I’ve been in a few rooms this week where I heard an argument that made me cringe: 

“Videos don’t really matter anymore because AI search doesn’t use them.”

Respectfully. What the heck. No. 

Or, at least, not exactly. AI search can absolutely use video.

The bigger problem is that a lot of videos are not packaged in a way that makes them easy to understand, retrieve, or trust.

The video exists. The expertise is there.

The transcript is either missing, messy, or buried. The title sounds like an inside joke (or a local paper headline from 1997). The description says something deeply useful like “Great conversation with Sarah.” The chapters don’t exist. The page around the video gives almost no context.

And then everyone wonders why AI didn’t magically pull the right answer from a 38-minute recording. It’s the same reason your videos aren’t showing up in organic search. This isn’t new. 

Video is not invisible to AI search. Unstructured video is just much harder to work with.

Google is very clear about this: generative AI search experiences can bring in relevant images and video.

Google’s own guidance also says that if you’re already following image SEO and video SEO best practices, you’re already optimizing for generative AI search.

So the idea that video “doesn’t count” in AI search is already too simplistic. But, wait! There’s a catch. Isn’t there always?

AI search is not rewarding the mere existence of a video.

It still needs to understand what the video is about, what questions it answers, whether the source is credible, and where in the video the useful information lives.

That’s where a lot of B2B video falls short.

Teams spend weeks producing the episode, interview, webinar, demo, or customer story.

Then the searchable layer gets treated like admin work.

Upload it.
Add a title.
Maybe paste in a description.
Auto-generate captions.
Move on.

That might be enough to publish the video. It’s not always enough to make the video useful to search.

If video were irrelevant to AI search, YouTube wouldn’t be showing up the way it is.

Citation patterns across ChatGPT, Perplexity, and Google’s AI products and found that YouTube averaged a 20% citation share across AI platforms. Plus, 29.5% of Google AI Overviews cite YouTube, making it the top cited domain in that analysis.

That’s a pretty loud signal. AI search engines are not ignoring video.

But they’re also not treating every video equally.

Google AI Overviews and ChatGPT use YouTube differently (so do other conversational engines). Google appears to use YouTube more broadly as an authority source, while ChatGPT is more selective and tends to cite YouTube for specific use cases like instructional, how-to, and entertainment-related queries.

That distinction matters. Because the job is not simply “put the video on YouTube.”

The job is to make the video clearly answer something your audience is likely to ask.

A vague episode title won’t help much. A buried insight 24 minutes into a conversation won’t help much. A raw transcript with no structure might technically contain the answer, but good luck asking a machine or a human to do the archaeological dig.

AI search needs handles. The more legible the video is, the more useful it becomes.

This is where video SEO and AI-search strategy start to overlap. A video is not just a file. It’s a content asset with layers. There’s the actual video.

Then there’s everything that tells search systems what’s inside it:

The title.
The description.
The transcript.
The chapters.
The thumbnail.
The embedded page.
The surrounding copy.
The schema.
The internal links.
The source it’s published on.
The way other content references it.

Google’s video SEO documentation points to the same principle: videos need to be discoverable, crawlable, and indexable. Google also supports video structured data, including VideoObject and Clip markup, and can use timestamps to surface key moments in a video.

That’s not just good for Google. It’s good for humans, too.

A buyer doesn’t want to scrub through a 42-minute interview hoping someone eventually answers their question. (and they won’t)

An AI system does not want to guess whether your video contains the answer based on a cute title and a one-sentence description.

Both need structure. So when someone says, “Video doesn’t work for AI search,” the better response is:

What kind of video?
Published where?
With what transcript?
With what title?
With what chapters?
On what page?
Answering what query?
Connected to what authority?

Because a video that’s packaged as a searchable resource is very different from a video that’s basically a black box with a play button.

Use this before publishing a video you want to support search, AI discovery, or buyer education.

Before you title the video, ask what the buyer would actually search or ask. Yep, the real human you’re talking to. 

Absolutely not: “Episode 47 with Sarah Jones”

More like:
“How B2B teams can use customer videos to support sales”
“What changes when your podcast becomes a video-led content engine?”

The title should make the topic obvious. Clever is nice. Clear gets retrieved.

A good description should explain what the video covers.

Include:

The main argument.
The questions answered.
The topics discussed.
The audience it’s useful for.
The outcome someone should understand after watching.

“Great conversation with Sarah” is not a description. It is a polite shrug.

Captions are a start. A useful transcript is better.

Make sure the transcript is accurate enough to support the topic, names, terms, and key ideas.

If your transcript turns “AI search” into “I search” 19 times, the machine is not the only one having a bad day.

Chapters turn one long video into a structured resource. They tell people and search systems where the important moments live.

Use them for:

Main questions.
Topic shifts.
Frameworks.
Demos.
Examples.
Customer stories.
Key takeaways.

If one section answers a high-intent question, make that section easy to find.

Here’s a fun exercise for you: Head on over to Google or Chat or your favorite search engine. Ask it a ‘how to’ question. If there’s a good video on the topic, I betchya it comes up in response. More: it comes up linked to a specific chapter. 

Do not make the embed do all the work.

Support the video with:

A short intro.
Key takeaways.
A written summary.
Related resources.
FAQs.
Internal links.
A clear next step.

This gives search systems more text to understand and gives buyers more ways to engage.

If the video lives on your website, structured data can help clarify the basics:

Name.
Description.
Thumbnail.
Upload date.
Duration.
Embed URL.
Key moments or clips.

Schema is not magic. It’s labeling. And AI search likes assets that are easier to label correctly.

Not every video has the same search job.

A demo might answer product-specific questions.
A webinar might support education.
A podcast clip might introduce a POV.
A customer story might validate a decision.
An executive video might explain how your company sees the category.

The structure should match the job.

YouTube matters.
Your website matters.
Transcript pages matter.
Blog posts matter.
Case study pages matter.
Owned and third-party authority both matter.
(We get it, Paige. It all matters. Move on.)

The goal is not to make one lonely video carry the whole search strategy.

The goal is to create a connected set of assets that make your expertise easier to find, understand, cite, and trust.

AI search doesn’t make video less useful. It raises the bar for making video understandable.