Written by Dom Smales
A lot happens in ten years. I was on stage somewhere in 2016 delivering a talk. Thirty talent on my roster. 55.5m Twitter followers. 21.3m YouTube subscribers. That was pretty much the whole map back then – Twitter and YouTube. Instagram was growing fast but nowhere near the scale it got to, and TikTok wasn’t really a thing yet.
A decade has flown by since that point. When I look back at pictures from that event – aside from the fact my skinny jeans were way too tight (even for that era!) – it makes me think a lot of what we did back then and about how it’s going to become valuable again. For slightly different reasons.
Search has moved
Search is changing faster than I’ve ever seen it change. More and more discovery now starts with an AI answer rather than a page of Google links. Pew Research tracked real Google users and found people clicked a result on 15% of normal searches, but only 8% when an AI summary appeared. Links inside the summary itself got clicked 1% of the time. I think the ‘answer’ is becoming the destination.
And it isn’t just information. John Lewis said this month when talking about their new long form YouTube show that the share of its customers finding products through ChatGPT and Gemini went from 0.3% to 2.5% in a year. Small numbers still, but eightfold, and across every age group. They launched their YouTube chat show, because the models favour third-party advice and live content. A department store is making long-form talk content so that a machine can find it. That should tell every Talent Manager something when it comes to thinking about strategy for growing talents’ audiences or increasing their relevance for specific categories.
What decides whether your cleint turns up in one of those answers? It isn’t reach. It’s whether an LLM can read them, quote them and attribute them.
The view count doesn’t count (here)
I find this really fascinating. OtterlyAI looked at over 100 million AI citations across ChatGPT, Google’s AI Overviews and AI Mode, Perplexity, Copilot and Gemini. For YouTube videos that got cited, views, likes and subscriber numbers had basically zero correlation with how often they were cited. More than 40% of cited videos had under 1,000 views. Long form made up 94% of YouTube citations, Shorts under 6%. The thing most linked to being cited was the length of the description. The text around the video!
YouTube has become one of the most cited sources in AI answers (Peec AI’s analysis of 30 million sources puts it second only to Reddit) because a video there turns up as a document; named creator, title, description, transcript, chapters, timestamps etc. etc. A Reel or a TikTok is, to a retrieval system, close to an invisible file. Sixty seconds is roughly 150 words, often sat behind a login so there’s nothing to chapter or cite.
So it’s about reference value, not views. Long form reference style content gets pulled much more, while the big viral numbers sit there doing very little for visibility on Ai.
What this means if you manage talent?
We’ve spent so long pricing creators on reach. Rate cards, media kits, pitch decks, all built on views and followers as well as engagement. We did it at Gleam for a decade. If AI answers become a real front door to discovery, that’s only half the story. Maybe less.
Think about what it does to the mid-tier. A skincare expert with 40k subscribers and two hundred properly explained videos could be more visible inside an AI answer than a creator with ten times the audience. That’s a very different roster conversation. And a different conversation with brands, who will start asking whether your talent shows up when someone asks ChatGPT for a recommendation (and some already are). We should also be including consideration for AI search when we suggest content responses to brand briefs.
It also changes what the back catalogue is worth. Five years of well-described long-form is an archive that keeps getting quoted. That’s a great asset, and assets are things we should be building on and owning, not just monetising once. Regular readers will know I have a bit of a thing about that.
For creators, the practical stuff isn’t glamorous but it matters. If you are interested in exactly what creators need to do and include to be more visible to AI then get in touch and I will explain. One thing to point out is none of this is a reason to stop making short-form. YouTube says Shorts get 200 billion views a day. It’s still where new audiences find you. It just isn’t the master asset of the future anymore. It’s distribution and general reach. The long form is the thing that gets referenced.
Back to 2016
One more thing, and it’s a gut feeling more than a data point really. I think audiences are bored of the perfect optimised flat lays and polished lives presented on social. Now that AI can make ‘perfect’ in seconds, ‘perfect’ stops being worth very much. What’s hard to fake is a real person talking at length about their real life and the thing they actually know about.
Which is…vlogging. Real lives, real people, long conversations and trust earned slowly. It puts us right back in 2016 again (minus the tight jeans!)
The creators we worked with back then at Gleam didn’t make long videos to game an algorithm. There wasn’t really one to game. They made them because that’s how you built a relationship with an audience. It turns out it’s also now how you get a machine to take you seriously.
So maybe we all need to be a bit more 2016 in order to be more visible in 2026.
