There’s a very interesting collision happening in search right now, and I don’t think we’re talking enough about what comes next. On 2 August, new transparency obligations under Article 50 of the EU AI Act came into effect. Generative AI providers must now make certain AI-generated and manipulated outputs detectable through machine-readable marking.
Anthropic has responded by introducing invisible watermarking into Claude-generated text, while other major AI players are developing their own approaches to provenance, labelling and detection. The technology isn’t foolproof. A watermark doesn’t conclusively prove who wrote something, how much AI was involved or whether a human substantially edited it. But I think focusing on whether the technology is perfect misses the much bigger shift. We are building an internet where the provenance of content increasingly matters.
And at exactly the same time, Google is making some very interesting moves of its own.
AI has made producing content extraordinarily easy. It has also made producing plausible-sounding content extraordinarily easy, and that distinction matters.
Google is no longer just ranking ten blue links and leaving us to decide which one to trust. AI Overviews and AI Mode synthesise information from multiple sources and present an answer directly within Search. That puts Google much closer to the information itself. If an AI-generated answer is built from generic, inaccurate or unaccountable AI-generated content, Google doesn’t just have a content quality problem. It has a trust problem.
So look at what Google has been rolling out. Preferred Sources now extends into AI Overviews and AI Mode, allowing users to nominate the websites and creators they want to hear from. Any website publishing fresh content can be eligible, and Google says people are twice as likely to click through to a site after selecting it as a Preferred Source.
Google has also introduced more prominent ways to surface fresh perspectives from discussions, forums and social media, alongside “Highly Cited” labels designed to identify influential original reporting. The language Google itself uses is worth paying attention to: original content, creator insights, unique perspectives, trusted sources and fresh perspectives. I don’t think that direction is accidental.
I don’t mean a literal ranking factor called HUMAN that somebody at Google switches on. I mean a trust signal built from experience, named authorship, firsthand knowledge, original research, expert opinion, accountability and a point of view that exists because somebody actually knows something about the subject.
We've been talking about these through E-E-A-T (experience, expertise, authority and trust) for years, but AI changes their value. When competent, technically correct, nicely structured content can be generated in seconds, competence is no longer much of a differentiator.
The question becomes what sits behind it. Who knows this? Who experienced it? Where did the information come from? What does this business know that 10,000 other websites regurgitating the same information don’t? And, increasingly, can Google verify any of it?
There has been an understandable assumption that generative AI would reduce the value of human-written content. I think we could see the opposite happen to genuinely expert content.
Not because Google has declared war on AI. It hasn’t. And not because every sentence needs to be lovingly typed by a human while AI sits untouched in another browser tab. That’s ridiculous, and it isn’t how I work either.
It’s because AI has dramatically increased the supply of information without equally increasing the supply of experience, expertise or original thought. Scarcity creates value, and evidence already shows that searchers care.
A YouGov survey conducted for Pangram in April 2026 found 69% of people who consume online content trust AI-generated content less than human-generated content, while only 8% said they trusted it more. Even more tellingly, 61% said they were unlikely to read or engage with content if they thought it was AI-generated.
That matters to search because Google’s job is not to reward businesses for publishing content. Its job is to give searchers information they find useful and trustworthy. If searchers become more conscious of who or what is behind the information they consume, provenance starts becoming commercially important to Google too.
"We’ve spent the last few years talking about how AI will change content creation. I think the next conversation will be about how AI changes content selection."
Renee Mariette
As more generic informational searches are answered directly by AI, there is less reason to click another generic article telling you the same thing. The click has to offer something more: the specialist’s interpretation, the business with 20 years of experience, the practitioner explaining what they see every day, original research, a case study, an unpopular opinion or a first-hand account. The person prepared to put their name behind the answer.
That changes what businesses should be creating for search. Publishing 50 generic AI blogs because “Google likes fresh content” is not a search strategy. Producing more information that already exists everywhere else becomes less valuable when machines can produce and summarise that information almost infinitely. Building a body of attributable expertise is a very different proposition.
Regulators are building infrastructure to identify AI-generated content. AI companies are implementing technology to make provenance detectable. Google is giving greater prominence to trusted sources, original reporting, fresh content and firsthand perspectives. Consumers are telling researchers they trust human-generated content more.
I don't think these are four unrelated trends. They are four responses to the same problem: when content becomes infinite, trust becomes scarce. That’s why I think the next era of search will place significantly more value on proving there is something behind the content worth finding. A person with experience and expertise. Original thinking. Firsthand knowledge. Something that existed before the prompt did.
AI isn't killing content. It's killing the value of content that has nothing behind it. And if I’m right, “human” is about to become one of the most valuable signals in search.
See what these changes mean for your search strategy and website content.
