Reflecting on the major WAN-IFRA World News Media Congress in France last week, it would be great to pinpoint a single dramatic moment which somehow changes the trajectory of news and journalism in an AI world.
There wasn’t, but who would expect that from within our industry anyway? Part of the issue is that people outwith any of our orbits now make the decisions about how journalism fits into the world, and they don’t come to our events, they go to balls and summits and forums and have large retinues of staff traipsing behind on their phones.
If they were in Marseille, they may have spotted a change of sorts though, which was a growing realisation and agreement from within the media industry itself that the future of news does not really rely on the open web any more, and the industry should take charge of setting the terms of engagement with the outside world and the AI companies themselves.
Basically, we’re up for a fight, and we need to be - audiences patterns are changing fast, and AI costs are about to soar.
A worn out mule
The open web is not dead by any means, but as a discovery and distribution channel for journalism, ye olde free webbe now seems about as critical to news as the croaking town cryer who was barely able to shout louder than the newspaper vendors on either side.
The connected dots in Marseille led to A.G. Sulzberger, chairman and publisher of The New York Times, a panel discussion called “How AI is transforming the news experience”, and the announcement of some 30 new publishers and organisations from around the world joining the UK-led “Standards for Publisher Usage Rights” SPUR coalition, who are trying to establish standards for AI companies using news content.
Sulzberger grabbed the headlines with his fiery assertions that AI firms have robbed us all blind, and only a collective industry action will do anything to correct it. There is Dot Number One.
Later in the event, a panel session would echo many of Sulzberger’s points, surmising the era of readers “seeking out” news is over, when they don’t navigate directly to publishers en masse and rely more than ever on search and, now, AI.
People typing media brand names into a phone or PC is not a business model by itself, unless they already know you and trust you, which is central to what publishers now own and are crucial to in an AI de-mediated world.
This owned audience view of the world is critical to this overall realisation. Click-based revenues reliant on traffic and the crossed-fingers of Google’s goodwill just don’t stack up. The panel wasn’t in some new stage of mourning however, and what mattered was its focus on what comes next: Dot Two.
Setting our terms before “they” do
Dot Three came in the SPUR collective announcement, showing a rapid rise in onboarded media brands and orgs, now to a few dozen. In the scheme of things, this is tiny, but they are highly influential and have formed a pattern and set a standard that can be tagged on to by others.
Rather than just shout, they are acting, setting standards for a whole industry, bluntly because if publishers don’t establish that framework themselves they will be picked off one by one by AI firms which want to silence a few well-known and influential brands, and leave the rest to burn.
This is what happened in music and movies in previous scenarios: by acting early and laying out a plan for the buyers to see before any talks took place, it put rights owners in much more control down the line.
SPUR has no teeth yet, but wants to create the norms which other brands worldwide in their thousands will refer to in the future when dealing with AI.
The open web did great things for the world and for journalism and media firms, while slowly ceding control over both to tech firms.
The short AI era has already made people realise this time around that what previously took 20 years might happen in two.
Action is not singular. Join SPUR, build your audiences around people you know, and make it so your content is able to be picked up by the new digital audiences of agents and robots.
It’s a fight, but at least we know the terms of engagement now.
Google's ad monopoly bill keeps rising
More than 20 European news publishers from eight countries are seeking £550m in damages from Google for adtech monopoly abuse. This move follows last year's European Commission fine of €2.95bn, which found Google favoured its own ad tech exchange by sharing competitor bid data and routing its buying tools to avoid rival exchanges. Publishers argue revenue was depressed and fees were higher than they should be. The case is funded by litigation funder LitFin, which will cover costs even if the suit fails, opening the door for smaller publishers who couldn't take on Google alone. It's not an isolated case, but joins a growing pile with cases such as 32 European media groups suing Google for €2.3bn in 2024, and the US DOJ proving Google monopolised digital ad markets last year.
Read
Visits lie, subscriptions don't
A recent study analysed 1.2 billion user sessions and 600 million article visits to an anonymous US metropolitan daily title and concluded that hard news about local government, public health, and politics is far more likely to switch a reader into a subscriber than entertainment or sport. However this isn't quite the panacea it seems to those who major in such news: newsrooms sustaining civic journalism only through digital subs is on shaky ground in even the most optimistic scenarios, when the digital subscriptions one reporter can generate may only cover about a quarter of their salary. Nieman Lab looks through the full dataset.
Read
Agent-ready with Glide
Publishers are already using AI to generate copy and summarise information, but what they need more now are agents that can work across systems: finding material, understanding context, and carrying out multi-step tasks within editorial rules. In order for that to work, the platform sitting underneath has to expose structured content, events, permissions, and publishing operations in ways an agent can understand - otherwise the new audiences of AI won't be able to work with what you have. Glide CMS has typed SDKs and MCP servers above its APIs, so agents see and access what they need to while editors still have control over the workflow, review, and approval.
Read
Chase loyalty, not clicks
With all the AI noise around, Google rolled out tools aimed at audience retention rather than traffic recovery. While Preferred Sources, Search Profiles, and Subscription Linking sound like helpful tools, it's important to remember that none of these replace lost traffic. These tools reward publishers who already have loyal readers, once again reminding everyone that cheap clicks aren't the important part, building something people come back for is. Barry Adams draws a new map.
Read
The free cake buffet no one agreed to
Europe's copyright directive turns five years old, and 250,000 authors across 52 associations want it gutted. The European Writers' Council claims that the directive's text and data mining exception, agreed upon in 2019, has been stretched into a free-for-all grabfest for non-EU tech firms. Creators technically have an opt-out, but even if they do there is no way to confirm AI developers acknowledged it, while the transparency rules only ask for the top 10% of domain names scraped, not individual works. The EWC is calling on the Commission to remove the TDM exception entirely, and replace it with voluntary opt-in licensing.
Read
The crawl stops, but data stays
Digital Content Next has sent a cease and desist letter to Common Crawl, the foundation whose free web archive was used as 60% of GPT-3's training data, demanding it stops collecting publisher content, and remove material already stored. The core argument: copyright law is not an opt-out regime, and publishers shouldn't have to ask to be removed. Blocking CCBot stops future crawling but that still won't remove the years of content already sitting in downloadable databases. Around 79% of top sites already block at least one training bot, but DCN's question is what does that accomplish if the material is still available in the archives.
Read
Zero-click acceleration
In the first four months of 2026, over two-thirds of Google searches ended without anyone clicking anything. That's up from 60% in 2024 and 49% in 2019, the fastest acceleration of zero-click in a decade. The culprit? AI Overviews, which reduce CTR by nearly 60%. Google's financial performance has improved during all these changes, so don't expect any changes or a reversal while that remains the case. One way for publishers to deal with this is to stop treating traffic as the primary KPI and start building brand influence, because even as clicks disappear, the content will still shape what AI answers say about you.
Read
AIs and their media diets
According to new Muck Rack data, across 25 million AI citations, earned media falls under 84% of what ChatGPT, Claude, and Gemini tell people. In the UK, the Guardian shows up in around 45% of AI answers, while paywalled sources barely even feature. Each model has its own diet: ChatGPT leans on Reddit and Wikipedia, Perplexity goes for the fewest sources and favours YouTube, while Claude and Gemini prefer primary and financial sources.
Read
Hidden SEO goldmine
While server logs capture every request search engines make, most organisations never take time to analyse them. Unlike Search Console and third-party crawlers, logs show exactly where the crawl budget gets wasted, how the servers respond under real crawl load, as well as which pages search engines actually prioritise. The data is already there, but not much use unless retained, analysed, and used.
Read
Parental controls, political pressure
Big Tech can move when it wants to, it appears. Apple recently previewed new parental controls, on the same day UK Prime Minister Keir Starmer gave both Apple and Google a window to implement device-level child-safety protections. The new features include Ask to Browse, category-wide time limits, as well as an expanded Communication Safety which will block content unfit for children. Whether this will satisfy the demands is yet to be seen, as for some of these Apple simply warns and doesn't block. In the US, the Kids Online Safety Act has already cleared committee in March, and billions in settlements and verdicts are already likely.
Read
Breach in plain sight
Microsoft has disabled more than 70 of their own GitHub repositories after hackers planted malware to harvest credentials from tools such as Claude Code, Gemini CLI, or Cursor. The same group previously compromised Microsoft's durabletask package from May, publishing three malicious versions. While Microsoft says they've notified users and are investigating, if you're using any of these services do check your exposure.
Read
Error or early preview?
A flag in Chrome Binary briefly confirmed fears that all Google search queries will be AI Mode by default, though Google's VP of Engineering for Search hurried to say it was an error and that there is no plan to make AI the default. Believable?
Read
No more WhatsApp gatekeeping
The EU Commission has ordered Meta to give competing AI assistants free access to WhatsApp, reverting to rules applicable prior to October 2025, when it began restricting and then charging rivals to use the messaging service's business tools. Italy and Brazil have pursued near-identical cases, and Meta will most likely appeal those just the way it will appeal this one. As the broader antitrust investigation continues, Meta is facing fines up to 10% of annual revenue for ignoring the order.
Read
Output up, outcomes down
iOS app releases have nearly doubled since agentic AI coding tools emerged, and 51% of all code committed to GitHub is either AI-generated or substantially assisted. However, the number of apps with significant usage hasn't moved much. A study of 100,000 GitHub projects, titled "Agentic Much? Adoption of Coding Agents on GitHub", found that while AI coding tools increased activity by 180%, by the time those gains reached actual releases most failed to make any sort of dent. There are more apps being built than ever, yet no sign any of them are gaining traction.
Read
Mythos lite goes public
Remember when big bad Mythos was too dangerous to release to everyone, but safe enough to talk about prior to IPO filings? Anthropic has released a sort of Mythos lite, Claude Fable 5, a "Mythos-class model made safe for general use". It is basically a public version of Mythos which exceeds every Anthropic model previously made. The catch? It won't tackle or answer anything sensitive like bio-weapons - good - but also costs twice as much as Opus 4.8. Considering the growing AI cost concerns spreading around the industry, that is not going to age well. The full Mythos still remains under the Project Glasswing safety initiative, for now.
Read
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.