Advertisers and publishers are being urged to look beyond whether content was created by AI and focus instead on whether it delivers genuine value, as new research reveals more than one-third of webpages published since ChatGPT launched show signs of AI authorship.
The findings, from Pew Research Center, have put fresh focus on the relationship between generative AI and the growing Made-for-Advertising (MFA) ecosystem, as the technology makes it dramatically cheaper and faster to produce content at scale.
Pew analysed a random sample of 10,000 English-language webpages collected in July 2026, finding that 10 per cent showed significant signs of AI authorship. But among pages published after ChatGPT was released in November 2022, the figure climbed to more than one-third.
For the advertising industry, the concern is not that AI-generated content is inherently MFA but rather, AI has removed much of the cost and time traditionally involved in creating the enormous volumes of low-value content that can sit underneath the MFA business model.
IAB Australia technology lead Jonas Jaanimagi told B&T the economics of MFA have changed dramatically since the organisation published its MFA guidance in June 2024.
“What has changed since we published our MFA guidance in June 2024 is a dramatic drop in the cost of feeding the underlying MFA business model,” he said. “Ad arbitrage has been around for many years and GenAI has increasingly removed any of the constraints on producing the content that arbitrage requires.”
However, Jaanimagi stressed that AI-written content and MFA are not the same thing.
“The industry needs to stop treating them as interchangeable,” he said.
IAB Australia’s MFA guidance defines MFA sites through characteristics including low-quality automated content, sometimes AI-generated, high numbers of ad units relative to content, low levels of organic traffic, clickbait headlines and reliance on paid traffic. The central issue for buyers is wasted ad spend, with users typically less engaged and advertisers receiving less value from their investment.
Jaanimagi said Australian publishers using AI responsibly for research or subediting should not be penalised simply because AI has been part of the production process.
“What defines MFA is the business model of buying cheap traffic, running very high ad loads, and monetising attention that is not really there,” he said.
“Verification providers are working to close this gap, but buying behaviours will also have to shift.”
While the Pew research does not provide a specific Australian breakout, Jaanimagi said the production tools and programmatic supply chain are global, meaning Australian advertisers are likely to face similar trends.
However, he said exposure would not be evenly distributed, with the anonymous long tail of the web presenting a particular concern.
The Pew analysis found AI-authorship signals were not evenly distributed across the web, with the research also showing substantially higher levels on .com pages than .org, .edu or .gov domains.
Jaanimagi said advertisers should therefore be moving towards supply-path optimisation, curated inventory and direct relationships with publishers they trust. He also urged buyers to use dynamic detection rather than relying on static blocklists and to judge campaigns on business outcomes rather than simply viewability and clicks.
“Static domain lists are certainly now a thing of the past as malevolent operators can quickly now spin up hundreds of near-identical sites from a single template overnight.”
AI text to outpace human text?
DoubleVerify’s head of sales ANZ Sophia Warren said the sheer volume of AI-generated content means advertisers cannot simply take an “AI = bad” approach.
“As generative tools improve and adoption accelerates, AI text will outpace human text within a year, with visuals likely following soon after,” Warren said.
That means advertisers need to distinguish between AI-generated content that is useful and appropriate for a brand, and low-quality content that offers little value to consumers.
DoubleVerify’s Global Insights: Media Quality in the Age of AI research found 59 per cent of Australian advertisers are concerned about advertising next to low-quality AI-generated content. The study surveyed 22,000 consumers across 22 markets and 2,020 marketers and advertisers across 21 markets.
The brand-safety implications are significant.
In Australia, 52 per cent of consumers said seeing a brand advertised alongside low-quality AI-generated content would negatively affect their perception of that brand, compared with just 16 per cent who said it would have a positive impact.
But Warren said the answer is not to block AI content altogether.
“The real question is no longer whether content is AI- or human-generated. It’s whether that content is suitable and aligned with a brand’s values and goals, regardless of how it was created.”
She said advertisers would benefit from developing tiered suitability guidelines that account for context, as AI content could be appropriate in environments such as gaming or entertainment while posing greater risks elsewhere.
Clearer signals are also needed to distinguish legitimate creative experimentation from content designed to mislead or manipulate.
Alison McKinnon, managing director at CM.OSX also spoke with B&T.
“AI hasn’t invented low value content, but it has fundamentally changed its economics, which now supports enormous volumes of disposable content designed primarily to capture search traffic and advertising impressions rather than provide genuine value to the reader,” she said.
“When the marginal cost of producing another article approaches zero, content that would never have been commercially viable suddenly can be.
“The challenge for advertisers is less about identifying AI content and more about distinguishing genuine value from content manufactured primarily to generate an impression.”
There is also a potential longer-term threat to the MFA model itself.
McKinnon said the growth of AI interfaces could reduce the number of consumers visiting generic articles in the first place, as people increasingly get answers directly from tools such as ChatGPT and Google AI Overviews.
“LLMs can remove the click altogether,” she said.
“If people increasingly get their answers directly from ChatGPT, Google AI Overviews or other AI interfaces, there is less reason to visit a generic article created primarily to capture search traffic.”
That could ultimately make the economics of producing endless pages less attractive, while increasing the value of original thinking, proprietary data, genuine expertise and authoritative content.
Meanwhile, marketing and public relations agency AZK Media said generative AI had removed “one of the biggest barriers to creating MFA sites” – the cost and time required to produce huge volumes of content.
“What once required a team of content writers can now be generated, repurposed and published at industrial scale in minutes,” managing director Azadeh Williams told B&T.
However, she cautioned against treating every AI-generated website as an MFA operation.
“There’s an enormous difference between a credible publisher, brand or expert using AI intelligently to support high-quality content and someone using it to manufacture thousands of low-value pages purely to attract advertising dollars.”
Williams believes the answer is to focus on a combination of provenance, quality, transparency and intent instead.
