AI Overviews (AIOs) now appear on nearly 65% of question-based searches, which are the same queries informational content is built to win. The citations inside the AIOs represent one of the most meaningful positions in the SERP, as they are the closest element we have to a first position link or a featured snippet. It's the new position zero. And that makes it prime real-estate for Google search visibility.
The SEO playbook remains largely unchanged, and Google confirmed this with the recent release of Google's Guide to Optimizing for Generative AI Features on Google Search. This study was conducted to see if the first citation in AIOs aligns with standard SEO best practices, or if the current narrative is overgeneralizing the new state of search.
This study looks at which pages Google’s AI Overview (AIO) actually cites in the first-citation slot and what those pages have in common.
The Goal: Determine which SEO factors correlate with earning the #1 citation slot in AI Overviews.
The Methodology
-
Generated 214,056 candidate keywords across 30 industries and 9 intent shapes (definitional, how-to, comparison, best-of, troubleshooting, etc.).
- Verified search volume with DataForSEO, kept 18,260 with ≥50 monthly searches.
- Drew a stratified sample of 8,500 keywords and captured AI Overviews via SerpAPI (May 7–13, 2026) 1.
- Found 7,225 AIO winners and crawled 6,354 of those winning pages for on-page SEO signals: word count, headings, schema, link graph, freshness, E-E-A-T (Experience, Expertise, Authoritativeness, and Trustworthiness)
- Ran two rounds of in-house validation (105 PASS/0 FAIL) plus a 6-day headed-browser drift check against live Google. No SerpAPI bugs found; trigger-level findings stable.
Our Expectations:
Going into this analysis, our prevailing hypothesis was that the AIO citation slot would reward the same signals that classical SEO has always rewarded.
We expected the "winners" to be the sites with: heavy schema markup (FAQ, HowTo, Article), strong E-E-A-T signaling through author bios and credentials, and long-form comprehensive content. We thought that AIOs would just be a slightly tighter version of the existing SERP, where the same brands and the same on-page patterns came out on top, but...
That's not what we found.
The two signals we'd weighted heaviest going in, schema and E-E-A-T, turned out to be the two that most clearly didn't correlate with first-citation wins. Sites with the most FAQ and HowTo schema underperformed by roughly 10x. The cohort with the highest author-bio coverage had the lowest AIO citation share.
Most of the findings below contradicted what we were expecting, which is the reason the rest of the analysis was worth doing.
3 Key Takeaways: What Gets You the #1 AIO Reference
- Reddit is the new authority and remains as such.
- Reddit gets cited 10× more than Forbes, NerdWallet, Investopedia, and a dozen other big publishers combined. Another blog post won't beat a Reddit thread. Build a real presence in the subreddits where your customers actually hang out.
- A good portion of the SEO playbook isn't earning citations anymore.
- Author bios, FAQ schema, 5,000-word ultimate guides…the sites doing all of that hardest are losing to sites doing none of it. Article schema and breadcrumbs are enough.
- The opportunity is where SEO money hasn't been.
- AIO mostly fires on long-tail queries, and the industries where it's most active and least locked-up — logistics, career, fitness, pets, energy, manufacturing — are exactly the ones to go after.
The nine findings below break the headline down, and more importantly, show where each one contradicts standing SEO orthodoxy.
Where AIO Rankings Contradict SEO Best Practices:
Reddit beats every "real" publisher 10× over
Reddit has no schema, no E-E-A-T, no authority signals. SEO best practices didn't need to weight Reddit as a priority.
But looking at AIOs, Reddit captured 20.4% of all first-citation slots.
The 14 textbook-optimized publishers tracked (Forbes, NerdWallet, Bankrate, Investopedia, Wirecutter, The Spruce, RD, PCMag, CNET, TechRadar, Tom's Guide, Digital Trends, Good Housekeeping, AllRecipes) captured 1.94% combined. Reddit alone has 10.5× more AIO citations than that entire cohort.
Major news is functionally absent
Classical news authority equals untouchable ranking authority... or so the thinking went.
NYT + WSJ + BBC + Forbes + Reuters + Bloomberg + Wired + Verge combined = 0.6% of first-citation slots.
Publishers the SEO industry has long deferred to as "untouchable authority" are a rounding error in AIO.
E-E-A-T author bios don't help AIO
SEO logic was: author bios + credentials + bylines = trust = ranking.
But the cohort with the highest author-bio rate (Perfect SEO publishers, 76%) has the lowest AIO share (1.94%).
The cohort with zero author bios (Reddit) dominates at 20.4%.
It wasn't even a neutral correlation, it was an inverse correlation one.
FAQ + HowTo schema is irrelevant for AIO
Add FAQ + HowTo schema for visibility (Google's own docs even say so!)
Except Reddit (0% FAQ schema) dominates. Wikipedia (1% FAQ) wins definitional.
"Perfect" SEO publishers (69% FAQ schema) underperform 10×.
Article + Breadcrumb appears sufficient; the rest is wasted markup as far as AIO is concerned.
The 5,000+ word "ultimate guide" captures only 4.4% of definitional slots
Long-form wins; Skyscraper content beats short content. More words = more authority.
But definitional winners are bimodal: ~25% are under 250 words (dictionary entries, government Q&A), another ~25% are 1k–2k (medium explainers). The Skyscraper-style ultimate guide captures 4.4%.
The target is "be quotable," not "be comprehensive."
AIO triggers less for high-volume head terms
It's been considered as a conventional approach to optimize for head terms because of search volume × CPC; long-tail is just leftovers.
In AIO, mean search volume for AIO winners: ~6,100/mo. Mean SV for no-AIO keywords: ~178,000/mo.
Head terms ("python," "amazon," "youtube") trigger AIO ~30× less than long-tail.
SV-weighted trigger rate is 16.24% vs unweighted 85%2.
Reddit still wins…big3
Keep Reddit and your blog separate. Reddit is for users, your blog is for ranking.
But for topics where Reddit has 15–33% AIO share (software engineering, food, gaming, e-commerce, parts of finance), publishing another blog post is fighting the wrong battle.
A real Reddit presence (substantive answers under a real account in relevant subreddits) may produce more AIO citations than another blog post.
Industry opportunity is inverse to where SEO money is
Follow the CPC: high-cost verticals are where the SEO opportunity is.
Classical SEO money flowed into the monopolized verticals because of high CPC; AIO opportunity is the inverse.
AIO-monopolized: personal finance (39% Fidelity), DIY (32% YouTube), health/medical (30% Cleveland Clinic), software engineering (33% Reddit). Diffuse — top-1 incumbent <20%: logistics, career/jobs, fitness, pets, energy, manufacturing.
AIO is a broad citer — 11 winners per query, not 1
Position 1 is everything. Position 2 is half the clicks. Position 10 is a rounding error.
But AIO cites a mean of 11.4 sources per query (median 11, p90 16, max 32).
Position-zero is the prize, but positions 2–11 are real exposure.
SEO tools have started tracking "AIO position" as a separate metric for this reason.
What This Means For Content Strategy
| Lever | Action |
|---|---|
| Intent targeting | Prioritize definitional, how-to, and comparison query shapes. Trigger rates 95–98%. |
| Content shape | Be quotable, not comprehensive. Aim for the 1k–2k word band, or under 250 words for dictionary-style entries. Avoid the 5k+ ultimate-guide format. |
| Internal links + outbound authority | Definitional winners average more internal links and cite ~2× more .gov/.edu sources outbound4. Build the link graph; cite real authority. |
| Schema + author bios | Don't over-invest. They aren't moving AIO citation rates. Article + Breadcrumb is fine; FAQ/HowTo schema appears irrelevant for AIO. |
| Where to play | Diffuse-incumbent industries: logistics, energy, career, fitness, pets, manufacturing, crypto, AI/ML. Avoid head-on competition with Fidelity (finance) or Cleveland Clinic (medical). |
| Reddit & YouTube as a channel | Where Reddit has 15–33% AIO share, a substantive presence under a real account may produce more citations than another blog post. Treat UGC platforms as a distribution channel, not a competitor. |
| Track AIO positions 1–11 | AIO cites ~11 sources per query. Track citation share across positions, not just the first slot. |
The Question Behind the Data
All of my thoughts surrounding this study point to something that I don’t know how to prove:
Do AI platforms have a structural incentive to surface and reward the kind of content they need for training?
But if the answer is even partially a yes, it reframes everything.
It means the signals Google built its original ranking system around (authority markup, structured schema, long-form comprehensiveness) may be exactly the signals an AI system learns to discount, because optimized content is less useful as training data than real human interaction.
That would mean the best thing you can do for AIO visibility is being more human, more specific, and more willing to say something that really matters in the places where your customers already are.
1 US/English only, May 7–13, 2026 snapshot. AIO behavior shifts over time.
2 The 85% trigger rate is the sample-shape number. Search-volume-weighted, it's 16.24% — head terms still get traditional SERPs.
3 Reddit and YouTube can't be deep-crawled (auth wall / JS SPA), so the analysis shows both "incl R/YT" and "excl R/YT" baselines wherever the cohort choice matters.
4 The +69% internal-link gap is the most cohort-sensitive number in the report — direction is robust, magnitude depends on which denominator you use.
Matt Buxbaum
SEO Associate
.png)
