How Youtube Affects AI Overviews for SaaS: A 259-view YouTube video can win the AI Overview citation panel.

When we pulled the AI Overview citations for 100 SaaS shopping queries last week, one finding stuck out: YouTube was the second-most-cited content type after vendor-authored listicles. About 10% of all top-3 citations were videos. So we went back and pulled the metadata on every YouTube video that showed up anywhere in the AIO source panels — 176 distinct videos, 119 distinct channels — and asked the obvious follow-up question: what makes a video cite-worthy?

The answer is more interesting than “big channels win,” because big channels do not consistently win.

The dataset

FULL DATA SET HERE

For each of the 100 SaaS shopping queries — things like “best CRM for small business,” “best applicant tracking system,” “best fraud prevention software” — we extracted every YouTube reference Google’s AI Overview returned (1–15 references per AIO). For each video we then queried DataForSEO’s YouTube Data API for its full metadata: view count, like and comment counts, channel name, channel subscriber count, publish date, duration, category. The end result was 176 distinct YouTube videos with full metadata, 28 of which appeared in the top-3 citation slot of at least one AI Overview.

That 28-video subset is the one that matters. The top-3 citations are the videos getting the click — most user attention collapses after the first few entries in the source panel.

Top-3 cited videos do skew bigger — but the floor is shockingly low

Comparing top-3 cited videos to videos that only made it to rank 4+:

Metric	Top-3 cited (n=28)	Rank 4+ (n=147)	Lift
Median views	6,892	1,335	5.2×
Median subscribers	65,400	9,880	6.6×
25th percentile views	2,087	434	—
Smallest cited video	259 views	—	—

Both differences are statistically significant (Mann-Whitney p < 0.005 for each). So yes, on average, the videos Google promotes to the top of the AIO citation panel are bigger than the long-tail videos that just barely got cited.

But the floor is the part marketers should look at twice. The smallest video in our top-3 set has 259 views. Several top-3 videos are under 1,000 views. The median is just under 7,000 views. None of these numbers describe a YouTube partnership channel — they describe what for most B2B marketers would be considered noise-level content.

This is not Google rewarding small channels for the sake of it. It is Google reading the content match between the query and the video — and content match dominates view count once a video clears a minimum threshold.

Views matter slightly more than channel size — but they overlap

The natural follow-up question: do top-3 cited videos win because the channel is big, or because the individual video is well-watched? A multivariate logistic regression with both predictors (log-views and log-subs, standardized) tells us:

Predictor	Standardized coefficient	Odds ratio per +1 SD
Views per video	+0.62	1.85×
Channel subscribers	+0.42	1.53×

Once you control for the fact that big channels publish videos that get more views, views per video is the slightly cleaner predictor of top-3 placement. Channel popularity matters mostly because it correlates with video performance — not because Google is brand-leveling the channel.

A useful way to internalize this: a 200-subscriber channel publishing a 50,000-view “I tested every CRM” video is a better AIO citation candidate than a 200,000-subscriber generalist channel whose category-specific video gets 1,000 views. The video does the work, not the channel logo.

(Important caveat: with n=175, this gap is real but modest. Don’t over-index. The honest framing is “both views and subs are positive predictors of similar magnitude, and they overlap heavily.” Views just edges ahead in the multivariate model.)

The Stewart Gauld pattern

When you sort the channel-level data by top-3 placements, the leaderboard is short and revealing. Most channels get one cited video. A handful get two. One channel breaks that pattern: Stewart Gauld — 4 videos cited, all 4 in top 3, across 4 different SaaS categories. He’s not the largest channel in the dataset. His videos are not the most viewed. What he has is a format-and-niche fit: he publishes meticulously structured “I tested X, Y, and Z” comparison videos for software categories that small businesses search for.

A few other channels are pulling the same play in narrower lanes:

• Hugo Pod — 2 top-3 placements, both for “best AI voice agent platform.” Channel has 4,560 subscribers. Google cited the same channel twice in a single AIO panel because no other creator is producing the comparison content that query needs.
• KodeKloud — 2 top-3 placements in the DevOps category.
• AssemblyAI — vendor channel, 1 top-3 placement.

The pattern is consistent: the AIO citation panel rewards category-specific comparison reviewers, not generalist tech channels and not vendor-owned content. There are 119 distinct channels in the dataset, which means the creator landscape for SaaS shopping is heavily fragmented — almost no one owns more than 1–2 categories.

YouTube Shorts are AIO-citable

15 of the 176 cited videos are YouTube Shorts. That’s 8.5% of the dataset, which is small but not zero. Google’s AI Overview is willing to pull from short-form video when the content matches — examples in our dataset include Shorts cited in the top-3 for “best CI/CD tools” and “best AI coding assistants.”

This is news because most B2B content strategies treat Shorts as a top-of-funnel awareness format, not as a SERP asset. The data suggests a Short with high content density and clear product comparison can do AIO duty just like a long video.

The dominant format is “I tested every X”

Across the top-3 video set, the title pattern is overwhelming:

• “I tested every Voice AI platform. Here is the BEST.”
• “I Tested 5 Accounting Software So You Don’t Have To”
• “Top 5 Best Cold Email Tools (2026)”
• “Best Payroll Software for Small Business (Watch BEFORE Buying)”
• “The 4 Best POS Systems of 2026”
• “I Tested Every Cold Email Outreach Tool So You Don’t Have To”

This is the same format pattern that wins on the text side of the AIO panel: ranked, comparative, opinionated, recent. Google’s model appears to weight the same content shape regardless of medium. If your category does not currently have a creator publishing “I tested every X” content, that is a gap, not a saturation problem.

What this means for B2B SaaS GTM

A few concrete shifts the data suggests:

Stop optimizing for the biggest YouTuber in your category. The data argues for partnering with the most format-aligned creator — the one who already publishes comparison reviews for software in your space, regardless of channel size. Stewart Gauld at 4K subs is the AIO citation play. The 500K-sub generalist who occasionally covers your category is not.

Treat Shorts as SERP assets, not just awareness assets. AIO will pull from them when content density is high. A 60-second product comparison Short can earn a top-3 citation in markets where no long-form comparison content exists.

The category is open. 119 distinct channels across 100 queries means there is no incumbent creator owning B2B SaaS reviews. The first SaaS company to systematically build creator partnerships across the long tail of comparison-format reviewers — fund the videos, sponsor the format, ensure the comparisons are fair — is going to harvest disproportionate AIO citations across multiple categories.

Track creator-channel fit, not just creator size. A creator who consistently publishes ranked comparison content is more valuable than one with 10× the audience but less category alignment. The metric to optimize for is “videos in this exact format about this exact category that get watched through” — not raw subscriber count.

Vendor-owned YouTube content is mostly absent from top-3 citations. AssemblyAI is the rare example of a vendor channel breaking through. The pattern in this dataset is that AIO trusts third-party product comparisons more than vendor demos — which means your owned YouTube channel is unlikely to be your fastest path to AIO citation. Earned creator coverage is.

Methodology and caveats

We pulled YouTube video metadata for every video referenced in the AIO citation panels of 100 SaaS shopping queries, via DataForSEO’s YouTube Data API (/v3/serp/youtube/video_info/live/advanced). Data collected May 2026, US, English. Each video was queried once. Total YouTube API spend: about $1.10 across 176 videos.

Important limitations:

• Top-3 vs rank 4+ is a within-AIO comparison, but our analysis treats top-3 placement as a global property of a video. We do not have a counterfactual showing which videos Google considered but did not cite.
• View counts are point-in-time. A video cited in May 2026 may have had a very different view count when Google’s index built the AIO. We’re seeing the current state, not the state at citation time.
• Sample size is modest. 28 top-3 videos across 119 channels means channel-level patterns (Stewart Gauld, Hugo Pod) are real but not statistically robust against deeper category-by-category breakdowns.
• US/desktop only. Other markets and mobile AIO behavior may look different.

The full dataset — every video, every channel, every keyword attribution, every view count — is available alongside this article. If you replicate this for a different vertical or a different geography, we’d like to compare notes.

---

Data and analysis by AUQ. Pulled May 2026 via DataForSEO YouTube Data API.