AI Overviews Cost Publishers 40% of Clicks, La Gazette Readers Revolt Over AI Sub-Editors...
A shout out to this month’s sponsor » PressReader Editions+ transforms replica or PDF magazines and newspapers into modern digital experiences, with built-in text-to-speech, translation, interactive content, and white-label app solutions. A fully turnkey solution, it increases reach and revenue, improves accessibility, and strengthens engagement.
When it comes to audience measurement, Stephen Jones is as experienced as anyone in publishing. Previously the Head of Analytics at the Press Association, he became increasingly frustrated by the number of platforms he had to use to pull together a single view of content performance. His answer? He built his own.
I spoke to Stephen and Sophie Vokes-Dudgeon, Chief Content Officer at Hello! Magazine, about what happens when publishers can see what each story is worth across Apple News, MSN, NewsBreak, SmartNews and other channels. TL;DR: page-view growth of 33% upwards, with revenue growth to match.
Let’s jump in…
AI Overviews Represent a Straight Loss of Audience
Google insists that although AI Overviews reduce clicks, the clicks publishers lose are the ‘low quality’ ones. It’s simply not true. Research by Saharsh Agarwal (Indian School of Business) and Ananya Sen (Carnegie Mellon) also found a whopping 40% decrease in organic clicks when the summaries are shown. The publishers’ friend…
ChatGPT Commands 92% of AI Referral Traffic
Previsible’s latest AI traffic study, based on 6.7 million LLM sessions, finds that ChatGPT accounts for 92% of AI referral traffic. Monthly LLM sessions grew 9.9x between Nov 2024 and May 2026. BUT ChatGPT referrals fell by half in Nov 2025, from 448k to 213k, before recovering the following month due to a ‘model-related change’. A clear warning…#volatile #traffic
Advertising Is Now a Product People Pay to Avoid
Sir John Hegarty, one of the legends of British advertising, used Cannes to argue that the latest generation of marketers have no understanding of branding and are fixated on sales promotions. The result? People have stopped liking the work. Key quote: “I’ve never read a business book that says, to succeed, make a worse product.” The World Media Awards shortlist - just out - is an exemplary exception.
Readers Use Le Monde Op-Ed to Push Back Against AI Sub-Editors
In an extraordinary opinion piece, readers of ‘La Gazette des communes’ - a highly respected French B2B title - have used a full-page in France’s leading newsbrand to attack Infopro Digital’s plan to replace sub-editors with AI. Their argument? Human judgement is needed to check, contextualise, edit and maintain reader trust. Too right…
Philly Inquirer Turns Cancellation Requests Into Renewals
The Philadelphia Inquirer saw long-term retention exceed 75% by replacing generic special offers with a dedicated retention team consisting of human agents, phone and live chat. The publisher combined “deep listening” with specialised capability aimed at “delivering the best possible experience” for subscribers thinking of leaving. NB: One-size-fits-all programmes were dropped…
Open-Source Tool PxPipe Cuts Claude Fable 5 Token Costs Up to 70%
Big deal? Yes. For publishers developing AI tools, token optimisation is everything. In an AI group I belong to, one media exec ran an agent task overnight and ended up with a £2,500 bill. This workaround renders source code or text into PNG images, which are billed differently, so the total token usage falls significantly. P.S. Costs are only going much higher » https://isaiprofitable.com/
What It Really Takes to Build a Membership Programme
Around 48% of the events industry is either already operating or planning to launch a membership proposition within the next 12 months - yet many fail because publishers mistake an audience for a community. AMO’s Kari McMahon outlines exactly what it takes to create a membership community. Rule #1: Don’t use the community as an upsell machine…
Europe’s AI Compensation Framework Starts To Take Shape
Following the UK’s CMA pushing Google to let publishers opt out of AI Overviews without being punished in search rankings, Italy is using EU copyright rules to support collective negotiation over publisher compensation. If policymakers don’t set the terms, Silicon Valley will. Key quote: “Voluntary compliance remains the weakest link in the entire system.”
WAN-IFRA Appoints Ezra Eeman To Lead AI In Media
Known as one of the world’s leading voices on the use of AI in media, WAN-IFRA (World Association of News Publishers) has landed Ezra Eeman, who joins from Dutch public broadcaster NPO. He will lead WAN-IFRA’s AI strategy, as well as the development of new initiatives to help publishers accelerate AI adoption and transformation. A significant appointment…
AI Tool: Honeylog
Honeylog lets publishers see who is really consuming their content: human readers, search engines, AI crawlers, commercial bots, and scrapers. By analysing server logs, it shows how much machine traffic reaches their sites, which bots generate value, and which consume resources without giving anything back. Turns invisible technical data into strategic insights and AI negotiation leverage…
Webinar: The AI-Era Monetization Blueprint for Publishers
As AI systems increasingly rely on machine-readable content, publishers that can structure their content accordingly, and make their journalism easy to license, track and control will be better placed to benefit. Features media licensing luminary, Creative Licensing International’s Paul Gerbino, and others… Tues 14th July | 10am EDT | 3pm UK
Event: Data Science Day 2026
WAN-IFRA’s annual gathering for senior data scientists, analysts and CDOs in news media. Seven publisher case studies, two expert workshops, and speakers from News UK, The Times, Schibsted, Axel Springer and others. Early registrants get complimentary access to selected Digital Media Europe sessions the next day. 21 October | The Royal Institution, London
In partnership with Bridged, WNIP has created Media Genie - a smarter way to explore 17 years of publishing insight. Ask a question about the media business and get a direct answer grounded in WNIP’s archive. 👇
How Hello! Turned an Industry-Wide Google Traffic Drop Into an Off-Platform Audience Success
With Google referrals falling across the publishing industry, Hello! built a cross-platform data strategy designed to grow off-platform audience without handing editorial control to any single channel.
Hello! Magazine has been a fixture of British media since 1988, when it brought the celebrity journalism of its Spanish sister title, ¡Hola!, to UK newsstands. Nearly four decades later, the brand’s challenge is certainly now less about recognition and more about distribution, with readers encountering the brand across numerous different platforms.
Like many publishers, Hello! built a significant share of its digital audience through Google referrals. As search behaviour has evolved and AI-generated answers and algorithm changes have reduced referral traffic, particularly over the past year, the underlying strategic problem has been how to measure the performance of all its various channels and allocate resources accordingly.
Rather than pivot from one platform dependency to another, Hello!’s response has been to bring all its platform data into one daily view, so editors can see what works across Apple News, MSN, NewsBreak, SmartNews, and others, without letting any of these platforms dictate editorial commissioning.
Whilst this has been a difficult balance to achieve, the results over six months have been concrete: page view growth of between 33 and 102 percent across aggregator channels, depending on the platform, with revenue growth to match.
The risk of off-platform growth
The obvious response to falling Google referrals is to find new sources of external traffic. But publishers risk replicating the same structural problem: building audience on ground they do not control, and making commissioning decisions shaped by incentives that belong to someone else.
Sophie Vokes-Dudgeon, Chief Content Officer at Hello!, is clear about the lessons she has already learned. The question her team now asks is not what is driving traffic but whether a story makes sense for the brand.
“Once bitten, twice shy. Before, we’d ask ‘what’s driving traffic?’ Now the question is always ‘does this make sense for the brand?’ first.” — Sophie Vokes-Dudgeon, Chief Content Officer, Hello! Magazine
Strategically this is important because off-platform growth is only commercially useful if it serves the publisher’s own audience relationship. Hello!’s VIP membership programme also depends on readers who have chosen a direct relationship with the brand. Aggregator platforms are therefore now being looked at as routes into this relationship and not as a replacement for it.
The data challenge underpinning the strategy
The key problem Hello! encountered was obtaining a single bird’s eye overview of its off-platform distribution, and discovering what was actually happening across multiple platforms.
This is far harder than it sounds because although each platform produces analytics, the data sits in separate systems, updated at different intervals, and in different formats. In short, any meaningful analysis required days of manual work consolidating information that was not designed to sit together.
“When you’re writing for website traffic, you’re optimising against one metric. Suddenly you’re looking at page views across half a dozen platforms, each behaving differently from week to week.” — Sophie Vokes-Dudgeon
Hello! used Maro, a content analytics platform built by journalist and former Press Association analytics head Stephen Jones, to pull performance data from across its distribution footprint into a single daily view.
The system uses a traffic light scoring model, ranking every story as a top performer, mid-table, or content that is working nowhere, across each measured channel simultaneously.
Jones built Maro to address a problem he had encountered directly at PA Media, where large volumes of data from multiple sources did not translate into clearer editorial decision-making.
The gap was not in the quantity of analytics available but in whether editors could use them quickly enough to act on them. “Most of the analytics platforms out there are designed for an age where everything’s on your website,” he told me. “Nowadays, for publishers, that’s not the case.”
What cross-platform views reveal
The first data point a single unified view unearthed for Hello! was that a small number of stories were underperforming across every measured platform simultaneously. This had been cloaked when data lived in separate systems, but by identifying and reducing low-return stories, the team immediately redirected its commissioning effort.
The other significant finding was that Hello!’s content did not behave uniformly across channels. Royal coverage would perform strongly on one channel, while celebrity and lifestyle features, by contrast, delivered stronger returns on another.
Another learning was around tactics to ensure best results - differences in headline formats or publication times to optimise for best performance on each channel. All of these learnings led to growth, and would have been missed if only analysing their on-site numbers.
“We needed that data presented clearly enough that we could see, by writer and by platform, what was actually working.” — Sophie Vokes-Dudgeon
Six months in, top-performing content is up 12 percent and bottom-performing content is down 11 percent - as measured by Maro’s traffic-light system. Editorial and audience teams are now working from the same daily numbers, creating a shared sense of mission across a large and multi-functional editorial team.
“We’ve got more levers to pull now, which makes the whole approach feel a lot safer than relying on any single channel ever did.” — Sophie Vokes-Dudgeon
Bottom Line
Hello!’s experience shows that platform analytics alone no longer give editorial teams enough information to judge story value, because the same content can perform very differently across Apple News, MSN, NewsBreak, SmartNews and other channels.
The Google dependency problem is not solved by finding new platforms, it is solved by knowing what your content is worth across all of them and commissioning from that position.
Hat tip: Monojoy (Oddly Obvious), Jim Bilton, Stephen Jones, Sophie Vokes-Dudgeon, Lorenzo Diaco, Charlotte Panther, Dean Roper, Giselle Ho, Carvoeiro PT, Fantic Caballero 500.