How to Measure YouTube Citations in AI Overviews

Most AI visibility tracking asks one question: did the answer link to my domain? That question quietly misses a large share of the citations a brand earns, because Google’s AI features often cite a brand’s YouTube video instead of its website. This piece walks through how to measure video citations properly, using four weeks of real data from the PushLeads agency as the worked case.

The blind spot in domain-only tracking

An AI Overview, an AI Mode answer or a Perplexity response comes with a list of source links. A typical tracker reads that list, extracts the domain of each link and checks whether the brand’s domain is among them. A link to a brand’s own video is reduced to “youtube.com” and is either ignored or, worse, counted as a third-party source.

PushLeads ran into exactly this. On November 8, 2026, its weekly check of buyer questions found 29 Google AI Overview answers citing pushleads.com. When the same answers were re-read with the brand’s own videos counted, the figure was 79.

A measurement method that credits video

The fix is not complicated, but each step matters:

1. Store full source URLs, not just domains. The video ID lives in the URL path or query string, so a domain list cannot recover it later.
2. Resolve redirect links. Some AI answer data returns Google redirect URLs for sources. Decode the destination before matching anything.
3. Keep a list of the brand’s own video IDs. Pull them from the channel, and include Shorts, which use a different URL pattern from standard watch links.
4. Match both ways. A citation belongs to the brand if its host is the brand’s domain or if its URL contains one of the brand’s video IDs.
5. Record the source position. Note where the brand’s first link sits in the source list. Position one is the first source shown — the closest thing AI answers have to a ranking.
6. Track presence separately from citation. When the AI feature does not appear at all for a query, that is not a lost citation; nobody took the slot.

What the PushLeads data showed

Applied to PushLeads across four Sunday measurements, October 18 to November 8, 2026, the method attributed 1,484 citations to the brand. 55% of them (823) pointed to YouTube videos. 160 of the 564 public videos on the PushLeads channel were cited at least once.

The engine split is the part practitioners should pay attention to:

Video citation is close to a Google-only phenomenon. A brand’s video strategy can lift its visibility in AI Overviews and AI Mode while doing nothing for assistants that build answers from written pages.

The effect on a composite score is large. PushLeads’ AI Visibility score for November 8 was 29.5 with video citations credited and 19.8 without them.

Embeds: a weaker signal than the video itself

The data also lets you ask whether embedding a video on a page helps that page get cited. In the PushLeads case, pages with a working embed were about 45% of the site but 52% of the cited pages, and they received 60% of the site’s citations. That is a small edge, and it is confounded: videos were embedded on the strongest topical pages first. A before-and-after comparison on the same pages is the right test, and PushLeads is running it.

Dedicated watch pages, which wrap a player with little supporting text, performed worst of all: 16 of 329 were cited, 25 times in total.

Practical takeaways

• Audit your tracker. If it matches on domain only, your Google AI numbers are likely understated.
• Report video citations as their own line, broken out by engine.
• Treat Google AI features and text-first assistants as two different channels with two different levers: video for the first, written answers for the second.
• Measure source position week over week, not just whether you were cited.

For the full write-up with PushLeads’ own commentary and video, see Are YouTube Videos Driving Your AI Visibility? on pushleads.com.