In the course of writing a eulogy for a dearly loved and recently departed relative last week, I had cause to marvel at the sheer complexity of the human writing process. The speech in memory of their life was well received by all those who heard it, and indeed was praised by many, it making them by turns happy, sad, and reflective, which is the purpose of a eulogy as far as I’m concerned.
Yet as I told my partner that I, as a writer of some modest skill was pleased with the result of my efforts, and she said “I’m glad you’re happy with what you’ve written”, I immediately responded “Well, truth is I’m never really happy with anything I’ve written, and if I’d written it today it would be almost entirely different.”
It’s true. Even if the aim of the writing - evoking the memory of the departed as warmly and effectively as possible - was an irreducible purpose, the characterisations used, the words, the structure, could all have been different. The human brain still yet has a few tricks up its cerebral sleeve in its almost endless capacity for creation.
Which leads us today on to the fast-warming topic of content generated by AI, how to identify it, and moves to make it easier to do. And why it may encourage the very things we are trying to avoid.
Read the full article on GPP.io
Same platform, real numbers
Games site AltChar ran a full redesign alongside a migration to Glide Deliver V3, and the numbers are speaking for themselves - even though we are about to too: Lighthouse mobile performance jumped from 78 to 98 on deployment, SEO scores hit a sweet 100, and server response times sit at an uberpeppy 6ms in production. The site runs serverless on AWS Lambda, scales automatically on traffic spikes when major games release, and is obliterating its previous user numbers. Got an old clonky CMS? Read on... the full case study breaks down the before-and-after.
Read
Drop the middleware
Most publishing teams have CMS and email tools as separate systems, bridged by manual effort. If you're hitting one send weekly, you can tolerate it, but when you grow and want to send perhaps dozens of daily briefings across multiple brands and audience segments, the bottleneck needs removing. Glide CMS does exactly that: editors write, target, and send from one place, with taxonomy-based audience selection and entitlement-aware delivery handling the complexity underneath.
Read
The Globe's anti-AI playbook
The Boston Globe grew its subscriber base 180% since 2019, now sitting on a cool 313,000 paid digital-only subs, by investing in regional bureaus, specialist topic coverage, and bundled subscriptions to give people more reason to subscribe. On AI Overviews, organic search has only dropped 3% year-on-year, with the Globe reminding everyone that original local reporting is not something AI can replace. Press Gazette has more details.
Read
Easy joining, easy leaving
From January 2027, the UK government is introducing new rules requiring subscription cancellations and departures to be as easy to do as signing up in the first place. There is an estimated £1.6 billion a year spent on unwanted subscriptions across 155 million active plans, and new UK Prime Minister Andy Burnham believes it is a basic cost-of-living issue as well as good consumer rights practice. Also on the block are "phoney bargains" which purport to show discounts where none really exist. Run a subscriber programme? You have a few months to get your ducks in a row.
Read
Quality journalism = quality comments
A recent study analysed nearly a million comments on 6,400 New York Times stories, and revealed that data-heavy and analytical pieces result more often in comment sections where readers reply to each other and debate the evidence presented. Overall, such articles actually attract fewer commenters, as people are more drawn towards emotional pieces rather than scientific ones, but the substance is stronger and it is where more engaged, loyal readers hang out.
Read
Listicles out, product pages in
New research from Peec AI shows that ChatGPT's latest free-tier model is conducting far more fan-out queries per prompt than before, and is pickier about what it retrieves. Product pages are rising, while lists and how-to guides are dropping. The model is leaning heavily on site searches and the word "official" in order to control source quality. What's interesting is that pages previously on a pedestal for GEO purposes are now falling out of favour, reminiscent of the early days of E-E-A-T except moderated by chatbot.
Read
From typo to takeover
According to researchers at pwn.ai, a WordPress vulnerability chain can escalate failed logins into remote code execution, with no account required. The exploit relies on a parser disagreement around since version 4.7, which allows attackers to smuggle HTML into the login page and then hijack a leftover script to run arbitrary code. It has been patched in 7.0.3 and a fix backported across every supported branch, so if you're self-hosting do look for the patch.
Read
Signed out, opted in
Google is testing a homepage for non-logged-in users where the Search button is replaced with three AI shortcuts: image generation, file Q&A, and a Gemini brainstorm. The familiar layout returns when you log in, for now. Search Engine Watch reproduces it across multiple browsers, while others couldn't trigger it at all. Of course the signed out user is who Google knows the least about, so enticing them to sign up or sign in will be a goal.
Read
The Section 230 shield has limits
A US appeals court has decided that, despite Meta and TikTok's arguments that Section 230 grants them blanket immunity, the thousands of lawsuits accusing them of designing addictive, harmful products can proceed. According to the court, while the law provides a defence against some liability, it doesn't offer blanket protection from everything. This comes after a jury verdict that found both Meta and YouTube liable for harmful social media addiction. There is also a separate trial going on for Meta over allegations it used children's data to keep them on platforms.
Read
Meta's reapy crawly
Reports of soaring activity by Meta bots scraping sites and content hint at the firm building its own search engine. While Meta failed to build a search engine previously in an attempt to compete with Google, this time the motivation seemingly stems from not wanting Facebook and Meta content to be exposed to Google, and thus boosting a rival AI. Mark follows where others lead.
Read
Nostalgia as a growth strategy?
The original owners of MySpace confirm a relaunch is coming, built around features which people remember and love, such as profile songs and a chronological feed with no algorithm. The pitch? You-curated social media without the infinite scroll.
Read
Watermarks you can't see or trust
Anthropic promised to embed invisible watermarks in things Claude generates, to comply with the guidelines set out by EU's AI Act. This will be applied at the model level across every product and third-party provider worldwide, and the company says watermarks will travel with copied text, persist through some editing, and won't affect quality or meaning. However, the marks aren't conclusive proof of AI authorship, especially since there are open-source tools which can strip the file-level metadata.
Read
It's not three men in a trenchcoat - it's AI
An investigation into a company which offers peer-review-ready manuscripts and systematic reviews to academics, claiming zero AI involvement, suggested that half of the listed team are, err... AI-generated fakes, and the other half are real people whose likeness and work is being used without their knowledge. It looks like everything from phone, email, chat support and more are simply AI agents cosplaying as humans. If the folks in academia aren't worried yet, this will surely do it.
Read
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.