completely absent from AI platform responses, and the 23% that are present convert AI-sourced traffic at 3× the rate of Google search. Smaller pool. Higher conversion. Higher stakes. The clients still being measured solely on Google rank are being measured on the wrong layer.
What predicts AI citations (and what doesn't)
If Google rank is no longer the leading indicator, what is? The published research is now consistent on three signals, in roughly this order of measured effect.
1. Earned brand mentions across trusted publications
Per RivalHound's correlation analysis, brand mentions correlate 0.664 with AI visibility while traditional backlinks correlate 0.218. That is roughly a 3:1 advantage for unstructured mentions over structured links, which inverts most agency playbooks built on link-building. Per ZipTie's study, domain authority from earned mentions outweighs schema markup by approximately 3.5:1 in citation probability. The lever the SEO industry under-invested in for a decade is now the largest AEO lever, and it compounds slowly enough that starting six months early is the difference between in-the-answer and out-of-it.
2. Reddit, YouTube, and niche-source presence
AI engines pull from a wider citation surface than Google's SERP. Per Discovered Labs' analysis, Reddit accounts for 46.7% of Perplexity's top-10 citations. Per CMSWire, 73% of AI product recommendations referenced Reddit in 2025. YouTube is cited in 23.3% of AI Mode answers; Wikipedia in 18.4% (per Ahrefs). Per Search Engine Land, the citation surface is structurally different from Google's and that difference is permanent.
The detail that matters most for agencies: per Semrush's study of 248,000 AI-cited Reddit posts, more than 80% of cited Reddit content has fewer than 20 upvotes or comments. The bar is lower than agencies assume. AI engines are pulling moderate-engagement comment chains, not viral threads — which means a small sustained presence in the right subreddits is feasible for an agency to deliver, and it earns disproportionate visibility on the engine where it matters most (Perplexity).
3. Content structure tuned for chunk extraction
Per Am I Cited, sections in the 100-150 word range receive ~4.7 citations per page vs 4.3 for sub-35-word sections. Per Frase, pages with FAQPage markup are 3.2× more likely to appear in Google AI Overviews. Per AI Boost, FAQ schema combined with inline citations is weighted approximately 40% higher in ChatGPT source selection.
What does not work: per SE Ranking's analysis of approximately 300,000 domains, llms.txt has zero correlation with AI citations. And per Growth Marshal's study, generic schema actually underperforms no schema (41.6% citation rate vs 59.8%) — the schema lever only works with attribute-rich Product, Review, or specific FAQPage markup at 61.7%.
If you only have bandwidth for one lever this quarter, work on earned brand mentions in trusted publications. The 3:1 effect size over backlinks reorders most agency playbooks, and the lever compounds slowly enough that starting six months early is the difference between in-the-answer and out-of-it.
Why the systems decoupled
The decoupling happened for three reasons, all measurable.
Reason 1: Different source pools. Google's algorithm crawls and ranks the open web. AI engines pull from training data plus retrieval-augmented sources that are heavier on Reddit, Wikipedia, YouTube transcripts, and niche publications. The citation surfaces overlap, but they are not the same surface, and the gap is widening as AI engines invest in their own retrieval pipelines.
Reason 2: Different signal weights. Backlinks correlate 0.218 with AI citations; brand mentions correlate 0.664. Google's historic algorithm leans hard on link graphs. AI engines lean hard on textual brand-mention frequency in their training corpus and retrieval results. The same brand can rank #1 in Google because of inbound links and #11 in ChatGPT because nobody is talking about them in plain prose.
Reason 3: Per-engine variance. Per Profound's public data, ChatGPT mentions brands in roughly 73.6% of answers while Claude mentions brands in 97.3%. The same brand can be #1 on one engine and invisible on another. A single “AI visibility score” averaged across engines hides this. The GenPicked Research Team's 2026 Fitness Wearables Study (Bradley-Terry maximum-likelihood estimation, four models) documented Oura ranking #1 on GPT-5 (1.91) and Claude 4 (1.74) but #3 on DeepSeek V3 (1.12). Same brand, three different rankings depending on which engine the buyer happens to open.
That same study also documented that Claude is approximately 6.7× more reactive to brand anchoring than GPT-5 in sycophancy conditions — meaning a brand that wins on aided prompts (where the user names the candidate vendors) can lose badly on unaided prompts (where the user asks the open category question). For agencies, the strategic implication is direct: track unaided and aided performance separately. The gap between them is itself a diagnostic about how much of the client's visibility depends on buyer recall vs AI surfacing.
The agency reporting upgrade
If the metric is no longer average position, what replaces it? Four reporting changes earn the retainer this quarter.
The fifth quiet change: fix attribution. Per Coalition Technologies, only about 0.5% of ChatGPT-sourced traffic is correctly classified as “organic” in GA4. The rest gets bucketed as “direct” or vanishes into last-touch. Per Yotpo's tracking guide, custom landing-page filters and UTM hygiene can recover 3-5x the AI attribution that default GA4 reports show. If your AEO retainer is producing pipeline and your reports show “direct,” the work is invisible to the client. Fix attribution before you fix anything else.
Common objections (and the data answer)
Three objections come up on every agency owner call. Each has a clean data answer.
“My client ranks #1 in Google. They'll be fine on AI too.” Per Ahrefs, only 38% of AI-cited pages rank in Google top 10. Per Profound, 28.3% of ChatGPT's most-cited pages have zero Google organic visibility. Google rank is a weak proxy. Run the per-engine check and be ready for the gap.
“We're going to add llms.txt and call it AEO.” SE Ranking studied 300,000 domains and found zero correlation between llms.txt and citations. Adding it costs nothing and is harmless, but it is not a strategy. Lead with brand mentions and structured chunks instead.
“We'll just optimize one site-wide AI visibility score.” Engines disagree. Per Profound, ChatGPT brand-mention rate is 73.6%; Claude is 97.3%. The GenPicked Fitness Wearables Study documented the same brand ranking first on two engines and third on a third engine. An averaged score is statistically a lie of omission — it hides the actionable finding. Always split by engine.
What to do this week
Time investment for ten clients is roughly an afternoon. Cost is zero. The client whose Google-rank-vs-AI-citation gap is largest is also the client whose retainer is most at risk — lead with that one.
At ten or more clients this stops being a manual job. GenPicked's Growth plan ($197/month) automates per-engine citation tracking across all five engines, runs daily sweeps, computes the ACS (AEO Citation Score) per brand, and produces white-labeled monthly reports the CMO actually opens. Built for agencies running this at scale.
The vendor landscape (and which signals to trust)
Agencies are being pitched by a dozen AEO vendors right now. The funding signals from the past 12 months are unambiguous: per TechCrunch, Peec AI raised a $21M Series A in November 2025 with 1,300+ brands and agencies onboarded. Per Profound's announcement, Profound raised $96M Series C at a $1B valuation in February 2026, with 700+ enterprise customers including 10%+ of the Fortune 500. The category is being capitalized hard, which means the agency-side procurement window is closing on opportunistic adoption and opening on competitive necessity.
Three signals are worth trusting when evaluating any vendor in this space. First, per-engine reporting. If the vendor demos a single averaged “AI visibility score,” the methodology is structurally hiding the strategic finding. Second, confidence intervals on rankings. Per the Conductor State of AEO/GEO Report, methodology rigor is the single biggest differentiator in vendor selection. Third, source-pool transparency. The vendor should be able to tell you which engines they query, how often, and how they classify a citation (text mention vs URL attribution vs both). If the answer is fuzzy, the data downstream of it is fuzzy.
Start your 14-day free trial
Growth plan free for 14 days. Five AI engines. Full agency dashboard.
Start free trial