What Gemini Citation Data Tells Us About Winning Google AI Overviews
AI Overviews — the AI answer boxes at the top of Google Search — are powered by the same Gemini models we analyzed across 190 standalone Gemini citations. The citation patterns look nothing like ChatGPT and only partially like Perplexity. YouTube dominates. Amazon appears. Google-owned properties cite other Google-owned properties. Here is what the data actually says about winning those answer boxes.
What we tested, and what transfers
Our study measured standalone Gemini (gemini.google.com) citations — not Google AI Overviews directly. The two surfaces share the underlying Gemini model but draw from different indexes. AI Overviews pulls from the Google Search index with additional authority signals applied. Standalone Gemini has somewhat more freedom to reach for video, social, and platform content.
That said, the directional patterns transfer strongly. YouTube dominance, Google-ecosystem preference, and the way Gemini grounds structural quality against platform authority all appear consistently across both surfaces based on available field observation. Treat the specific citation counts as standalone Gemini data; treat the patterns as a strong proxy for AI Overviews behavior with higher authority weighting applied.
Gemini is the platform-ecosystem engine
Across 190 Gemini citations, the source pattern differs clearly from the other engines. Gemini pulls heavily from Google-owned and platform-scale properties — most notably YouTube — while still rewarding well-structured brand pages when they exist. It is the only engine in our study willing to cite Amazon product listings and Google’s own surfaces as authoritative sources.
The top-cited single domain is YouTube with 24 citations — nothing else is close. The next highest domain is Yaskawa (8 citations), then Wolverine, Amazon, Ariat, and Terex at 4 each. That is an unusual top-10 for an AI engine: a video platform followed by a mix of industrial brands and consumer product manufacturers.
What makes Gemini different from ChatGPT (which concentrates on brand primary domains) and Perplexity (which reaches broadly into Reddit, LinkedIn, and financial aggregators) is Gemini’s preference for Google-adjacent platforms: YouTube, Amazon, Google Business Profile pages, Google Maps results, Google Shopping listings. These surface in Gemini citations at rates the other engines do not match.
YouTube is not optional for Gemini visibility
24YouTube is the single most-cited domain in the Gemini dataset — more than Yaskawa (8), more than Wolverine (4), more than Amazon (4), more than any brand primary domain. It represents 12.6% of all Gemini citations. When combined with direct google.com citations (3 more), the Google-owned share of Gemini’s citation surface reaches about 14%. No other engine in our study came close to this pattern.
This is not accidental. YouTube is owned by Google. The Gemini model was trained with access to video metadata, descriptions, captions, and engagement signals at scale. When Gemini needs visual, procedural, or demonstrative content — which it does more often than ChatGPT — it reaches for YouTube first.
The content patterns that drive YouTube citations in our dataset are specific, not generic:
- Model-specific walkthroughs — videos titled with exact product numbers and technical specifications
- Problem-specific how-to content — “how to diagnose X” or “how to replace Y” framing
- Technical demonstrations — showing physical procedures, installations, troubleshooting
- Accurate closed captions — this is the hidden layer. Gemini extracts from captions more than from titles or descriptions.
Generic brand videos — overview reels, trade show footage, thought-leadership talks — do not appear in Gemini citations at meaningful rates. The content that gets cited is specific, technical, and captioned.
What to do about it
A YouTube presence is not a nice-to-have for Gemini or AI Overviews visibility. It is one of the two or three highest-leverage investments available. For B2B and industrial brands skipping video entirely, there is a measurable citation gap you are volunteering to keep open.
Google properties cite Google properties
Google.com itself appeared 3 times in the Gemini citation pool — and zero times in ChatGPT or Perplexity. Gemini is the only engine in our study that cites Google’s own surfaces as authoritative sources. Add YouTube’s 24 citations, and the Google-owned share of Gemini’s citation surface is about 14%.
For AI Overviews specifically, this pattern intensifies. AI Overviews are surfaced on Google Search result pages and have direct access to Google’s structured data layer — Knowledge Graph entities, Google Business Profile data, Google Shopping listings, Google Maps entity information. Content that lives inside the Google ecosystem has structural advantages content outside it does not.
Google Business Profile
For any business with a physical location or local service area, a complete and verified Google Business Profile is non-negotiable for AI Overviews visibility on local queries. This is not just about Maps — it feeds directly into the entity data that AI Overviews uses to ground local answers.
Google Knowledge Graph entity
Establish your brand as a recognized Knowledge Graph entity. This happens through a combination of Wikidata entries, Wikipedia presence where eligible, schema.org Organization markup with a complete sameAs array, and consistent NAP data across major directories. Brands with Knowledge Graph entities receive more consistent AI Overviews citation authority than brands without them.
Google Shopping for product brands
If you sell physical products, your Google Shopping presence matters for AI Overviews product queries. Product feed accuracy, structured product schema on your site, and GTIN/MPN data all feed into the entity layer AI Overviews draws from.
YouTube channel verification and activity
Beyond just publishing videos, a verified YouTube channel with consistent uploads signals entity legitimacy to the broader Google entity graph that AI Overviews relies on.
Amazon appears in Gemini, barely anywhere else
Amazon.com was cited 4 times in the Gemini dataset, 2 times in ChatGPT across 130 citations, and 0 times in Perplexity across 226 citations. Gemini is effectively the only major engine in our study that treats Amazon product pages as a meaningful citation surface.
The Amazon citations in our dataset appeared primarily in consumer product queries — specifically work boots and related industrial apparel. For commercial-only B2B brands with no Amazon presence, this lever does not apply. For any brand with a consumer or mixed-channel component, Amazon optimization matters for AI Overviews visibility in ways it does not matter for other AI engines.
Specifically: complete product details, accurate specifications, high-quality imagery with alt text, A+ content with structured information, verified customer reviews, and consistent GTIN/ASIN data. These are the same signals that drive Amazon’s own search ranking, but they also feed into Gemini’s citation layer through the Google ecosystem.
Gemini rewards structure — and adds a platform layer on top of it
A common misconception about Gemini is that it tolerates weaker-structured pages than the other engines. The data pushes back on that. Gemini’s median Page Structure Score across cited pages is 66, compared to ChatGPT’s 70 and Perplexity’s 65. Gemini sits between the other two, not beneath them.
Looking at the distribution: 45% of Gemini citations score PSS 70 or higher — strong-structure content. Only 13% score 20 or below. The shape is not a barbell of “strong brand pages plus weak platform junk.” It is a distribution clustered around solid structural quality, with YouTube and Amazon adding a second lane on top of the brand-domain foundation rather than replacing it.
Implication for own-site content
Your brand primary domain pages should still target PSS 70+ for AI Overviews visibility. The structural pillars — answer extraction, content formatting, technical readability, schema markup, topical alignment — work the same way on Gemini as they do on ChatGPT. Pages that score PSS 80 on ChatGPT tend to score well on Gemini too.
Implication for off-site content
For content hosted on platforms Google recognizes as authoritative (YouTube, Google Business Profile, Amazon), the PSS bar is lower. A YouTube video with a weaker description still gets cited if the content is relevant and the captions are accurate. An Amazon product page with minimal schema markup still appears for product queries because Amazon’s platform authority carries the citation.
This is a second lane of investment available for AI Overviews that is not available for ChatGPT. Strong brand domain pages remain the foundation. Platform content adds a layer ChatGPT cannot access.
The 30/60/90 AI Overviews playbook
If you are starting from zero, this is the sequence that builds AI Overviews visibility fastest based on what the Gemini data actually rewards.
First 30 days — Google ecosystem foundation
- Claim and fully complete your Google Business Profile. Every field, all photos, services, hours, attributes.
- Verify your YouTube channel and make sure the channel About section, links, and branding are complete.
- Add Organization schema with a full sameAs array linking to Wikidata, LinkedIn, and YouTube.
- Audit your Google Shopping feed (if you sell products) for accuracy, GTIN/MPN data, and image quality.
Next 60 days — content investment
- Produce 5-10 YouTube videos with specific, descriptive titles covering model-specific or problem-specific topics.
- Add accurate closed captions to every video — this is the extraction layer Gemini actually reads.
- If you sell physical products, audit your Amazon product pages for A+ content, accurate specs, and review integrity.
- Add Product, Service, FAQPage, or HowTo schema (as applicable) to priority pages on your brand domain.
Ongoing — compounding presence
- Publish 2-4 YouTube videos per month with technical depth and specificity.
- Keep Google Business Profile updated with posts, Q&A responses, and recent photos.
- Rescore priority brand pages for PSS quarterly; target moving pages from the 40s into the 70+ range over 2-3 quarters.
- Monitor AI Overviews for priority queries monthly to track which of your properties are being cited.
What does not move AI Overviews
Ignoring YouTube
This is the single biggest gap most B2B brands have for AI Overviews visibility. A brand with no YouTube channel, or a YouTube channel stocked only with generic marketing content, is voluntarily sitting out of the roughly 13% of Gemini’s citation surface that video occupies.
Uncaptioned video content
Publishing YouTube videos without accurate closed captions means Gemini cannot extract the content for citation. Auto-generated YouTube captions are usable but imperfect; manually reviewed or professionally captioned content performs significantly better.
Neglected Google Business Profile
Local businesses skipping Google Business Profile completion miss one of the highest-leverage AI Overviews levers for geographic queries. This is not a 2015 Local SEO concern — it is an active 2026 AI Overviews citation source.
Amazon presence with weak listings
For product-selling brands, having Amazon listings without A+ content, accurate specifications, or review integrity means you exist on Amazon but do not benefit from it in Gemini citations. Either invest in listing quality or treat Amazon as out of scope.
Ignoring the Google ecosystem entirely while investing in ChatGPT AEO
The most common pattern we see: brands doing substantial schema markup and content structure work for ChatGPT visibility, with no corresponding Google ecosystem investment. That strategy optimizes for one engine and leaves another equally large one on the table. AI Overviews operates inside the Google ecosystem. You cannot win it from outside.
Continue the series
The pillar guide
The three-lane AEO framework. Why single-engine strategy covers only one-third of citation surface area.
Read the pillar →Dominate ChatGPT
The authority engine. Brand primary domains, Wikipedia entity presence, and schema-rich page structure.
Read the guide →Dominate Perplexity
The platform engine. YouTube, LinkedIn, financial aggregators, Reddit, and Facebook as five surfaces.
Read the guide →Carrier HVAC case study
How architectural choices — a dedicated subdomain with depth — drove 43% owned citation share for Carrier.
Read the case study →The earned media myth
Why owned content dominates AI citations across 547 data points and 16 verticals.
Read the analysis →Audit your AI Overviews readiness
Tampa Web Technologies maps your presence across Google ecosystem properties — YouTube, Google Business Profile, Google Shopping, Amazon, and brand domain structure — and identifies where AI Overviews visibility is leaking.
Request an AI Overviews Audit