Enter, the age of provenance — or the week AI started tagging itself

Something’s changed in artificial intelligence (AI). You will love it or loathe it. The world appears somewhat divided on this one. 

Anthropic⁠ says new Claude models launched in the European Union (EU) on or after 2 August 2026 will support machine-readable marking from launch. That will fan out globally. Anthropic says greater transparency via provenance data, revealing where content comes from, can give people useful context, while also helping it meet its commitments under the EU AI Act. 

Substack⁠ has introduced “Scan for AI text”, powered by Pangram, which estimates how much eligible writing is human-written or AI-assisted. Creators can also add a “How I make this” statement explaining their use of AI. 

Substack put its position rather neatly: “The problem isn’t AI, the problem is when you pretend something written by AI was written by a person.” 

LinkedIn⁠, meanwhile, is reducing the wider distribution of low-effort AI-generated material that lacks genuine perspective or substance. Importantly, LinkedIn is not rejecting AI-assisted writing. It says AI can help refine language, but that posts still need to represent the author’s own voice and perspective. Its conclusion is simple: “The ultimate value comes from the human behind the tool.”

These are three different interventions. Anthropic is working on provenance. Substack is introducing detection and disclosure. LinkedIn is trying to suppress generic synthetic content.

But they point in the same direction.

What counts as human work?

After several years dominated by generation and adoption, we appear to be entering the age of provenance. And that creates a much harder question. What exactly counts as human work?

• If the original thought is mine, but AI improves the structure, who created the work?

• If I write the first draft and AI edits it, is that AI-generated?

• If I develop the argument, build the framework and use AI to improve the language, where exactly does authorship sit?

I hate making PowerPoint presentations. If AI turns my thinking into a better-looking presentation because I would rather not spend six hours aligning boxes, has the machine created my work, or packaged it?

The client may not care. This week I presented my idea within a Claude deck to one of the largest brands on the planet. I told them it was a Claude deck, personally the idea was so robust it didn’t matter what wrapping it came in. They wanted a great idea, so I simply chose to box it up for presentation with speed. They loved it – they wanted the strongest thinking presented clearly, rather than an invoice reflecting my endurance and struggle with PowerPoint.

That exposes something AI is forcing knowledge workers to confront. Effort and value are not the same thing. If an experienced strategist can use AI to turn 10 years of domain knowledge into an excellent recommendation in two hours, that recommendation is not necessarily worth less than mediocre work that took somebody 12 hours to produce manually.

The more interesting question is where the intellectual value came from. And perhaps that gives us a more useful distinction between AI assistance and AI substitution.

AI proofreading, AI organising, plus how to package my strategy into slides is assistance? AI manufacturing an opinion I have never formed, which I then present as my own, feels different, as well as using AI to create insights and expertise I do not possess, experience I never had, or certainty I have not earned.

One accelerates capability and the other can substitute for capability the author is claiming to possess. That distinction matters far more to me than whether a detector decides a paragraph is 42% AI.

Even Anthropic acknowledges the problem. Its own documentation says marked content may have been processed by Claude without Claude being the original author. Someone may have used it to proofread, translate, summarise or convert material whose underlying ideas came from somewhere else. Equally, the absence of a detectable watermark does not prove that AI was not involved. 

In other words, machine involvement does not automatically establish intellectual authorship. Authenticity is not one thing. It sits across the original thought, research, experience, framework, draft, editing, expression and decision to publish. Reducing that entire chain to HUMAN or AI may be convenient for software. It tells us remarkably little about where the value originated.

There is also a governance lesson here.

I spent nine years working in social media and online safety⁠, and we watched platforms become deeply embedded in childhood, politics, relationships and public life before regulation caught up.

The UK is now following Australia and others, moving towards prohibiting under-16s from accessing certain social-media services from 2027, alongside further restrictions on potentially harmful online features. The measures follow years in which these technologies became part of ordinary childhood.

The recurring pattern is that governance often arrives after behaviour has already changed. AI is travelling considerably faster. By the time we agree on precisely how AI-assisted work should be labelled, hundreds of millions of people may already regard AI assistance as completely normal.

Power of economics

And there is a powerful reason for that. Economics. Humans constantly calculate effort, time and reward. If something that once required four hours can be completed well in 40 minutes, people will use the faster route. It may allow them to serve more clients, earn more money, leave work earlier or simply remove a task they dislike. The economics are too compelling.

That is why I remain sceptical of predictions that a backlash against AI content will somehow return us to a golden age in which everything valuable is lovingly handmade by humans.

Human work may acquire a premium precisely because it becomes scarcer. Originality may become more valuable. But convenience will continue winning often enough to reshape the market.

For that reason, the next phase of AI literacy may have less to do with knowing how to prompt and more to do with knowing where the machine should stop.

• Where should AI accelerate the work?

• Where should human judgement remain decisive?

• When should AI involvement be disclosed?

• When does assistance become substitution?

• And when should something remain entirely human?

At the moment, many of us are becoming rather pleased with our ability to “spot AI”. We recognise the predictable phrasing, immaculate structure, polished emptiness and strange universal tone. There is almost a club forming around the sentence: “That’s so AI.”

We should be careful. We are learning to recognise the fingerprints of today’s models. Tomorrow’s will have different fingerprints. Detection will evolve. Generation will evolve. Provenance will evolve. People will attempt to circumvent all three.

Value of human contribution

Which is why I suspect the enduring question will be what the human contributed. And that takes us somewhere more interesting. AI has radically reduced the cost of producing content. When another article, image, presentation, post or video becomes increasingly cheap to create, supply explodes. Scarcity moves elsewhere; attention becomes scarce, alongside trust, judgement, taste. Experience, to some degree, becomes scarcer.

A genuinely original point of view is harder to find, which is why I have relaunched my content agency to focus solely on thought leadership and domain authority⁠ that may become more valuable. The C-Suite has to stop confusing thought leadership with producing more content. Paddle fast, or get lost in the flood.

The valuable person will be the one who has actually done the work, made decisions, understood a market, noticed something others missed and developed a point of view worth hearing. Depth, provocation, humour, quirkiness, creativity.

The ability to understand a psychographic segment well enough to surprise them, move them or make them reconsider is now carrying higher value than ever.

Caveat: Human-created content has never automatically been good. We produced mountains of forgettable advertising, corporate waffle and derivative thinking long before ChatGPT arrived. AI simply makes it possible to manufacture vastly more of it, way faster.

Everyone has a viewpoint, value system, ethical boundaries and set of objectives. Which is why this new provenance spotlight will be loved by many, but loathed by many too. Sorry I can’t give you answers, only provocation.

AI and its morphing, squirming evolution are guaranteed to persist. No one knows what’s coming next week. Buckle up. DM

Dean McCoubrey is the founder of MediaWeb Group.