Search isn't just Google anymore
Your buyers have stopped scrolling ten blue links. They ask ChatGPT, Perplexity, Google AI Mode, and Claude a question and take the answer. That answer names a few brands and moves on. If you are not one of the names, you do not exist in that conversation, and you never find out you lost.
This guide is about how to become one of the names. Specifically, it is about why video is the most underused way to get there, and the practical steps to make your video readable, citable, and recommended by the engines your buyers now trust.
We built it from interviews with the people doing this work: Christopher Penn of Trust Insights, Andy Crestodina of Orbit Media, Jason Barnard of Kalicube, Kaleigh Moore, Jess Hennessey, Adrian Dahlin, Cassie Clark, and Carolyn Outhwaite. Their words are throughout.
Can AI engines actually read your video?
Not the way you think. An AI engine does not watch your video. It reads the language around it: the transcript, the page, the title, the chapters, the context you wrap it in.
"It is the video transcript that AI engines reference and cite, not the video file itself."
That single fact changes the whole game. A beautiful video with no transcript is a black box to a model. A plain video with a clean, chapter-formatted transcript is a feast.
"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, SEO/GEO/AEO expert
So the question is not "can AI watch my video." It is "have I given AI something to read."
Why this matters right now
The surface where buyers get answers is exploding, and the number of brands that get named is shrinking.
FACT: Google's AI Mode has surpassed 1 billion monthly users, with AI Mode queries more than doubling every quarter since launch. (Google, "100 things we announced at I/O 2026")
FACT: Google AI Overviews now appear on 71.7% of niche B2B SaaS searches (up from 63.3% the prior month). (Overthink Group, State of Niche B2B SaaS Search, May 2026)
At the same time, the answers cite fewer sources.
FACT: The average number of links in an AI Overview dropped 69.7% in two months, from 7.8 links down to roughly two. (Overthink Group, May 2026)
More queries, fewer slots. The brands that lock in those slots now are building a lead that compounds.
"There's a lot of white space here that's still left to be claimed when it comes to AI visibility and citation rate. So video is a really smart play. Not a quick win, but a smart long-term investment." — Kaleigh Moore, B2B SaaS content strategist
And the window is not open forever.
"Take control of it today, because in two years time it's going to be too late. Your reputation in the minds of these machines will be set in stone." — Jason Barnard, founder of Kalicube
How AI search decides what to cite
AI search is not a ranking. It is a recommendation. It does not hand the buyer a list to evaluate. It hands them an answer with a few names in it. To be one of those names, you have to look like evidence, not like marketing.
Models are trained to trust independent, community-validated sources over brand-owned copy. A product page reads like a pitch. A third-party platform reads like proof.
FACT: 85% of AI citations now come from third-party platforms rather than brand-owned properties. (AirOps, via Source Signal Stack)
FACT: 88% of Google AI Mode citations do not appear in the organic top-10 search results. (Moz, via Source Signal Stack)
Read that second stat twice. The thing you ranked for in classic SEO is largely not the thing AI cites. It is a different surface with different rules.
"It's essentially community-driven. It's third party, it's not brand-owned assets, this is community validation, which is what LLMs trust the most." — Kaleigh Moore
This is exactly why video on YouTube is so powerful. YouTube is a third-party platform that the models train on heavily, and Google surfaces it because Google owns it. Video lets you put your own voice and language into AI's training set, on a platform that reads as community rather than as a brochure.
"AI engines train on the language inside YouTube videos. If that's what your company does, and you didn't make that video, you're not in the game." — Christopher Penn, Trust Insights
One more thing the engines do that breaks old assumptions: they personalize. Google tailors AI Overviews and AI Mode using the searcher's own Gmail, calendar, photos, and YouTube history. The same question returns different brand recommendations for different people. There is no single "ranking" to check anymore.
How video specifically gets cited
When AI cites a video, something happens that almost never happens with text: your brand comes along for the ride.
A text AI summary usually strips out who said it. But when an AI Overview cites a YouTube video, it carries the channel name, the thumbnail, the title, and your face. The model cannot detach the brand from the concept without breaking the citation. You get visible credit by default.
"Video is the single best way to generate signals that the AI actually needs to understand how your ideal audience is acting online, and to place you as a centrality within that community. If you're generating more signals with your audience than your competition, you're going to win the GEO game." — Jason Barnard
Two practical truths about which videos get cited:
- Helpful beats promotional. Product demos and how-to videos that solve a real problem are what surface in AI Overviews. Brand films and promos are not.
- Zero-click is the new normal. Sometimes AI names your brand without a link. Do not panic.
"If your brand turns up by name without a link, I wouldn't panic too much. It's kind of the new normal that we have to adapt to." — Carolyn Outhwaite, digital visibility expert
"Today's awareness searches are tomorrow's pipeline searches, and they're next week's deal closings." — Carolyn Outhwaite
How to optimize your video for AI search
This is the playbook. None of it requires a film crew.
1. Transcribe everything, and format it for a lazy reader. Break the transcript into chapter sections with H2 and H3 headers that match your YouTube chapter markers.
"LLMs are lazy. They don't really want to think. So a wall-of-text transcript is causing them too much thinking." — Jess Hennessey
2. Add an FAQ block with schema. Put a Frequently Asked Questions section at the bottom of the video page and mark it up with FAQPage JSON-LD. It mirrors the structure of the pages already getting cited.
3. Use embeds that include the transcript. Most embedded videos give AI nothing. An LLM-friendly embed bakes the transcript into the embed code so crawlers can actually read the video instead of treating it as a black box. Keep the video on a property you control.
4. Let your inbox write the content plan. Answer every real customer question on camera. Christopher Penn's whole strategy is this simple:
"Today someone asked me this question, and here's the answer. That's your strategy." — Christopher Penn
5. Favor speed and consistency over polish. The fastest, most consistent publishers win citations. Penn records unscripted, no crew. The point is volume of helpful answers, not production value.
"It's easier to talk than to type. Your language when you're talking is just more forthright and informal and direct and personal, which is all the hallmarks of good writing." — Andy Crestodina, Orbit Media
6. Pull the proof out of the video and onto the page. Stats, case studies, and testimonials trapped inside a video file are invisible to the model. Make them exist as readable text.
A warning on the easy button:
"If you're just clicking on a button and it's producing video and it's easy, you're doing it wrong. Because that doesn't have your soul." — Jason Barnard
What NOT to do
The single biggest mistake is treating AI search as a daily algorithm to chase. The tools are unsettled. A change you panic about today can reverse within a month.
"The best thing to do is to revert to best practices and what makes sense to your customers and what makes sense to you. Because if something doesn't make sense, chances are it's not going to last, or it's just a rumor." — Carolyn Outhwaite
A few specific traps:
- Don't cancel top-of-funnel. Awareness is where AI search does its work. Awareness searches become pipeline.
- Don't chase rankings as the target. Focus on what your buyer is actually asking and answer it.
- Don't pay for "AI visibility" prediction software. Personalization makes it impossible to know what any individual buyer sees.
"Anyone who is selling you software saying we can tell you what your AI visibility is, they have zero insight now. There's just no polite way of saying it, they're lying." — Christopher Penn
- Don't audit your way to one channel and copy it blindly.
"Don't assume that because X website is the number one cited source across all LLMs that that's where your strategy should be." — Adrian Dahlin, founder of searchtosale.io
- Don't expect a quick win, and don't quit early.
"You're not going to do a one or two month experiment with video and change your business." — Adrian Dahlin
How to measure whether it's working
Forget dashboards that promise to read AI's mind. Measure at your own front door.
The gold standard is a free-form "How did you hear about us?" text field, with no drop-down, on every point of intake: contact form, call center, email auto-response. A free-form field captures the exact tools and queries a preset list would never surface.
Once a month, run the responses through AI and count how many name ChatGPT, Claude, Perplexity, Google, or any AI tool as the source. Zero means your strategy is not working yet. A number that climbs month over month means it is.
FACT: Christopher Penn reports 11% of Trust Insights' business now comes from AI referrals, measured this way, in the last six months. "It is astonishingly high."
And set your expectations to the real timeline.
"Plan for 6 to 9 months. AI search visibility is a long-term play, not a quick win." — Kaleigh Moore
"Test, don't theorize. These engines are probabilistic, so they don't give the same answers consistently, but they give a kind of consensus." — Cassie Clark
Future-proofing: the two-year window
Here is the strategic bet. The models are forming their picture of your market right now, from the content that exists right now. Video is the most defensible way to be in that picture, because it carries your brand into the citation, it reads as community proof, and almost none of your competitors are doing it well yet.
"If agentic commerce is eventually going to be making a purchase on behalf of someone, and there's not a person pushing the button, it's going to be so much harder to be the answer when that happens if you've not already done the work now." — Cassie Clark
"Getting people on video, leveraging your humanity in this world of AI, that is one of the most advantageous things you can do." — Kaleigh Moore
The white space is open. It will not stay open. Start putting your answers on camera, wrap them in readable context, and become the brand the machines recommend before your category locks in.