If you’re a marketer, you’ve spent the past 5, 10, or 20(!) years thinking about SEO.
You’ve spent long days and late nights: finding the right keywords, crafting content hubs around them, optimizing your pages, slowly building authority, earning some links, and hoping Google deems you worth of a spot on page one.
And after all that time in SEO-land, you’re now seeing the landscape turn into something else entirely.
Now, people are asking ChatGPT, Gemini, Claude, Perplexity, and whole host of other AI tools questions they used to type into Google. And increasingly, those tools are pulling information from across the web to formulate their answers, sometimes linking back to their sources and sometimes not.
That raises an interesting question for video marketers:
How do you create videos that AI systems can actually find, understand, and potentially cite?
Though it’s always evolving, much like LLMs, the answer is both similar and unique to what you’re used to with SEO.
The big idea is pretty simple:
Don’t just make your video engaging. Make it really easy to understand.
Here's how.
1. Start with the questions your audience is actually asking
LLMs are fundamentally trying to answer questions, and in more nuanced or conversational ways than traditional query-based searches. So if you want your company to show up when someone asks, "What's the best way to safeguard employee AI use", creating a two-minute video called "Our New Product Overview" probably isn't going to get you very far.
A video called "How to Best Safeguard Employee AI Use in Under Two Minutes" is a much more obvious match.
Think about the questions your customers ask your sales team, customer support team, or Google. If possible, fire up SurveyMonkey and ask your customers directly. Then build videos around those questions.
Some examples:
- How does [category] work?
- What's the difference between X and Y?
- How much does [solution] cost?
- What should I look for when choosing [product category]?
- How do I solve [specific problem]?
- Is [common approach] worth it?
- What are the biggest mistakes people make with [topic]?
This is actually pretty close to how Demo Duck has approached business video production for years: start with the audience's problem and figure out the creative treatment from there. Their own creative process begins with understanding the company, audience, goals, and common objections before getting into the script.
AI search just gives us another reason to do it.
2. Write the script for two audiences: humans and machines
Man vs. machine…a scripting tale as old as time, but more relevant now than ever.
LLMs sometimes use video in a more direct manner in formulating their answers, but today, a lot of their understanding of web-based video still comes from the text surrounding it: transcripts, titles, descriptions, metadata, and the webpage itself. Wistia's research makes the point pretty clearly: for the foreseeable future, if the information in your video isn't available as readable text, AI systems may have a hard time understanding it.
That means your script suddenly matters for two audiences:
The person watching the video.
And:
The robot trying to figure out what the video is about.
While it may work on your website (we've been known to maybe do this on a few pages...), please don’t start stuffing your script with keywords. The people watching won’t like it and there could be negative consequences for the robots scrubbing it too.
Instead, be as specific as possible without being overly repetitive in your language.
If you're talking about customer onboarding software, don't say:
"Our platform makes onboarding easier."
Say what "easier" actually means:
"Our customer onboarding software helps companies automate tasks like account setup, training, document collection, and progress tracking."
That's useful to a human being. And it's also much easier for an AI tool to understand and associate with relevant queries.
In other words, write the way your customers talk when they're looking for an answer. Just like in middle school English class, don’t be afraid to restate that question in your answer too.
3. Give every video a transcript
This is probably the single most practical thing you can do, and luckily it’s quite an easy task.
Create a transcript for every meaningful video you publish, and make that transcript available as actual HTML text on the page.
Actual, crawlable text.
Why? Because a transcript turns your video from a piece of media into a searchable body of information.
And it gives AI systems significantly more context to work with.
This is especially important if you're embedding a YouTube video on your website. Wistia notes that YouTube's iframe-based embeds don't necessarily expose the video's transcript to web crawlers on your page, making a transcript alongside the video particularly useful.
At Demo Duck, we use Wistia and Rev to generate our transcripts, pending the video project needs and publishing plan.
There's a nice bonus with this step, too: your transcript can become the foundation for the rest of your content.
One video can kick off a glorious content journey:
Video → transcript → blog post → social clips → newsletter → FAQ → sales enablement content
4. Make the video itself genuinely useful
This sounds obvious, but it's worth emphasizing.
Much like a website's blog section isn’t a keyword junkyard, AI search isn't an excuse to make boring keyword-targeted videos.
If anything, it's a reason to make better videos.
Think about the kind of content an AI assistant would want to recommend to someone. It's probably not a 90-second commercial that says, "We're innovative. We're different. We're passionate."
It's something that actually answers a question.
This is where explainer videos can be particularly effective.
Take our project for Crazy Egg. The video quickly establishes the problem, explains the product, and shows how the solution works. That's why the video helped increase Crazy Egg's conversion rate by 64%.
You don't have to carb copy that video, especially as it's over a decade old, but it's important to make something that clearly teaches the viewer something.
If your video genuinely explains a concept, answers a question, demonstrates a process, or helps someone make a decision, you've created something with value beyond your own marketing message.
And that's exactly the type of content you want associated with your brand.
5. Give every video a very literal title and description
Your creative team may want to call your video something clever. Your AEO strategy needs to fight against this urge.
If your video answers the question "How does employee onboarding software work?", there's nothing wrong with calling it:
How Employee Onboarding Software Works
You don't need to name it:
Welcome to the Future of Onboarding™
Save the clever copy for the OOH campaign. Your video title should clearly communicate the video's subject. Your description should take that a step further.
Explain:
- What the video covers
- Who it's for
- What questions it answers
- Important concepts mentioned
- Products, people, companies, or technologies discussed
- Related topics
Think of the description as the Spark Notes version of your longer video. Google's own guidance similarly recommends using structured video data to communicate information such as the video's description, thumbnail, upload date, and duration, and notes that structured data can help Google discover videos.
6. Break longer videos into chapters
If your video covers five different questions, don't treat it like one giant blob of content. Break it into chapters for the viewer, the metadata, and inside of YouTube itself (if applicable). Also, the more descriptive and clear you can make the chapter titles, the better.
For example:
00:00 — What is customer onboarding?
01:15 — How does automated onboarding work?
02:40 — What does onboarding software cost?
04:05 — What features should you look for?
05:30 — Common onboarding mistakes
Now you've given both humans and machines a much clearer map of what's inside. And it'll be easy to implement these directly in Youtube or Wistia.
This also makes it quite easy to create shorter videos from the same underlying content, chop it up by chapter. And importantly, each chapter can answer a slightly different search query around the same topic. That's a much more powerful content strategy than making one 30-minute video and hoping someone discovers it, and can actually find what they need out of it..
7. Put the video on a page with supporting content
We talked about content hubs importance as it pertains to SEO so let's carry through that same line of thinking here. Don't think of the video as a standalone asset. But instead, think of it as the centerpiece of a content page—it's own mini-hub.
So, if you've created a video called "How to Choose Project Management Software." Don't just embed it on /videos/project-management-video/ and call it a day. Create a useful page around it.
Include:
- A clear headline
- A short introduction
- The video
- A transcript
- Key takeaways
- Relevant statistics or examples
- FAQs
- Links to related resources
- A clear next step
Now you have a webpage that gives an AI system multiple signals about what the video is actually about. And you're not just optimizing for AI. This simplicity and clarity makes it a genuinely useful resource for humans, too. That's the balance we're trying to strike here. Hooray for us!
8. Use video structured data
Google supports VideoObject structured data, which can provide additional information about your video and help Google understand and discover it.
At minimum, make sure you and your team is thinking about these things when creating the script and putting the final touches on the render:
- Video title
- Description
- Thumbnail
- Upload date
- Duration
- The page where the video lives
- Video URL
Does JSON-LD and the idea of a VideoObject file make you a bit nervous? Don't worry, Demo Duck is able to create these files for our clients to make sure their videos are getting found.
If you're going at it solo, don't let your beautifully produced video disappear into a JavaScript-heavy page with almost no text explaining what it is.
9. Build a series, not a pile of random videos
At Demo Duck, we've been approaching educational video production series like TV shows, for years. Both internally and externally, we have referred to the various scrips and videos as "episodes" in order to make sure we're properly crafting the right viewer experience with the project. So, as you think about an LLM-optimized video, one of the smartest things you can do for AI search is create topical depth.
Instead of making:
"Company Product Video #17"
Make a series around a specific topic. For example:
The Complete Guide to Customer Onboarding
Episode 1: What is customer onboarding?
Episode 2: How to build an onboarding process
Episode 3: 7 common onboarding mistakes
Episode 4: Automated vs. manual onboarding
Episode 5: How to measure onboarding success
Episode 6: Customer onboarding software explained
Now you're not creating six unrelated videos. You're building a library of expertise around one subject.
We've long advocated for video series as a way to move beyond one-off videos and create multiple pieces of informative content around a shared goal. That strategy makes even more sense in an AI-search world. The more consistently you cover a topic—and the more consistently your company is associated with that topic—the more opportunities you create to be surfaced when people ask related questions.
The big takeaway: make your videos easy to understand
There's a temptation to think about "ranking in ChatGPT" as a new version of keyword stuffing. Crank out a bunch of blogs and videos, stuffed with the right phrases, prompts, and keywords and watch the KPIs soar.
But of course, much like with SEO, effective engaging content will find a way to win long term. So, the better question is:
"How can we make the expertise in our video as easy as possible for an AI system to understand, retrieve, and trust?"
That starts with making useful videos around real prospect or customer questions. Then you support those videos with transcripts, clear titles, detailed descriptions, chapters, structured data, and useful webpages.
And most importantly, you keep the content focused on helping someone—not gaming a machine. That's actually one of the nice things about optimizing video for LLMs. The tactics that make your content easier for AI to understand are, for the most part, the same tactics that make your content better for humans. But maybe will a slightly less pithy way about the script and promotional copy.
So don't throw out everything you've learned about video marketing.
Just add one more viewer to the room. The robot.