top of page
Search

When Every Video Becomes Your Video: The Personalisation Endgame

  • Writer: Adrian Munday
    Adrian Munday
  • Oct 12
  • 6 min read

ree

It's been a wild few weeks for AI announcements. Anthropic alone launched a new frontier model in Sonnet 4.5 and an experiment in Imagine with Claude. But two things caught my eye. Firstly Open AI’s Sora 2 video model and secondly Google’s incoming Veo 3.1 release.


The capabilities are jaw-dropping. Sora 2's cameo feature lets you create videos of yourself in any situation. Veo 3.1 reportedly generates much longer clips while maintaining character consistency. I've been experimenting with these tools myself (you might remember my AI influencer from earlier posts, built with Veo3).


This isn't another post about AI killing Hollywood or drowning us in synthetic content.

After tracking video AI developments and testing the tools firsthand, I've realised we're entering what I'm calling the "personalisation endgame" for video content. Creators and short-form marketers will soon test and adapt content for an audience of one.


You.


This will reshape marketing, education, and communication at a scale most organisations haven't begun to process.


With that, let's dive in.


From A/B Testing to Infinite Testing

Back in 2002, you actually needed budget approval to test two versions of a landing page. Before ChatGPT and tools like Lovable, that's just how things worked.


We've come a long way from those days, but speak to marketing teams today and they typically still only run tests with handfuls of variants. Video has always been the exception. Too expensive to produce. Too time-consuming to edit. You got one shot to nail it, maybe two if the budget allows.


Those constraints just disappeared.


We've crossed a threshold where generating personalised video variants has become not just possible, but trivially easy. The tools I have tested among many others can take a base video and, if you can fund it, create endless variations adjusted for viewer demographics, viewing context, prior behaviour, or literally any data point you want to drive the production.


Picture a future where brands generate product videos dynamically based on each visitor's browsing history. Looked at hiking boots? The outdoor gear video emphasizes durability and weather resistance. Browsed running shoes? Same product category, completely different narrative arc focusing on performance and training.


The shift isn't just quantitative. When you can generate micro-targeted variants, you stop thinking about optimisation in the traditional sense. You're not trying to find the one best version anymore. You're creating a dynamic system that matches content to context in real-time.


The Mathematics of Personalisation

What does this look like in practice? A traditional A/B test might compare two versions across 10,000 viewers. Statistical significance requires weeks of data collection. You learn which version performs better for your aggregate audience, implement the winner, and move on.


Now consider what happens when you can generate multiple variants and match them to viewer segments of 100 people each. Yes this is expensive today (as my AI subscription bill testifies), but cost of inference is falling continuously. When you generate at scale, you're not just learning "Version A beats Version B." You're matching each version to each customer segment and testing versions far beyond what you would have been able to do traditionally.


The data accumulation rate is exponentially different. You're gathering nuanced insights about what resonates with specific audience types that would take years to discover through traditional testing.


Personalisation progresses through increasingly refined levels of audience targeting: from a classic “one-to-all” model at the most basic level, through customer personas and micro segmentation, all the way to hyper-personalisation.  This is where each customer experiences an individualised journey defined by their real-time intent, prior interactions, and a deep, data-driven understanding of their needs, often leveraging AI and predictive models.


The Second-Order Data Effect: The Real Moat

The obvious value from these developments is the personalisation itself: better engagement, higher conversion rates, more effective communication. But that's only the first-order effect.  When people talk about AI killing the marketing industry they are talking about these first-order effects. 


My personal view is that new industries will emerge from the second-order effect that this technology introduces.


The real value is in the data you accumulate about what works for whom.


I’ve been thinking about the amount of data Netflix holds recently (for a work-related topic). It’s not the 3-4 petabytes of video content they carry that interests me, rather the multiples of that number they store around customer behaviour. Netflix's competitive advantage isn't that they can recommend movies (lots of companies can do that). Their moat is the massive dataset they've built about viewing preferences and patterns. They know what follows what, what keeps people engaged, what causes drop-off. That knowledge base becomes more valuable than the algorithm itself.


I have a personal passion for this kind of ‘telemetry’ (which some of you have already discovered) which will likely to be the subject of another blog.


With US TikTok effectively relaunching the platform for the West and Sora introducing AI social media, new platforms will emerge over the next few years. Those who build early will accumulate the datasets that become tomorrow’s competitive moats.


The Ethical Dimension

I can’t ignore this as personalisation at this level raises significant, legitimate questions about manipulation and privacy. Cambridge Analytica wasn’t that long ago. When video content dynamically adapts to exploit individual psychological triggers, we're entering territory that many will find uncomfortable.


As someone who spends their days thinking about risk, I believe organisations need clear ethical frameworks for personalisation. Some boundaries I'd suggest:


Transparency about personalisation. Viewers should understand when content has been adapted for them. Some level of disclosure seems appropriate, even if it slightly reduces effectiveness.


Limits on behavioural triggers. Just because you can personalise based on emotional state indicators doesn't mean you should. Some targeting dimensions are ethically questionable even if technically feasible.


Data governance. The datasets that power personalisation contain insights about human psychology and behaviour that deserve protective custody. Clear policies about data use, retention, and access matter.


These questions don't have simple answers, but ignoring them isn't an option. The organisations that think seriously about ethical personalisation now will be better positioned than those who treat it as purely a technical challenge.


Principles from organisations like the Institute of Electrical and Electronics Engineers have developed comprehensive frameworks for ethical AI design and legal frameworks such as Europe’s GDPR or the California Privacy Rights Act provides overarching consumer protections. But are these adequate for this coming wave? I’m not so sure.


The Bottom Line

We're witnessing video's evolution from a broadcast medium to a bespoke, individually adapted format. The implications extend far beyond marketing effectiveness. We're changing the fundamental economics of video content production and the business models built on top of it.


But as I said earlier this isn’t about the hollowing out of the marketing industry. Second-order effects will launch entirely new industries.


Look at the early days of Facebook or Instagram, or later TikTok. Organisations or creators that were in early gained massive advantages at low cost of customer acquisition.


The confluence of these trends has created a new and complex paradigm. It presents an unprecedented commercial opportunity: the ability to achieve true one-to-one engagement at a scale previously unimaginable, tailoring every interaction to the individual's context, behaviour, and inferred intent.


Yet, this opportunity is shadowed by significant strategic and ethical risks. The pursuit of hyper-personalisation can easily cross a line into intrusion, triggering consumer alienation and actively damaging brand equity.


As a result, organisations that recognise this shift early, that start building the capabilities and datasets alongside the appropriate ethical frameworks, will have genuine competitive advantages.


Until next time, you'll find me on a Saturday morning, generating personalised video variants for the audience of this blog and wondering whether my enthusiasm is coming through in Version A but not Version B...


Resources & Further Reading

Industry Context

  • HubSpot. (2025). 2025 State of Marketing Report. HubSpot, Inc.

  • Willemsen, B., & Gillespie, P. (2025). Gartner Marketing Personalization Survey (2025). Gartner, Inc.

  • Wistia. (2025). State of Video Report 2025. Wistia, Inc.   

  • Gartner, Inc. (n.d.). Gartner CX Maturity Model.

  • Gartner, Inc. (2025). Magic Quadrant for Personalization Engines, 2025. Gartner, Inc.   


Practical Tools

Ethical Frameworks

  • California Office of the Attorney General. (n.d.). California Consumer Privacy Act (CCPA). State of California Department of Justice. https://oag.ca.gov/privacy/ccpa   

  • Data & Society Research Institute. (n.d.). Algorithms and Publics. from https://datasociety.net/research/algorithms-and-publics/   

  • European Parliament and Council of the European Union. (2016). Regulation (EU) 2016/679 of the European Parliament and of the Council of 27 April 2016 on the protection of natural persons with regard to the processing of personal data and on the free movement of such data, and repealing Directive 95/46/EC (General Data Protection Regulation). Official Journal of the European Union, L 119, 1-88.   

  • The IEEE Global Initiative for Ethical Considerations in Artificial Intelligence and Autonomous Systems. (n.d.). Ethically Aligned Design: A Vision for Prioritizing Human Well-being with Autonomous and Intelligent Systems. IEEE.   


 
 
 

Comments

© 2023 by therealityof.ai. All rights reserved

bottom of page