Content teams fight back, comedy beats conspiracy, and simple things still win

The Mail+ rewrites the playbook, The Independent's ten-year proof, Meta's workforce turns to a dataset, and 75,000 fake songs a day - all in this week's Content Aware.

Published: 15:37, 23 April 2026

Corbidge comments on… three species of media
Francesco Marconi, who previously worked for AP and WSJ, has published a paper which argues that the media industry is splitting into three types: The Intelligence Business, the Attention Aggregator, and the Public Good. And only one of them has pricing power. Our Chief Scroll Officer Rob shares his thoughts on the paper and how it brings a different thinking to an industry which has been used to reacting to Big Tech instead of being ahead of it.
Read

Publishing & Media

Media owners agog as Google says content theft is harmless
Google has told the UK's Competition and Markets Authority that using publisher content to fine-tune its AI models is not harmful to publishers. The Guardian, Financial Times, DMG Media, and the News Media Association all pushed back on this, arguing that fine-tuning obviously adds value to Google and that at some point it might make the need for real-time content retrieval obsolete, undermining the licensing deals publishers are trying to build. Publishers are aiming at separate opt-out controls for training, fine-tuning, and retrieval, so they can decide what they want to permit. Currently, it's all one bundle, and DMG Media suspects that Google could separate fine-tuning from its ranking algorithms but commercially has no reason to do so. The CMA is also urged to not wait a year before addressing payment for content, because a delay will leave the existing power balance untouched.
Read

Satire takes over the conspiracy factory
After 17 months of battling it out in court, The Onion has secured a licensing deal for the Infowars brand, website, and IP. The deal still needs a judge's approval, but if it all goes through, America's most famous satirical news outlet will be running the platform used for 25 years by hyper-controversial Alex Jones. After a defamation lawsuit in 2022 brought Infowars to bankruptcy, the deal came to fruition and now The Onion is paying $81,000 a month for six months with an option to extend. At the helm as creative director, they've put the comedian Tim Heidecker. The Onion's own description says it all: "America's Finest News Source has entered into an agreement to operate America's Source of Disinformation for Sovereign Citizens Who Reject The Idea of Child Support".
Read

The Mail's silent reinvention
Around 60% of the Daily Mail's traffic is direct, and that's the stat the media giant is now betting on. Instead of page views, eyes are now on time spent, repeat visits, and a "golden metric" which blends dwell time with sharing, commenting, and saving. An AI-powered dynamic paywall has been tested on half the global audience for three months, gating content based on user behaviour rather than a fixed list. Games have gone from slideshow to front-of-app: the daily word game pulled 50,000 to 100,000 users in its first week, more than half returning daily. The Crime Desk hub bundles true crime stories, podcasts, and a daily crime game into one destination. Mail+ is targeting 1 million subscribers by 2028, up from 400,000 today.
Read

No print, no problem
When The Independent decided to drop print in 2016, revenue fell 73% and nearly half of the staff moved on. Fast forward 10 years, and digital revenues are now higher than print, The Independent on Sunday, and The i Paper combined. The title has been profitable every year since then, with two thirds of its 62 million monthly visitors coming from outside the UK. The model that replaced print is built on global reach, multiple revenue streams, first-party data, podcasts, a content operation that spans six languages, and licensing of brands like BuzzFeed and HuffPost in the UK. The entire revenue model rewards scale and mass appeal, and although it was built for an upmarket UK reader the brand now bets that AI tools, data-driven targeting, and editorial innovation can balloon scale and editorial quality together.
Read

Useful content is still the strategy
With the emergence of AI Overviews, chatbots, and all of these voice assistants answering questions before anyone even gets the chance to click a link, the model of publishing evergreen content and waiting for Google to send some traffic is fading. The argument is that utility content - straightforward service journalism that answers topline questions clearly - is still what works across platforms, whether breaking news or seasonal evergreen. Instead of the "set it and forget" approach, publishers should become more active: mapping out evergreen targets in advance, tracking the news cycle for content gaps, refreshing explainers, and recirculating related content - the fundamentals haven't changed.
Read

Your site isn't where AI finds you
Around 85% of AI citations come from third-party pages rather than brand-owned content, which means coverage on other sites matters more for AI visibility than your own optimisation. Mark Williams-Cook makes the case in one of his must-read SEO tips: if you're only optimising your own site for AI visibility, you're missing most of the picture. Being in more places equals increased chances of showing up.
Read

The end of bad software, not all software
Will AI agents make all SaaS products obsolete? Or will they wipe out those built around a bad premise: "bespoke" feature billing, screens only for humans to click through, bad APIs, and bad interoperability? Our CEO Denis Haman shares his thoughts on the flawed SaaSpocalyspe narrative.
Read

Big Tech

Meta employees become training data providers
Meta has started installing tracking software on employee computers, recording their mouse movements, keystrokes, and screens as they work, with the data going into training AI agents and LLMs. According to the company, the programme is an "opportunity" to help models get better, promising it wouldn't be used for performance reviews. Unconnected, the company is looking to cull 10% of staff. Meanwhile, Meta is also under attack from the Consumer Federation of America over scam ads on Facebook and Instagram.
Read

Bing to the rescue
The Internet Archive, which has been backing up the web since 1996, has run into some crawling obstacles, and Microsoft Bing came to the rescue. They offered to help with the adoption of IndexNow, a protocol that notifies search engines when content changes rather than requiring them to crawl for it. This should help the efficiency and overall work of the website that has been saving the web.
Read

Is that really you? 
YouTube is rolling out a tool that lets celebrities scan the platform for AI-generated videos that use their likeness, flag them, and request takedowns. Think Content ID but for faces instead of music. Celebrities or their agents can upload their likeness and the tool will scan YouTube content for offensive deepfakes, which will be flagged and reviewed. First in line were politicians, then creators, and now it's the turn of actors, athletes, and musicians. Not everything flagged will be taken down — parody and satire are still valid — but direct replicas that could replace someone's work are fair game.
Read

So much for a restricted release
Remember Anthropic's Mythos, the AI model so dangerous it can't be released to the public? A group of anonymous users has gained access to it through a third-party vendor, guessing the model's online location on the same day it was announced. The group, part of a Discord channel that hunts for unreleased AI models, even provided Bloomberg with screenshots and a demo as proof. Anthropic says it's been investigating, yet claims to have found no evidence it has been compromised. The group says it just wants to "play around, not cause mayhem", the entire restricted release programme was supposedly to prevent this sort of scenario.
Read

AI & Copyright

When the algorithm discriminates, who pays?
A jobseeker rejected from over a hundred positions on Workday's platform is suing the company for algorithmic discrimination, to which Workday claims it is not liable for the decisions of its platform's users. The court allowed the case to proceed, and insurers are watching closely. Who is legally responsible? Insurers are already stripping AI-related losses from policies, while others are capping payouts via sub-limits. The playbook looks like it will create a new coverage gap that will eventually turn into its own market.
Read

The AI slop hits music
According to Deezer, 44% of new tracks uploaded are now AI-generated, amounting to 75,000 new AI songs each day. Most of the streams they receive are fraudulent, generated by bots. Deezer's detection technology catches AI tracks with a false positive rate below 0.01%, and keeps flagged content out of recommendations and editorial playlists. While AI music accounts for only 1-3% of actual listening on the site and 85% of AI music is demonetised, the problem is that 97% of users surveyed cannot tell the difference between AI and human-made tracks. While AI music gets cheaper and watermarks get easier to remove, publishers fighting their own battles with AI can already see the parallels.
Read

Your Jiras will feed AI
From August, Atlassian will start collecting metadata and in-app data from its 300,000 global customers to train its AI models. If you're on a Free, Standard, or Premium plan, metadata collection is mandatory with no opt-out. In-app data - the actual content of your Confluence pages, Jira descriptions, and comments - is turned on by default for Free and Standard users. Premium and Enterprise customers are safe from this, for now. The collected data will be stored for seven years; users have until August 17th to make data decisions, so do look.
Read

Silicon Valley's copyright queue
The Northern District of California currently has 47 AI copyright lawsuits, more than every other country in the world combined. The Southern District of New York holds second place with 31, while every other jurisdiction has only a handful. Soon every active judge in the Northern District will likely have multiple AI copyright cases to deal with. That part of the US has picked up most of the legal reckoning of AI, and it's about to get busy in those courtrooms.
Read