Does YouTube Video Actually Show Up in AI Search? A Live Test With Jess Hennessey

Dane Frederiksen and Jess Hennessey are running a live experiment to figure out whether YouTube videos actually show up in AI search yet. This is the mid-test check-in. They open a video page, paste a question into Google, and look at what AI Overviews, ChatGPT, and Perplexity do with it. The results are messier than expected and more useful than expected — both at the same time.

Key Takeaways

• YouTube videos do show up in Google search results, but as the top Video result — not consumed inside the AI Overview answer itself.
• Most buyers will stick with Google search, not jump directly to Gemini or ChatGPT. That means Google is still where the GEO test matters most.
• The lift from a video for AI search comes from the context around the video — transcript, description, FAQ block, chapters — not the video file alone.
• Long-form video pages get richer AI treatment than YouTube Shorts. Shorts don't get the same description and transcript surface.
• LLMs are lazy. Chapter-formatted transcripts beat wall-of-text transcripts because they require less reasoning to summarize.

Why this test exists

Jess Hennessey has spent almost 25 years in web design, development, and digital marketing. For the last year she's been all-in on GEO and AEO. Her take on why this matters:

"I think it's really important to understand how video is going to be leveraged by AI answers. With the recent announcement from Google that it was moving more towards an AI-first search, I've been playing a lot with Google AI Overviews and AI Mode. One of the things I noticed is that it pulls in YouTube links into the responses." — Jess Hennessey

The mechanism is straightforward. LLMs synthesize answers from the content they were trained on, and they often answer without crediting any specific brand. A blog post can get summarized into an AI Overview that never names you. A video can't be summarized the same way. If the AI cites a video, it links to your YouTube page — your brand, your channel, your face.

That's the bet behind the test. Put videos in places where AI engines can find them and see whether the citation lift is real.

Why B2B can't keep treating video as a nice-to-have

Dane's framing on why this is the moment:

"Video feels more like a need-to-have rather than a nice-to-have. B2B companies are losing so much control. They're losing so much visibility. They're drowning in a sea of slop. This is something they can grab onto that has real meaning." — Dane Frederiksen

The "sea of slop" point is the punchline of the last 18 months of AI content. Everyone shipped more written content. Everyone's written content started to sound the same. The differentiation problem got worse, not better. Video is harder to fake, harder to commoditize, and easier for an AI engine to cite back to its source.

What we're actually testing

Dane and Jess set up a test document a few weeks ago. They picked specific GEO questions, recorded video answers, posted them to YouTube, and embedded them on the website. The plan was a two-to-eight-week window before checking citations.

The mid-test sneak peek: Dane searched the seed questions on Gemini, ChatGPT, and Perplexity. No citations yet. Not surprising — eight weeks is still the upper bound, and the videos are only days old in some cases.

But that's the wrong place to look first.

The finding nobody expected

When they searched the same question directly in Google (not Gemini, not ChatGPT, just Google), the AI Overview rendered at the top of the page, and below the sponsored results, the video showed up. Top of the Videos carousel. Their video.

The AI Overview didn't cite the video. The video was a separate placement below the AI block. But the placement was the top spot.

"Going back to Google's announcement about AI-first search, people are going to go to Google directly. They're not going to go to AI mode. At least my hunch is, they're not going to go directly to Gemini. They're going to stick to Google. That's what people are used to." — Jess Hennessey

This reframes the test. If most buyers stay on plain Google and the AI Overview surface is the new top-of-page, then the question isn't only "are AI engines citing my video?" The question is also "is my video the top result that the user sees right under the AI summary?"

Their video was. That's a useful finding even before any LLM citation lands.

Why the context around the video matters more than the video

Jess's strongest point in the call. She was looking at one of Dane's blog pages with a video embedded and a transcript pasted underneath, and she said:

"How you have framed the context around those videos is really where the lift is going to come from. It isn't coming from the video directly. It's coming from the context around that video." — Jess Hennessey

This is what GEO actually is, in practice. The video is the asset. The page around it — transcript, description, FAQ block, schema markup, chapter markers — is what AI engines parse to decide what the video is about and whether it answers the question being asked.

A bare YouTube embed on a blank page won't get the same treatment as a full transcript with chapters and FAQs. The video is identical in both cases. The page is the difference.

Long-form vs. shorts

Side-by-side check on the test pages: the long-form interview pages had a full transcript pasted in. The Shorts pages had only the short-clip transcript. That gap matters more than it looks.

"The shorts don't get the same treatment as the long-form. That's the difference." — Dane Frederiksen

A short by itself is too thin for an AI engine to learn anything useful about your brand or expertise. The fix is to link every short back to its long-form parent — and to publish the long-form with the heavy GEO context (full transcript, FAQ, JSON-LD schema). The short becomes a discovery surface; the long-form is what the LLM actually learns from.

The chapter-formatted transcript insight

Maybe the most actionable finding of the whole call. Dane and Jess pulled up a competitor's video page and compared transcripts. The competitor's transcript wasn't a wall of text. It was structured with headers — Chapter 1, Chapter 2, Chapter 3 — and the transcript text fell under each chapter.

"From an LLM standpoint, that's much easier to follow. Because with yours, they're going to have to read it. And my favorite saying is: LLMs are lazy. They don't really want to think. So a wall-of-text transcript is causing them too much thinking." — Jess Hennessey

This is a small structural change that compounds. Add chapter headers. Group transcript text under each header. The LLM doesn't have to summarize the whole page to figure out what each section is about — the structure tells it.

Frequently Asked Questions

Do YouTube videos show up in Google AI Overviews yet?

In this test, the AI Overview itself didn't cite the test videos. But the video did appear as the top result in the Videos carousel directly below the AI Overview, above the sponsored links. That's still a high-value placement. The AI citation may come later — the test is only mid-window — but the Google Videos placement is already live.

Where should you actually run your GEO video test — Google, Gemini, ChatGPT, or Perplexity?

Start with plain Google search. Jess Hennessey's hunch is that most buyers will stay with Google and use AI Mode or Gemini sparingly. Google's own AI Overview surface is the new top of the page for most users, so that's where the citation has the most reach. Run secondary tests on Gemini, ChatGPT, and Perplexity to spot LLM-specific patterns, but Google is the volume play.

Does the video file itself drive AI search visibility, or is it the page around it?

The page around it. The video file gives you the right asset, but AI engines parse the transcript, description, FAQ block, schema markup, and chapter markers to decide what the video is about. A YouTube video embedded on a bare page won't get treated the same as the same video on a page with a full transcript, FAQs, and JSON-LD schema. The context is the lift.

Why don't YouTube Shorts work as well as long-form for AI search?

YouTube Shorts pages don't get the same description and transcript surface that long-form pages get. The clip itself is too thin for an AI engine to learn meaningful context. The pattern that works: use Shorts as discovery, and link every Short back to a long-form parent page that has the full transcript, full description, FAQs, and chapter markers. The long-form is what trains the LLM. The Short is what surfaces it.

How should you format a video transcript so AI engines actually use it?

Don't paste a wall of text. Break the transcript into chapter sections with H2 or H3 headers. Group the relevant transcript text under each chapter. This matches the chapter structure on the YouTube video itself and gives the LLM clear structural cues about what each segment is about. As Jess Hennessey put it: LLMs are lazy. They don't want to do extra reasoning work. Format the page so they don't have to.

What's next

Dane and Jess are extending the test window and adding a transcript-formatting comparison — taking the same video page and republishing two versions, one with a wall-of-text transcript and one with chapter-formatted transcript and FAQs. The goal is to isolate whether the page structure alone moves AI citations.

"We're definitely in the middle of this. It's messy. This is us building in public, but that's a good thing." — Dane Frederiksen

The next check-in lands in a couple of weeks.

Full Interview Transcript

Dane Frederiksen: Hey everybody, my name is Dane Frederiksen and I'm a video expert in the B2B space. I'm joined by Jessica Hennessey, an expert in the SEO, GEO, AEO world. Jess, give us a little background about you before we get into it.

Jess Hennessey: Sure. Thanks so much for having me again. I'm Jess Hennessey, as you said. I've been in the web design, development, and digital marketing space for almost 25 years — both from a development and design side, and from the digital marketing side. And really for the last year, I've been all in on GEO, AEO, trying to understand what it means.

Dane: Yeah, you and me both. It's SEO's newer baby brother or baby sister.

Jess: Exactly.

Dane: Got to get to know the space. Just a little context for this video: Jess and I have recently started to test the GEO best practices when it comes to video. How can video help move the needle and help you show up in AI search? So we've come up with a test.

Jess, what do you think is really important to point out about this test and why we're doing it?

Jess: I think it's really important to understand how video is going to be leveraged by AI answers. And I think we have a really great opportunity right now. I might be diverging from your question, but with the recent announcement from Google that it was moving more towards an AI-first search, I've been playing a lot with Google AI Overviews and AI Mode. One of the things I noticed is that it pulls in YouTube links into the responses.

And I think that's really critical when you think about the way LLMs are synthesizing answers and not necessarily giving your brand credit for the answer it's giving. So with video, you have control of that narrative by combining your brand with the concept being conveyed. And it can't take that away from you. That's been a really big issue from an LLM response standpoint — it's kind of taking the copyright. It's stealing your information and coming up with something new.

Dane: It steals your candy.

Jess: Right. It's stealing your information and coming up with something new and not giving you credit. So I think if you'd asked me this question a month ago, in our first conversation, I probably wouldn't have been so bullish on video. But with the recent developments on the Google side, there's a real opportunity.

Dane: Yeah, I think for me — I'm a video expert. I've been doing this for a long time. I inherently see the value of what video can do for a company and for your personal brand. But now more than ever, we've gotten to a point where video feels more like a need-to-have rather than a nice-to-have, for the reasons you just pointed out.

This is a really important conversation to be having for B2B companies. Because they're losing so much control. They're losing so much visibility. They're drowning in a sea of slop. This is something they can grab onto that has real meaning. And there's a much clearer path to follow in that direction. How can we do more with video? Seems like an obvious question we should be asking.

So anyway, that's my soapbox. Just to go back to how we started here — the whole idea was for you and I to put together a test to see how we can move the needle with video. What are the right ways to do this, and how soon can we see some results? So I'm going to share the document we created together.

The idea was to take some videos and put them in places where we can see if they're showing up in search results. At the top of the document, I've pasted a lot of stuff in here, Jess.

Jess: That's quite all right.

Dane: So this is your outline of how we were going to test this. Just to summarize, we're going to put some long-form content out there. We're going to put some short-form content out there. We're going to do different variations about where this content lives. It's all going to be on YouTube, but some will be embedded in the website. And we can compare and contrast how those things are showing up.

As we've been going along, I've been doing a dump of information in this long document — what, 22 pages now. But at the very bottom, I've left us some next steps. This is supposed to be on the 10th of June. We were thinking maybe two to eight weeks we'd see some results.

I did a little sneak peek the other day and tested on Gemini, ChatGPT, and Perplexity — some of those first questions we asked. I didn't see any results yet. So these are the file names for the videos we posted, which are the answers to the questions we came up with. When I did those searches, I did not see any results coming up in Perplexity, ChatGPT, or Gemini. So I don't know if that's a surprise to you. Are you surprised by that?

Jess: No, I'm not really surprised yet. It hasn't been all that long.

I'd love to see — and I think it'd be great to show everybody — how you have put them on YouTube and embedded them on your website. Is the transcript included? Is the description included? How you have framed the context around those videos. Because that is really where the lift is going to come from. It isn't coming from the video directly. It's coming from the context around that video.

Dane: Right. I'm pulling that up as you're talking. Where did I put that stuff? It's been a minute. I don't know if you do this, but I've been working all weekend on Claude Code. I've got this AI overwhelm thing happening now — just in AI madness mode. We spend so much time on it.

Okay, this is the blog post. I think this is one of them here. We've got this embedded with YouTube. It's answering the question, and it has the full transcript. This one doesn't look like it necessarily has the full transcript.

This is on YouTube. This is one of the videos. I ran an agent to optimize these pages. Looks like it's got pretty much everything we would want in here. Anything jumping out as not right?

Jess: Well, I'm not seeing the transcript yet.

Dane: Yeah, I'm not either.

Jess: That could be part of the challenge.

Dane: Here's the transcript on YouTube. This is auto-generated. So maybe the transcript didn't make its way into the page.

Jess: Are you using a YouTube plugin?

Dane: YouTube automatically transcribes when you post it.

Jess: Sorry, I meant on your website when you put it on the blog.

Dane: I built an agent that takes the transcript I've downloaded from Riverside and uses that as the basis to make the blog post. I intended to have the transcript in these posts, but I'm not seeing it. Although no, this one has the full transcript. This is the full interview, the long 15-minute version.

Jess: But that's on YouTube, correct?

Dane: It's on YouTube and in this blog post.

Jess: I'm still seeing the YouTube. Can you show me the blog post?

Dane: Yeah, let me close this. I'm going to try this another way. See, this is us having an organic real conversation where mistakes get made.

Jess: We are not AI.

Dane: Exactly. So this is the full video in the blog post. And if I scroll down, we can see that it does have the full transcript.

Jess: Oh, so that is the transcript. There it is.

Dane: Yeah, full. So this is all in there. This is the full transcript for the long-form one. Now we need to see if it's in the shorts as well.

So the shorts will be... that's the full one again. Let's look at this next one. This is the short. I don't think this one has the transcript. Oh, it does. It says it has just the transcript from the clip. And it's linked to the full-length one as well.

Jess: So you have the quick answers, which I think are really good from an LLM standpoint. Can you scroll down?

Dane: Got direct quotes.

Jess: Yeah, so you get the FAQs. But is there anything? If our test is looking for your brand to appear, it may not. Kind of what we were just talking about — if it's pulling up your answers into the responses but not tying it to your brand, you wouldn't know that. So it would really be: are they linking to the video directly?

Dane: If we wanted to test this further, we'd put that question into Perplexity now and see how that comes up. So these are the questions we came up with. "Does YouTube content show up in AI answers?" So we could go to —

Jess: Which video is tied to that question?

Dane: This is number one. So the file name would be toward the end, where we had that laid out. Highlighted, I think. Here we go. Right here. Video number one. This is the file name of the video file, the MP4.

Jess: And where's the YouTube one that relates to that? Where's the YouTube post and the page post?

Dane: Let's go right to YouTube, see how we can do that.

So this should be shorts, and it should be with Sherri. Should be one of these. Why am I not seeing that? That's strange.

Jess: Put in one of the questions of the video. Let's choose that one.

Dane: We'll search for this name then.

Jess: No, the question. Go back to YouTube. Click on one of Sherri's videos, like the first one: "How can B2B software videos..." Copy and paste that question.

Dane: Into ChatGPT.

Jess: Or Google. Yeah.

Dane: Drum roll. I'm not seeing any links. We should ask it what its sources are.

Jess: First, actually, go back to the main Google search. Yeah. Put it there.

Dane: Right here.

Jess: Nice. Okay, so now go into "show more."

Dane: Maybe over here.

Jess: Go over to the right where it says "show all." Yeah.

Dane: Not showing up, but it is under "Videos" down here. Interesting.

Jess: That is good. I think this user behavior is what most people are going to do. Going back to Google's announcement about AI-first search, people are going to go to Google directly. They're not going to — at least my hunch is — they're not going to go to AI mode, like directly to Gemini. They're going to stick to Google. That's what people are used to. So that's where we want to do our initial searches when comparing it to the others.

Interesting that the video was right at the top of "Videos," while not being consumed within the AI answer, the AI Overview. I think that's still good. Because I'm just trying to think from a user standpoint — they're going to get to that AI Overview. They may or may not want to expand. So that video will be right there. So I think that's still an important finding.

If you could go back to the AI Overview, let's look at the links it's referring to. What are those compared to the video that you've put together? How are those different?

Dane: Should I click on this one here? This is the top result. So, blog posts. There is some video in here. Oh yeah, there's a bunch of videos in here.

Jess: Interesting. So they have a bunch of videos, but they don't have transcripts in it. No, sorry, go back. The one that's in pink. "The strategies." Click on that one.

Dane: This thing here. Yep.

Jess: So where is that? Is that on YouTube? Is that on their site?

Dane: Seems to be on YouTube. Why is it in this player by itself? Interesting. It's credited to YouTube right here. There are no other signals here.

Jess: Okay, so let's go into YouTube. Let's look at it on YouTube. All right, let's look at what they have for description, transcript, et cetera. They don't have... they do have a transcript.

Dane: This is going to be with all videos. This is auto-generated.

Jess: Right. So if we could compare this video to your video, how is it different? Do a quick side-by-side.

Dane: The one with Sherri was just long-form?

Jess: No, it was the short one. We were looking at the shorts.

Dane: Yeah, here we go. So the shorts don't get the same treatment as the long-form. That's the difference.

Jess: Okay. So you're right. Let's go to the longer one.

Dane: Yeah, if we go to Sherri's long-form, and we look at sort of a compare-and-contrast on the data. It looks pretty comparable. They've got some website stuff up at the top. Mine's down at the bottom. But I've got more other links here that they don't have.

Jess: How are their chapters laid out?

Dane: Same way. Seems like all of theirs have a link. Looks like some of mine didn't get generated.

Jess: It would be interesting to compare the transcripts and see exactly what they say versus what Sherri said. About that particular question.

So their transcript looks different.

Dane: It does. Maybe theirs isn't auto-generated. It looks a little more formatted than mine. I have no idea why that would be.

Jess: I think that's something to investigate.

Dane: Exactly. I'll take a screenshot of that.

Jess: Yeah, because it has like "Chapter 1." And then it gives it, and then "Chapter 2." From an LLM standpoint, that is much easier to follow. Because with yours, they're going to have to read it. And my favorite saying: LLMs are lazy. They don't really want to think. That would say is causing them too much thinking.

Dane: Yeah, that's a great quote.

Okay. So we've got some learnings from today about things being a little bit different. Some investigation to do. But we can say that our test is underway, and we've already learned that video did show up in results. It was not part of the AI summary, but it was the top video that was cited. So that is interesting. Anything else?

Jess: I think, if you could recall — were the videos above the sponsored posts or sponsored links?

Dane: Let me see if I can find that.

Jess: I want to say it was right below the AI Overview, wasn't it?

Dane: Below the sponsored ones.

Jess: Below the sponsored ones. Okay.

Dane: But it's the top video.

Dane: All right. Well, we've got some learnings.

Jess: We do.

Dane: We did our check-in. So we're definitely in the middle of this. It's messy. We're building in public, but that's a good thing. I think we're going to check back in in a couple of weeks and see what we found out. I'll commit to doing some research on these transcripts and see if we can learn the different approaches there.

Jess: All right. Sounds great. Thank you, Dane.

Dane: All right, thanks, Jessica. All right. Bye.