Morten E. Wulff

I went to a HeyGen workshop in London this week expecting to learn more about AI video. I left thinking about AI visibility. The workshop was genuinely hands-on. We created videos with AI avatars, and saw how quickly you can go from an idea to a polished piece of video content. I've been hearing for many years: “We need to create more video. We just don’t have the time.” That’s the biggest barrier. It’s not necessarily the production budget, but the commitment! 👉 Finding a person to be on camera 👉 Finding a location 👉 Writing the script 👉 Recording 👉 Editing And then doing it all again next week! AI is changing that, but it's another reason I think this matters. Video is no longer just a content format. It is becoming part of the information layer that AI uses to understand brands. YouTube is increasingly appearing as a citation in AI-generated answers. The title, description and transcript give AI systems a rich layer of information they can understand, retrieve and cite. So when we think about #GEO, #AEO we shouldn't only be asking: “How do we get our website into ChatGPT?” We should also be asking: “What information about our brand exists for AI to discover, understand and cite?” That could be 👉 Your website 👉 A review 👉 A Reddit discussion 👉 An expert article 👉 A retailer page 👉 And increasingly, a YouTube video This is exactly why the work we’re doing at 3RD around AI visibility is so interesting to me. The AI-mediated customer journey is not built from one channel. AI is pulling information from an entire ecosystem, and video is becoming an increasingly important part of that ecosystem. And this isn’t just an idea. A few years ago, we actually built a YouTube channel with HeyGen for one of my clients: https://lnkd.in/e9aiVKAs. The doctors behind the content are real doctors. But they aren't the ones appearing and speaking in the videos. Their AI avatars are. That means they can turn their expertise into a scalable video library without spending hours in front of a camera every time. This is where I think the combination of 3RD + HeyGen + YouTube gets really interesting. AI can dramatically lower the cost, and the time commitment of creating video content. And that content can then become another source through which AI systems understand your expertise, your products and services, your category and ultimately your brand. We are moving from: Create content → reach people to: Create useful information → make it discoverable → become part of what AI knows and recommends That is a very different content strategy. And probably one of the biggest shifts happening in marketing right now. Huge thanks to Bin Liu, Audrey Van Zee and Teodora P L from HeyGen team for a very practical and thought-provoking afternoon, and for making the “we don't have time to create video” excuse a little harder to use. 😉 #geo #generativeengineoptimistion #aeo #aivideo #videoavatar #heygen #youtube #3rd #aivisibility

Happy to share that I’ve obtained a new certification: Agent Engineering from Profound! Amazing to build these agents for InvestorFlow and collaborate with incredible marketing minds in the process including ashton dunn — thank you!

HYBRIDCASUAL ADS “SORT EXPRESS!” BY CIRCLE GAMES One more Turkish studio is blowing up in the hybridcasual ecosystem! 🚀 ✨ Circle Games 💫 Yesterday, they announced that they had received a $25 million Series A investment. 💰 The funding round was led by the Chinese giant Tencent 🇨🇳, with participation from e2VC, Bitkraft, APY Ventures, and A16Z/Speedrum. Investor euphoria is no accident. 🎯 In addition to an experienced team that includes former Dream Games and GoodJob staff, among others, Circle is scaling a hybridcasual title with the potential to become a massive hit: "Sort Express!" 🧸🥤 🥫 Circle found an uncovered gap in the mobile gaming market. 💡 "Sort Express!" is a shelf Sort Puzzle game (a trend with millions in in-app purchase revenue and strong download numbers) that adds a robust "Royal Match"-style metagame, alongside exceptionally polished game art. 🎨 The charts show steady revenue growth over the last few months, with formidable metrics. 📈 Their competent UA team has also sustained this success. 👏 One of the most successful concepts in their ad creatives is a simulation of the Sort Puzzle mechanic using colored blocks. 🟩 🟨 🟥 Although this concept doesn't exactly match the game, the block-movement logic and the ASMR generated by the blocks fitting together evoke its gameplay. According to Ad Insights from GameAnalytics, video ads with this concept have reached hundreds of millions of impressions, with a significant impact in the US 🇺🇸 market. Ad network: 100% AppLovin. I speak with many leaders of small and medium-sized gaming companies, and one of their biggest dreams is to become a self-publisher. 💗 It looks like Circle is yet another emerging Turkish company achieving just that. 💪 Yes, the dream of self-publishing can become a reality. This post is brought to you by GameAnalytics.

The Pets Table came in skeptical. We respect that. They ran an incrementality test with Haus to verify our numbers. We love that even more. The results? Truly incremental, strong customer quality. "I'm definitely going to have AppLovin as part of my marketing mix." — Dovas Zakas, Co-Founder & CMO, The Pets Table. #DTC #ecommerce #paidmedia #performancemarketing cc: HelloFresh

Why I’ve been struggling to create an updated version of our famous Funding Napkin for a while. ;-) More seriously - we will release an update at some point and it will be funny, but meanwhile, check out all of the data published by Carta. It’s honestly a much better, bigger dataset than what we collected for our napkins.

"GEO is moving fast, shouldn't we wait and see where it lands?" Sure…if you don’t mind the costs. Our buying behavior is heading one way, and a new Danish report backs it up: The “Payment Report Denmark 2026” (Dankort & Nets Nexi) finds 56.5% of Danes used AI for purchases/bookings in the past month, and 70.5% would use a secure payment solution inside a chatbot rather than being redirected. I get the worry: some AI outputs change with model updates, so nothing is guaranteed. But most of the work you should do isn’t fragile. When I hear “too early” it usually means one of three things: 1. Other priorities are more pressing 2. The field might change so much the work is wasted 3. It doesn’t feel urgent yet Only #2 is a real AI argument. #1 and #3 are timing and resourcing issues. The answer there is to start small, not to wait. What will last across model changes: - AI crawlers can access your site, pages render cleanly, sitemap present - Organization schema + SameAs links so models connect your name to you - Pages that clearly answer real buyer questions What won’t last: - Your rank for one specific prompt - The exact answer on any given day (models are probabilistic) What does waiting cost? Models re-crawl and re-learn on their own schedules. There’s a days-to-weeks lag between publishing a fix and seeing it in answers — so late starts produce late results. You also won’t know where you stand today: which competitors AI names instead of you, what it says about you, and which sources it relies on. Start small: - Make sure AI can read your site, knows who you are, and finds you on trusted third-party sources, that helps across models. So don’t bet on a single market outcome: start small, build lasting foundations, then let data tell you if it’s worth scaling. - Measure 10–15 prompts important to your business in one market on the models your buyers use - Get first results within an hour and a full picture after ~24 hours - Citation share is an early signal — when AI starts citing your pages, mentions tend to follow PS. If you’d rather start without a tool, check these three things this week: 1. Your robots.txt lets AI crawlers in (GPTBot, ClaudeBot, etc.) 2. Your site has Organization schema with SameAs links to your official profiles 3. Ask ChatGPT and one other LLM five questions your buyers would ask; note who’s named and which sources are cited If the answers surprise you, that’s your reason to measure it properly.

Fik du set Meta Connect denne uge? Det gjorde aktiemarkedet i hvertfald! (jeg har smidt et link i comments hvis du ikke gjorde) Det mest interessante fra eventet var måske hvordan Muse - din personlige AI agent powered by Meta, der nu er den mest downloadet App i USA - tvinger os til at forholde os til spørgsmålet omhvad vi gør når AI-agenter begynder at bevæge sig fra at finde information til faktisk at handle på vores vegne. Meta viste bl.a., hvordan Muse kan se det samme som dig gennem AI-brillerne, forstå hvad du kigger på og koble det til shopping, betaling og andre services. I mine øjne tydeliggør det en fundamental ting: Vi har i mange år designet vores digitale tilstedeværelse til mennesker. Nu skal vi også begynde at designe den til agenter. Det er selvfølgelig ikke nyt. Du kan let finde 1000vis af andre der har sagt dette de sidste 2 år, men med Muse bliver det lige pludselig ekstremt konkret. Og for marketing tror jeg især, det har to konsekvenser: 1. Du er nødt til at sikre at dine kunders agenter finder dig, forstår dig og  konkluderer at du er mere relevant end alternativet. Først var det Google. Så kom sociale platforme. Nu bliver AI endnu et sted, hvor virksomheder skal være synlige og forståelige, men denne gang er det ikke kun for mennesker. AI Discovery bliver en reel faktor i alle kunderejser og virksomheder bliver nødt til at have styr på om LLM'erne forstår hvem du er, hvad du tilbyder, og hvornår du er relevant? Jeg har den seneste tid gravet mig ned i dette emne og det er både ekstremt fascinerende og det går ekstremt hurtigt. Faktisk fandt jeg en rapport (link i comments til Impervas seneste Bad Bot Report) der estimerer, at mere end 53% af webtrafikken i 2025 var automatiseret. Det er selvfølgelig langt fra alt sammen AI-agenter. Men det siger noget om retningen: Det er ikke længere sikkert, at den primære “besøgende” på din digitale tilstedeværelse er et menneske. 2. Marketing skal måske samtidig længere op i tragten. Det lyder måske paradoksalt. Men hvis agenter i stigende grad overtager dele af search, comparison og fulfilment, bliver noget af det, vi i dag kæmper om nederst i tragten, automatiseret. Til gengæld bliver kampen om at skabe behov, forme kategorien og være det brand mennesker forbinder med løsningen endnu vigtigere. For inden en kunde siger til sin Muse (eller anden agent): "Find den bedste løsning på X og køb den for mig" …skal kunden først have forstået, at X er et problem, hvilken type løsning der findes - og måske allerede have en idé om, hvilke brands der er relevante. Agenten vil højst sandsynligt kende kunden og vide hvad hendes præferencer er ift. brands. Så måske får vi brug for at tænke på marketing til to audiences samtidig: Humans and agents. Vi skal både være top of mind hos mennesket og top of model hos agenten.

Anyone in my network working at OpenAI , we are growing and building aggressively fast, where are our startup credits so we can move faster?

AI has an opinion about your brand. Brand Perception shows where it came from 🔦 Niklas Springer and Antonia Breitenfellner walked through it live. The recording covers: → How LLMs describe your brand, and how that changes per model → The competitors they name organically — not the ones you track → The objections they raise against you before a buying decision → The claims they get factually wrong about your product → Where all of it comes from, source by source Recording and Q&A 👉 https://lnkd.in/eNwdGaGV