Growing referral traffic from AI is nowhere near making up for losses from search since the arrival of Google’s AI Overviews and AI Mode, leading publishers have said.
Speaking at Press Gazette’s Future of Media Technology Conference in London last week, representatives from the Daily Mail, Bauer Media, PA Media and Admiral – The Visitor Relationship Management Company agreed there is no meaningful value exchange from AI bot traffic to websites and that publishers need to bolster themselves via the use of human experts and original content.
Daily Mail director of SEO and editorial e-commerce Carly Steven said that although the brand is “seeing the impact” of clickthroughs from search dropping dramatically when an AI Overview is present alongside its keywords, there is less of an impact on traffic overall.
Earlier this year Steven revealed that when an AI Overview appears in Google, the Daily Mail’s average clickthrough rate was 56.1% lower on desktop and 48.2% lower on mobile.
But last week she warned against focusing on the “big, scary numbers” as a hit to clickthroughs does not always mean an overall slump in traffic.
“What we’re seeing is that, by the nature of being a big breaking news site, even if there’s an AI Overview there, if by the time it appears that story has moved on and our readers aren’t searching for it anymore, actually the impact is maybe negligible.
“That doesn’t mean to say that it doesn’t matter. It absolutely does. But I think we are in this very fortunate position… where over 50% of all of our traffic is direct, over 60% of our Google search traffic is branded [searches containing the brand name Daily Mail], so it makes us quite resilient in the face of these changes.”
Bauer building business on ‘genuine, unique insights’ from experts
Bauer global audience director Stuart Forrest said the brands he oversees are facing a “nuanced” impact from the changes in traffic referrals.
With brands across verticals like automotive, TV and entertainment, women’s lifestyle, golf, angling and motorcycling, he said they see less branded traffic than the Daily Mail “and so perhaps AI Overviews is going to disrupt a bit more”.
“Definitely we’ve seen an impact on some of our autos titles where, if you’re talking about, say, car specification data that is not necessarily proprietary to us, absolutely, but we don’t build a business that is contingent on providing answers to those questions for consumers. We build a business that’s based on genuine, unique insights from people who are expert in their field, and add something to that data.”
Forrest noted that zero-click searches were already increasing even before the arrival of Google AI Overviews as the search giant began providing boxed-out information for some factual queries – for example sports results or Oscar winners – above website links.
He said: “Zero-click search features definitely appear more, and they did pre-AIO, on information that is perhaps more homogenous and not open for debate and a clear fact.”
Forrest went on to say that new information sources like ChatGPT appear to be “mostly additive to Google search, not necessarily a replacement to Google search”.
He noted that total search volumes do not appear to be dropping, adding: “LLM traffic is alongside search, so I don’t see that as a material threat, but I definitely agree with Carly that the stuff that Google is doing to reduce the need for consumers to click to our sites is absolutely an existential problem.”
‘Actual magnitude’ of AI referrals ‘still extremely small’
Dan Rua, chief executive at Admiral which serves a network of thousands of publishers including NBC, CBS and Hearst collectively approaching 20 billion page views, said the overall network is not seeing drops in overall Google search traffic.
But he said clear winners and losers could be seen when broken out by verticals. For example, he highlighted that law, business and finance, and politics were among the verticals that appeared to be seeing the biggest declines in terms of their percentage of referral traffic coming from Google between January 2024 and mid-2025.
Meanwhile pop culture, games, trivia, quizzes, religion and video gaming were the types of websites “not getting hurt or maybe even doing a little bit better”, he said. The full breakdown can be seen via a chart at this link.
Rua also showed what looks like steep referral traffic growth to the Admiral network from AI platforms between the start of 2024 to the middle of 2025 but noted that “the actual magnitude is still extremely small” in terms of actual traffic arriving. In May, AI sites were still only bringing in 0.033% of total referrals, up from 0.003% in January 2024.
Rua said: “The problem is, when you look at the negatives lost from search, the positives, even on this trajectory, do not end up making up for the loss – that 0.03% can multiply a bunch of times before it ever gets to the search losses.”
Bauer’s Forrest echoed this idea and suggested growth in AI referrals is levelling off.
“Not only is that referral traffic tiny, and we all know that there is really no meaningful value exchange from a referral perspective from these platforms, it also looks like it’s plateauing.
“May, June, July, it was like 0.17%, 0.18%, 0.2%, whereas a year ago, it was 0.01%, so we’re all looking at this and thinking, well, what’s the mature position? Certainly based on the past quarter, we may have plateaued… and that’s a real challenge, because there is no value exchange for us here.”
Rua also demonstrated that the vast majority of AI referral traffic to Admiral’s network is coming from ChatGPT, followed by Perplexity, Google’s Gemini and Microsoft’s Copilot. (A version of this slide with ChatGPT taken out is also available.)
Bot traffic is cost with ‘no meaningful value exchange’
Forrest noted that this growth is “massively expanding total bot activity, which is a net cost to us as publishers”.
“Cloudflare’s step to shut down bots on behalf of its publisher customers really is a sign that the CDNs [content delivery networks] are seeing the operational cost of all of this scraping that’s coming to us as customers at some point. And I think that’s the real problem that we face, that not only is there no meaningful value exchange from traffic, while as publishers we rely on a referral model, it’s just a net cost.
“So if you look at the sheer volume of bots, bot traffic is now one in 50 web visits. At the start of this year, it was one in 200 web visits. So that’s the real challenge that I think we as an industry need to start talking about, and what our response should be to that.”
Steven of the Mail agreed that “the value exchange of that bot traffic eventually turning into a real human being if we appeared in ChatGPT is so small that we just don’t see the value in allowing the bots access to the site”.
The Daily Mail therefore blocks chatbots and LLM scrapers both in robots.txt signals and at a website level via providers like Cloudflare.
[Read more: UK and US publishers back move to block AI scrapers by default]
Steven said this lack of a value exchange in the new AI landscape was also why the Mail had signed a licensing deal with, and taken an equity share in, Prorata which says it will share advertising revenue with publishers when their content is used in chatbot answers.
“I think everybody feels like the value exchange has just completely changed,” Steven said. “So there used to be this fair level playing field with Google and publishers that crawl their content, and we get a click, and we make money from the advertising we generate from that traffic. And that’s obviously gone.
“And I think with Prorata, the reason that we’ve partnered with them is because they absolutely understand the value of journalism, and that AI only thrives if there is high-quality, original journalism, properly fact-checked on which to train those models. And without that, the whole thing kind of falls apart and the only way that we can continue to provide that is if we have the revenue in order to do that.”
Rua advised publishers to block the bots, if they have not already: “That doesn’t mean in the future you might not find more creative ways to work in the whole ecosystem, but you have to start with the block.”
However Forrest advised that blocking robots.txt is “really nothing more than a statement of intent where bots are concerned, because data released just a couple of days ago suggests that 15% of all scrapes are mimicking human behaviour”.
[Read more: AI bots bombard publisher websites with ‘no meaningful value exchange’]
Therefore, he said, if websites block the bots then they will “start mimicking humans to crawl our content”.
“We’re into a whole world of problems there, because they typically mimic as a Chrome mobile user, so they come to our sites and they hit our content. They continue to cost us money hitting our content.
“But it also opens up the possibility that Google, who use engagement data to directly inform their organic search rankings, start seeing a rise in bad mobile user behaviour, so we start seeing the potential of bots impacting negatively our organic search rankings.”
He suggested instead that publishers should raise the price of scraping to a controlled level “such that the most palatable option is to come to us and have a value exchange with us. But for that to happen, I think you need some regulation.
“So for me, the big one for the bots is to mandate authentic open identification. It should be illegal for bots to mimic humans, and we should have the tools to be able to regulate that.”
Call to regulators: Ensure original content is incentivised
Forrest also called for the industry to work together better instead of being tempted into silos by deals: “The history of publishing is littered with much bigger players coming in and picking off one or two people with some money as soon as they get a sniff of regulation and that has been very successful in dividing us as a sector.
“If, actually, we’re going to act as one to stop this activity and protect our IP we have to be able to come together and have some kind of common approach and common standards.”
Steven shared three of the biggest points on her wishlist from Google:
- For Google to share the data in publisher analytics that would show the impact from AI Overviews and AI Mode on traffic
- More transparency around algorithm updates: “With a bit more of transparency from Google, everybody wins. I think Google gets the sort of high quality content that it wants, publishers understand where there may be problematic areas of their website, and we can rectify them. And then, most importantly, users get a bit better experience, which is what everybody wants”
- Better attribution of how publisher content is signposted when it is used in AI Overviews and AI Mode, and whether it is used for training or as an actual link.
Martin Ashplant, product development and operations director at PA Media, called for regulators to make sure “that the environment is set up to incentivise creating original content, because the worst case end point is we get to a place where there is no value in creating original content, because everything is being summarised, and then you lose the value of that, because you’d end up having AI kind of summarising itself, and that cycle will go into the middle, and you’ll end up in a really, really bad place.”
Ashplant said that although there was a “negative mood” in the day’s conversation about AI, he was “actually really bullish about AI” because of its potential.
He cited the use of AI for PA Media being able to power a better search function for its journalists and its publisher customers to access its archive back to 1991, and to convert content into different formats such as wire stories being turned into audio.
Ashplant added that “we will start to see a premium placed on human generated content… I think we’re going to see increasing amounts of AI slop, and it’s going to become harder and harder to wade through that slop.
“I think as creators of good quality, original content with humans involved, that will become a real premium. That’ll become a point of difference. And I think that’ll play out even more over the next few years.”
Email pged@pressgazette.co.uk to point out mistakes, provide story tips or send in a letter for publication on our "Letters Page" blog
