The walled gardens gets taller, a new revenue stream no one expected, and Google trims the guest list

Two questions for every publisher, a domino case for Meta, Brussels is done with games, and mythical Mythos inches closer - all in this week's Content Aware.

Published: 15:17, 28 May 2026

Corbidge comments on… the faked future of truth
Oh the irony. A book about the effects of artificial intelligence on truth, The Future of Truth, by tech-and-media thinker Steven Rosenbaum has been found to, err... be pockmarked with AI-fabricated quotes attributed to real people. These include noted journalist Kara Swisher who confirmed that she, in fact, never said what she was claimed to have said. While Rosenbaum acknowledged the use of AI, it still doesn't make this situation any less sloppy or, frankly, satirical. Our resident Quote Cop Rob shares that if you can't be a good example, at least be a terrible warning.
Read

Publishing & Media

Google's garden getting higher
Google's latest Search update will allow users to jump from AI Overviews to AI mode automatically, with just a follow-up question, creating what an SEO manager would call "a zero-click loop" locking users into AI-generated answers. While AI Mode isn't the default yet, SEO experts agree that as soon as Google figures out how to make it match its current ad revenue from organic search, it will become so. Some of the new features include AI agents which track products and content in the background, interactive visuals, and AI-powered query suggestions before you've finished typing. Google claims it still sends billions of clicks to the web daily, and for now, breaking news is relatively safe. However, for publishers that rely on top-of-funnel search traffic, this is one more nail. Press Gazette knows more.
Read

Fourth leg of the chair
While local journalism's sustainability conversation has long been centred on ads, subscriptions, and philanthropy, now there is a fourth player in the game: public policy. According to Rebuild Local News' annual report, $129 million is flowing into local news through state-level action, with an estimated $74 million going out to newsrooms in 2026 alone from laws enacted in six US states. Small outlets aren't being left out: in Illinois, two-thirds of employment tax credits went to titles with staff of six people or fewer. The models gaining traction include employment credits, government requirements, and state-funded fellowships. The goal isn't favouring one outlet, but lifting an entire local news system.
Read

Blog for buyers, not clicks
Most blogs attract readers but not buyers, which kind of defeats the purpose for most sites. The problem is usually strategic: topics chosen for search volume rather than purchase intent, calls to action that don't match a reader's place in a journey, problems disconnected from solutions, and articles that sit in isolation rather than building topic authority. Add in a missing bottom-of-funnel layer and no clear conversion path, and traffic stays traffic. The fix is building content around buyer intent, not just clicks. Steven Wilson-Beales, who will be one of the speakers at Glide Live: London in June, writes more.
Read

The cognitive tax on the truth
If a dictator of the internet were to take away your ability to get accurate information, the result would look similar to what we can find on our screens today. That's how Matt Pearce started his essay, using behavioural economics to explain a news industry decline, leading with the argument that the dollar cost of encountering content has declined thanks to ad-supported platforms and subsidised AI agents, of which everyone is feeling the impact. Hardest hit are the smallest outlets who relied on search referrals. The result? An invisible tax levied on communities that we pay civically, cognitively, and sometimes literally.
Read

Teach AI your editorial history, not just words
Tipped all your content into an AI and hoped it would work miracles in categorising it all? How's that going? AI tagging without context is a guessing game, and while semantic search can spot what an article is about, editorial tagging IRL has a completely different requirement: applying the right terms based on how your publication classifies its content and what your view of the world is. If you are bogged down by vector-only methods, or flooded with suggestions that are plain wrong, here is how we do it. Glide CMS uses a multi-signal approach that combines semantic understanding, your publication's taxonomy structure, and even your editors' own tagging history, to suggest terms that actually fit how your site classifies content. Tested against 1,000 real articles, it found 97% of editor-applied tags versus 20% for semantic-only. The result is way fewer gaps, much less correction work, and tag suggestions that reflect your editorial conventions from day one.
Read

Reaping what you didn't sow
A Reddit user posted about building an AI tool that could scrape local news and generate podcast audio from it; minutes after a student newspaper published a story, an unrelated AI podcast released an audio version drawn from that same story. No permission and no payment involved. When your content can be so rapidly turned around by others, what could be the remedy to acknowledge your work? A century-old Supreme Court doctrine "hot news" could become the legal fix, as it protects labour and resources behind news production, on unfair competition grounds rather than copyright. While courts have rejected its use in recent cases - but never overturned the law - the recent 2025 ruling against the New York Times noted that AI scraping turned into competing slop could still qualify. Nieman Reports explores if it's time for a revival.
Read

Two questions every publisher should answer
WNYC's Head of Product, Sam Guzik, laid out a simple two-question framework at WAN-IFRA's Digital Media Asia conference for publishers to help guide discussions over strategy and future: is your mission civic infrastructure or community, and what are you best at - breaking news or explaining things? His framework reveals where to compete against the commoditisation of information, doubling down on content, sources, and the storytelling that humans do best, or leaning into authenticity and community where AI struggles. The upshot? News is just too important to assume it'll be okay without deliberately choosing where to focus. Elsewhere at WAN-IFRA, your favourite neighbourhood CMS Glide will be attending the 77th World News Media Congress in Marseille, France, so do come and say bonjour to the Glide team at booth 12.
Read

Google gets pickier
SEOs are reporting that Google seems to be de-indexing URLs at a higher than usual rate since early April. Some suspect it's a freshness or quality evaluation tied to reducing index size, while others believe it's just a reporting bug. With the volume of AI content online, the consensus among site owners is that Google is being more selective about what gets a spot in the index. Search Engine Roundtable has more details.
Read

Big Tech

The internet you knew is gone
Axios reports that Google's overhaul of the search bar washes away one of the last monuments of the early internet era, when social media was novel and search felt empowering and good. Blue links are now a secondary offering while zero-click answers are in the spotlight now, and the companies that build the web are now competing to see who will dismantle it first. Trust in Big Tech fell from the top in 2010 to just 24% in 2025, and the pattern is almost everywhere: platforms are upending products which once generated billions before they've even figured out what the new business model looks like, or even exists.
Read

Meta's apps on trial
A New Mexico judge appears ready to rule that Meta's social media apps amount to a "public nuisance", a determination which could force fundamental changes to both Instagram and WhatsApp. The judge is considering the requirement of an independent safety monitor, pausing notifications for young users during school hours, as well as stronger policies on harmful content involving children. This is coming after a jury verdict from March, where Meta was ordered to pay $375 million in penalties for violating child safety laws. Separately, Meta has settled a bellwether Kentucky case, ending the first school-district trial over youth mental health. With thousands of lawsuits pending from families, school districts, and others from across the US, whatever happens in New Mexico will echo beyond state lines.
Read

Brussels sends the bill
The European Commission is preparing to bring down the DMA hammer on Google, with a fine in the high hundreds of millions of euros for self-preferencing its own shopping, travel, and local search services in ranking. It would be the highest penalty ever issued under the DMA. These changes, which came as a result of an investigation opened in 2024, will result in what Google has labelled as "the biggest downgrade in the product's history" and of course it plans to challenge any decisions in court. Meanwhile, a separate DMA case is looking into Google's treatment of news publishers in search results, for which we're still expecting a formal decision.
Read

AI & Copyright

Spotify's AI remix gamble
In today's episode of "Huh?" we have Spotify allowing premium users create AI-generated remixes and covers using music from participating artists and framing it as a "controlled" alternative to piracy and unregulated AI slop. Artists can consent and earn from it - is this the music industry's "I've got mine..." moment? - but that is not the underlying problem: every minute someone is listening to an AI knock-off, someone real is losing out.
Read

AI labels front and centre 
YouTube will now make AI-generated content labels more prominent and automatically apply them if the system detects "significant photorealistic AI use", whether the creators have disclosed them or not. The labels won't affect recommendations or monetisation, and creators have the option to appeal incorrect flags, but if the content is made with YouTube's own AI tools or carries C2PA metadata, the labels are staying.
Read

Is Mythos waking up soon?
Anthropic seems to be preparing a broader release of its HAL-9000-esque Claude Mythos under the label Mythos 1. During internal testing via Project Glasswing, the Mythos preview managed to find more than 10,000 high and critical-severity zero-day vulnerabilities and demonstrated the ability to generate professional-grade offensive code. Offensive in the sense of attacking an exploit, that is. While there is no official release date yet, Anthropic said the Mythos models will be generally available "in the near future" once stronger safeguards exist.
Read