Gary Vaynerchuk published something this week that should interest you if you lead marketing, even if you don’t have a single video on YouTube.

He says YouTube Shorts went from the fifth or sixth most important platform at his agency to potentially number one. And the reason he gives isn’t the views. It’s that every piece of content you upload today into Google’s ecosystem is a deposit into the battleground coming next: AI answers.

That word, deposit, is the one I want you to read twice. A Short isn’t uploaded for today’s viewer. It’s uploaded for the model that will answer your customer tomorrow.

The thesis of this post, and it’s for your marketing team, not your CTO: the game shifted from optimizing for reach to optimizing for retrieval. Most teams still produce so people will watch, not so the AI will cite. And that difference, which looks minor today, will in a couple of years separate the brands that show up in the answer from the ones that don’t exist.

Reach is no longer the game. Retrieval is.

In the comments on Vaynerchuk’s own post, someone summed it up better than the post did: platform strategy is shifting from reach optimization to retrieval optimization. It sounds technical. It’s the most concrete thing you’ll read this year.

Optimizing for reach is what your team already knows how to do: chase views, likes, followers, engagement. Today’s metrics, visible, the kind that feel good in a report. Optimizing for retrieval is something else: getting your brand to show up when a prospect asks ChatGPT, Perplexity, Gemini, or Claude “who does this in Central America?” Either you’re in the answer, or you don’t exist for that person. There’s no second place in an AI answer the way there is on a page of Google results.

That’s AEO, answer engine optimization, and I’ve written that as a discipline it didn’t exist 18 months ago. What’s new in Vaynerchuk’s argument is where the data those engines will cite comes from. It doesn’t come from nowhere. It comes from the content brands upload today into the ecosystem where those models train and search. That’s why a Short isn’t just distribution. It’s raw material for an answer that doesn’t exist yet.

Why video, and why now

Vaynerchuk’s bet has a simple logic. He believes Gemini will be a big winner in the AI race, and that everything you upload to YouTube, owned by Google, is doing two things at once: it gives you access to a platform with enormous consumption today, and it deposits into the ecosystem that will feed AI answers tomorrow.

And the barrier to entry is almost zero. That’s the part your team needs to hear. If you have years of Reels, TikToks, Facebook videos, podcast clips, talk recordings, or educational material, you already own the inventory. Nothing to reinvent. Step one is simply publishing it where you weren’t publishing.

But here’s the trap, and it’s where almost everyone will get it wrong. Low barrier doesn’t mean easy win. Uploading the old file is the floor, not the strategy. And the difference between those two is exactly what separates a deposit from a bit of noise.

Reposting old Reels is not making a deposit

Another commenter on the same post said it flat: don’t mistake reposting your old Reels for being native to how people watch on YouTube. He’s completely right. Dumping the same Instagram vertical with the same watermark and the same caption tells nobody anything new, human or model.

So what counts as a real deposit? Four concrete things, and none of them needs a big budget.

The first is editing native to the platform. Same raw material, different finish: pacing, cut, and support copy built for how it’s consumed on YouTube, not copied from how it was consumed somewhere else.

The second came from a consultant in that thread and it’s the most useful: structure your support copy as question and answer. “Will rates drop next year?” Then the short answer. Why? Because that’s exactly how people talk to an answer engine, and content that already arrives in that format is far easier to cite. Your FAQ stopped being a boring section of the site. It’s the native format of the AI era.

The third is depth, not spread. One good video helps. A thousand answers around one subject build authority a model recognizes. The library beats the single post, because AI rewards consistency and topical depth, not the lucky strike of an isolated viral.

The fourth is text the AI can index: transcripts, real descriptions, not three emojis. The model still reads far better than it watches. If your only deposit is a video with not a word around it, you deposited into an account the bank can’t read. This is the other face of something I’ve written before, about having a brand an agent can read: if the AI can’t process what you publish, then for the purpose of the answer, you didn’t publish it.

What we do about it at IQ Source

I’ll tell you with the most honest example I have: our own blog. Every question in the FAQ block at the end of a post is written as a query someone would type into a search box or say to a voice assistant, not as a question that only makes sense if you read the article. That isn’t a copywriting detail. It’s AEO applied to ourselves, and it’s the proof we practice what we sell.

AEO is the concrete piece of IQ Source for the marketing team that understood the game changed. It doesn’t replace your reach work, it complements it: on top of measuring how many people saw the content, we start building so the AI uses it as a source. Question-and-answer structure, topical depth that builds a library instead of a pile of loose virals, and indexable text around each piece. The goal isn’t more views. It’s that when your customer asks a model who does this, the answer says your name.

Before you close the week, ask your team one question. Not “how many views did we get this month?” They already know how to answer that. Ask the other one: if a prospect asked ChatGPT today who does what we do in our market, would we show up? If nobody’s sure, that’s exactly the account your brand isn’t depositing into yet.

Turn your content into an asset AI will cite