AI Made Content Cheap. Trust Is Getting Expensive.

01 / Supply

We don’t have a content shortage

Consider what happens while casually scrolling now.

You see a photograph.

Is that real?

A glowing product review.

Did they actually use it?

A beautifully written article.

Was this researched, or generated from five other generated articles?

An influencer recommending an app.

Do they genuinely like it, or was this entire thing assembled by an AI marketing system?

A video of somebody saying something outrageous.

Did they actually say it?

None of these questions are entirely new. Humans were lying, Photoshopping and writing suspicious Amazon reviews long before ChatGPT arrived.

The difference is scale and cost.

Producing plausible material no longer necessarily requires much time, expertise or even a person particularly invested in whether the result is true.

That matters because trust has always relied partly on signals of effort. Reporting takes work. Testing a product takes time. Building expertise takes years. Photographing something generally requires being there.

When convincing imitation becomes dramatically easier, those old signals become weaker.

And the public already seems uncomfortable with the shift.

In a February 2026 Pew Research Center survey of 5,119 U.S. adults, 49% said they had used an AI chatbot, up from 33% in 2024. Yet 63% said AI was advancing too quickly, and 40% expected AI’s long-term effect on society to be negative compared with 16% who expected it to be positive.

That combination is interesting.

We’re not simply rejecting AI.

We’re using it while simultaneously becoming less certain about what its arrival means.

02 / Uncertainty

The real cost of fake content isn’t always believing the fake

This is where the trust problem gets stranger.

We tend to imagine synthetic media causing harm like this:

Fake thing appears → person believes fake thing → damage occurs.

But research suggests there is another path:

Fake things become common → people realise anything could be fake → confidence in real things also falls.

A well-known 2020 experiment on deepfake political video found synthetic media could increase uncertainty even when it failed to fully deceive viewers. Researchers Cristian Vaccari and Andrew Chadwick warned that the danger wasn’t merely persuading people of falsehoods, but weakening confidence in information more broadly.

Other researchers describe a related phenomenon as the “liar’s dividend.”

Once everyone knows convincing evidence can be manufactured, genuine evidence becomes easier to dismiss too.

“That video is fake.”

“That screenshot was generated.”

“That recording is AI.”

Sometimes they will be.

Sometimes that accusation itself will be the lie.

Academic work on deepfakes describes this erosion of trust as a serious secondary consequence of synthetic media: the existence of believable fakes can provide cover for people who want inconvenient real evidence dismissed.

That’s the part of this conversation I find more important than whether an AI-generated Instagram post has six fingers.

03 / Creators

The creator economy is our accidental experiment

Nowhere is this tension easier to see than influencer marketing.

Creators are valuable to brands for a reason traditional advertising has historically struggled to manufacture: people believe them.

A creator has spent years accumulating an audience through opinions, experiences, recommendations and personality.

A brand effectively says:

People trust you more than they trust our advertisement. Can we borrow some of that?

There is serious money attached to that relationship.

The Interactive Advertising Bureau projected U.S. creator advertising spending at $37 billion in 2025, with industry estimates putting 2026 spending around $44 billion. Yet in IAB’s survey of 453 creator-marketing industry respondents, 95% of brands reported concerns about using AI in creator marketing, with the top concern being loss of authenticity or human connection.

That’s a fascinating contradiction.

Brands want AI because it makes content cheaper and faster.

Brands want creators because they feel human.

Now they are trying to combine the two without accidentally destroying the thing they were paying for.

Business Insider reported this week that some creators are commanding large premiums for AI sponsorships—or avoiding them—because promoting AI products can trigger audience backlash. Some deals reportedly reach seven figures.

The creator isn’t simply selling advertising space.

They are spending a little piece of accumulated trust.

And trust, unlike content, is difficult to regenerate on demand.

04 / Judgement

This is where the “human premium” begins

It would be tempting to conclude:

Human content good. AI content bad.

Unfortunately, reality has refused to organise itself that neatly.

Research comparing human and AI influencers has found that AI influencer endorsements can reduce perceived brand trustworthiness and purchase intention compared with human influencers. Two experimental studies published in the Journal of Product & Brand Management found the effect was linked particularly to audiences perceiving less agency—a weaker sense that there was an intentional mind behind the endorsement.

But another 2026 experiment involving 721 women aged 18–40 found that AI influencers could feel more trustworthy and relational when audiences perceived stronger human-like mental qualities and self-disclosure. Interestingly, perceived mental humanlikeness mattered more than simply appearing physically human.

That’s an important wrinkle.

Perhaps what audiences value isn’t necessarily human production.

Perhaps they value human judgment.

There is a difference.

AI can help a photographer edit an image.

AI can help a journalist organise research.

AI can help a creator clean up a script.

AI can help someone articulate an idea they already genuinely hold.

None of those automatically removes the human value of the work.

The distinction that may increasingly matter is not:

Human vs. AI

but:

AI extending human thought vs. AI replacing the need for anyone to think at all.

05 / Signals

The authenticity paradox

This produces a slightly strange future.

For decades, the internet rewarded polish.

Sharper photographs. Cleaner edits. Perfect grammar. Perfect lighting. Perfect brand voice.

Generative AI is becoming extraordinarily good at polish.

Which means polish itself is becoming a less useful signal.

If everyone can produce a pristine image, a beautifully formatted article and perfectly competent marketing copy in seconds, those qualities tell us less about what went into making them.

Meanwhile, things we once edited out may gain value:

A strangely specific story.

A photograph that isn’t perfect.

A failed attempt.

A source link.

An unpopular opinion.

An admission that something isn’t known.

Changing your mind publicly.

Showing how something was made.

Not because imperfection automatically equals truth. Humans are perfectly capable of producing handmade, artisanal nonsense.

But specificity, process and accountability provide information that generic perfection does not.

The better AI gets at looking human, the less “looking human” proves anything.

So we may start looking for different signals.

Not does this look real?

But can I understand where this came from?

06 / Transparency

Even an “AI-generated” label doesn’t solve it

Transparency helps. But even that turns out to be more complicated than attaching a tiny robot icon.

A peer-reviewed study in the International Journal of Advertising tested responses to AI-generated charitable advertisements.

In one experiment with 226 U.S. participants, people shown an AI disclosure rated the advertisement’s credibility lower than those who weren’t shown the disclosure. In another experiment, participants exposed to an AI disclosure pledged an average $1.20, compared with $1.61 following a human-creation disclosure.

Simply announcing “AI was used” can therefore introduce its own trust penalty.

But newer research suggests how we disclose AI matters.

A 2026 Psychology & Marketing study found that people responded more positively when disclosures clearly explained AI’s actual role in producing an advertisement rather than merely giving a vague AI label. Greater perceived transparency increased confidence in the creation process, which improved attitudes toward the ad.

That makes intuitive sense.

“Made with AI” tells me almost nothing.

Did AI correct three sentences?

Generate the photograph?

Invent the person in the photograph?

Research the claims?

Write every word?

Those are completely different relationships between a human and a machine.

So perhaps the future of transparency isn’t simply declaring whether AI was involved.

It’s explaining what it did.

07 / Provenance

The internet is already trying to build receipts

Interestingly, this idea is moving beyond academic debate.

The Coalition for Content Provenance and Authenticity—C2PA—is developing an open technical standard called Content Credentials.

Rather than trying to magically detect whether an image “looks AI,” Content Credentials are designed to attach verifiable provenance information: where digital media came from, how it was created and what happened to it as it was edited.

The standard describes provenance as part of restoring transparency and trust around digital media.

That’s a significant shift in philosophy.

Instead of endlessly asking:

Can we detect the fake?

we begin asking:

Can the real thing prove its history?

We may eventually need some version of that idea far beyond photographs. For articles. Reviews. Research. Advertisements. Expert recommendations. Perhaps even AI answers themselves.

08 / A mental model

Three questions for the post-AI internet

Most of us aren’t going to inspect cryptographic metadata before liking a TikTok.

Nor should we have to.

But there is a simpler mental model worth keeping. When something online actually matters, ask three questions.

That final question may become particularly valuable.

Because the real premium in an AI-heavy internet isn’t necessarily going to be:

“A human made this.”

It may be:

“A human is accountable for this.”

09 / Scarcity

Content became abundant. Trust didn’t.

Generative AI isn’t going away, and attempting to divide the internet neatly into pure human creation and contaminated machine creation is probably futile.

AI will sit inside cameras, editors, browsers, search engines, phones, design tools and writing software.

Much of the best work will almost certainly involve both humans and machines.

The more useful distinction is whether technology is being used to extend someone’s capabilities or to manufacture the appearance that someone cared when nobody did.

For years, success online often meant producing more. More articles. More posts. More videos. More search results.

AI has effectively solved the supply problem.

We can now make more content than any human could possibly consume.

Which changes what becomes scarce.

Experience becomes scarce.

Judgment becomes scarce.

Original reporting becomes scarce.

Accountability becomes scarce.

And above all:

trust becomes scarce.

AI made content cheap.

The opportunity now belongs to the people, creators, publications and companies that can make themselves worth believing.

That may turn out to be the real human premium.