Published Media

Here's something urgent you need to know about OpenAI's plans for adverts.

They're introducing ads inside ChatGPT responses. Not boxed-out ads or feed interruptions we instinctively recognise as advertising.

It's possible the advertisers themselves won't know exactly what the ad will say when ChatGPT serves it up.

We won't be evaluating a product. We'll be getting advice that promotes a product, paid for by its producer.

You can see how this works innocuously enough — ask for 'shoes like these' with a photo after a night out.

But you don't have to think hard to see how it works dangerously. Corporate interests routinely align with harm — junk food, tobacco, gambling. The persuasive capacity of LLMs to shape how we see the world is already proven. I'm not anti-advertising. I quite like getting prompts to explore products I might enjoy, and I don't mind personalisation (even if I do mind the business models that drive it).

But this breaks a social contract that's pretty much survived the entire history of advertising: advertise all you like, but tell us you're doing it.

A technology used by hundreds of millions of people to shape their worldview, available to the highest bidder with no transparency? That's alarming.

We need to keep adverts and advice separate. We need new standards. Fast.

What's coming for communications in 2026? Here are six predictions.

One of them will definitely be wrong - but which?

1. Attention begins to shift from social media to chatbots

Social media loses some of its grip on our screentime as we start spending more time in conversation with AI. Ensuring your work is 'in the mind's eye of the AI' becomes a crucial communication challenge - restyling content norms just like SEO and keywords did decades ago - and each of us drifts further into a more personalised and less sociable online existence.

2. AI personalities become socialised

We are all familiar with the quirks and demographics of different social media platforms - TikTok for the kids, Facebook for pensioners etc. In 2026 we'll start to get more sense of who uses different LLMs and the quirks of their design. As we do, we realise the need to develop different channel strategies for different LLMs - reaching the institutional management class through Co-Pilot and the new economy through Gemini.

3. The slop backlash gathers pace

Even in the highly polarised US, AI slop can unite voices across the political chasm. Similar numbers of Republicans and Democrats dislike the use of AI video. Platforms like TikTok have said they will put 'slop filters' in place but the problem is - as we know from their failed (and now largely aborted) efforts to identify and remove harmful content - platforms can't police themselves. Expect to be wading through slop for a while yet.

4. An AI scam becomes a scandal

A company's AI chatbot is manipulated into offering up confidential information and the material extracted is immediately analysed with AI to set in motion a PR disaster (or whistleblowing triumph!) that is almost over before the comms team can open their laptops.

5. AI gets political

The use of AI by politicians continues to creep into public life, with firm numbers emerging around how often MPs use LLMs to draft routine correspondence and even speaking notes. But politicians themselves start to wake up to the distorting effects of AI when lobbyists begin deploying AI agents to flood their inboxes and casework systems - making it almost impossible to separate human signal from AI noise.

6. The bubble is a red herring

The AI bubble bursts or deflates - but the real challenge isn't the economic impact. It's the way that collapsing AI share prices enable people to convince themselves that the impact of AI on our lives won't be that big after all. 'Bubble Cope' sets in while AI technology is woven further and further into the fabric of our lives, hoovering up our data and reshaping how we see the world. Growth and productivity continue to stagnate - but AI companies are growing again by the end of the year….

You know that feeling your phone must be listening to you?

You ain't seen nothing yet.

Social media companies don’t need your microphone to know you better than you know yourself.

Apps already collect insane amounts of data: your typing patterns and deleted messages, how long you pause on each post, where you are, who you message most at 2am, playlist changes that reveal your mood, even items you add to your basket but never buy.

But LLMs can take this to a whole new level. With the data they collect they can understand context, predict needs, and take action.

Now imagine AI apps integrate with other data on our phones and start to send you behavioural prompts based on everything they know:

* "You've been researching symptoms for three days. Here are questions to ask your GP—and which ones to prioritise.

* "You’ve spent £250 more than usual on takeaways this month and your sleep is off. Stress or celebration? Either way, let's talk.

* "You're near your mum's house. You said last week you've been meaning to visit. You have 90 minutes free—want me to let her know?

* "You've walked past that gym you joined 6 times this month. Want me to book you a class or help you figure out what's holding you back?

* "Your breakup playlist is back and you're home alone most nights. Want to talk, or should I suggest some plans?

Does this feel dystopian?

It might well end up that way unless we pressure AI product designers to build with our agency in mind, not just engagement metrics.

At what point does "helpful" become "controlling"?

Would you want AI nudging you towards the gym? Managing your family obligations? Diagnosing your emotional state from your Spotify?

I hate to break it to you, but the media world you learned to navigate no longer exists.

And the numbers prove it.

This year we'll upload 364 exabytes of data to social media.

To put that in perspective: back in 2000, experts estimated that every single word ever spoken or written by humans since we first started talking amounted to just 5 exabytes.

We're now creating 70x that amount of information every single year.

Here's some more stats:

* One in six American teenagers is "almost constantly" scrolling TikTok or watching YouTube.

* People collectively watch a billion hours of YouTube content daily. The BBC's biggest news day? Maybe 2 million hours of viewing. Total.

* Some food influencers who reach over a million people every time they post. Meanwhile, the restaurant critic at a major newspaper is lucky to get 50,000 readers a week.

* The influencer economy is heading toward $500 billion by 2030, while UK newspaper sales have collapsed from 20 million copies a day to just 2 million.

And....two-thirds of people trust influencers more than they trust brands.

So let me ask you something...

When was the last time you honestly evaluated whether your communications strategy matches this reality?

Because I keep meeting brilliant comms professionals who's bosses want them to spend their time chasing newspaper coverage while creators with bigger audiences than national broadcasters are right there, waiting to be engaged with.

Quick reality check:

*Do you actually track conversations beyond traditional media? *Are you building real relationships with creators in your space? *Have you got a plan for when disinformation hits? *Are you using data to make decisions, or just gut instinct? *What percentage of your content is actually video vs text? *Are you experimenting with AI, or avoiding it?

If you're answering "no" to most of these, you're not behind the curve.

You're fighting a war that ended five years ago.

The information environment hasn't just changed - it's been completely rebuilt while we weren't paying attention.

The question isn't whether you should adapt.

It's whether you can afford not to.

If you’re not using AI in your day to day work then the person next you probably is.

Doesn’t feel like it?

That’s because of what I call the ‘Shy AI’ phenomenon.

This is where there’s far more regular users of AI than there are people willing to be seen using AI.

There’s good reasons for this -  rightly or wrongly being seen to use AI can open up a fear of judgement, job insecurity, ethical quandary and even existential angst.

But still, recent research shows that in leading companies 73% of staff are using AI on a weekly basis but…and yet when it comes to organisational adoption of AI things are moving much more slowly.

This is also understandable/sensible. The barriers are bigger and the stakes are higher. It’s one thing asking AI to draft social posts, it’s another thing letting it loose on your CRM.

In our work at Rootcause we try to stay sceptical about wild claims of just how fast and how far the AI revolution might go but we do believe that in some areas the revolution is underway.

One of the areas where we think things will happen is in information classification.

AI is good at making near-instant sense of large amounts of unstructured data and organisations have a lot of common knowledge needs where this ability can be useful.

There are lots of data sources where it could be possible to extract new insights.

*Emails *Surveys *Media content (including video and images) *Social Media Data (Proprietary and Open Source) *Publication Archives

And AI can be really helpful in devising and developing new database systems where AI tagging and other capabilities can add extra functionality and increase user-friendliness.

Rootcause Global exists to try and demystify influence in the age of AI.

Our latest experiment is focused on AI and data classification, so that we can show you how this might be useful in the future. If you want to be first to hear about the findings sign up to the Newsletter in the link below.