Skip to contentIs the Future of AI "Digital Mad Cow Disease"?
by Christopher Spencer B.A. (Hons) MCIM
Chartered Marketer and Co-Founder of Peak Media Marketing
Is AI Eating Itself?
The Threat of Synthetic Media Saturation.
The future of everything looks quite different now that AI is here. We already have AI embedded in CCTV in shops that is watching how we buy stuff and “optimising the shopper experience” so that they can orchestrate all in‑store activity at one time.
This means AI can see where people dawdle, linger and loiter so the shops can use dynamic promotions wherever our attention is and basically turn shopping into a building full of stuff for sale with an algorithm.
The AI's job is to streamline processes and enhance the shopping experience to optimise the spend per head.
AI is predicted to be doing pretty much everything in the near future and it's getting cleverer by the minute.
It’s not just being used to create homogenised “one size fits none” bone idle marketing that is beige, bland and boring; it’s also doing a myriad of things from autofilling one‑time codes for SMS verification techniques all the way to the stars, literally enabling spacecraft to navigate, detect anomalies, and reconfigure systems without needing to wait for ground control. (or Major Tom)
It was all looking hunky-dory.. until we got involved. That’s right. Humans.
Synthetic Media Saturation Without Provenance.
Just in the same way that the dinosaurs roamed the Earth for 200,000,000 years and were every bit the better tenant than we are because in a fraction of that time we have engineered everything to suit us at the cost of… pretty much everything that’s not us. We have pretty much engineered a human‑driven mass extinction which is already nicely underway. (well done, us)
And now we’re also creating Digital Mad Cow Disease.
In the same way that British farmers fed ground‑up meat and bone meal (from cows) to cows and caused “mad cow disease”… we are now doing the same to AI. (another slow clap for humans)
This is by means of a process called synthetic media saturation without provenance where we allow unlabeled AI content to flood the web. Crawlers then treat it as “real,” training future versions of models on AI‑created content which can be riddled with hallucinations, artifacts, and errors.
So the AI models are now learning from a blurrier copy every time they crawl “new info” and quality collapses over time because we are feeding them their own output.
(and we wonder why Skynet had it in for us?!)
What synthetic media saturation does to the internet
If we keep flooding the web with unlabeled AI output, the internet gets less reliable, less diverse, and harder to trust. Search starts to feel like déjà vu: many pages repeat the same points in the same voice, carrying the same mistakes. Confident answers will increasingly cite AI‑written sources that themselves cite other AI‑written sources, creating an authoritative‑looking echo chamber.
The quirky long tail of human knowledge gets buried under smooth filler, while spam and scam content scale up because it’s cheaper than ever to produce.
Under the Bonnet: Provenance Rot and Feedback Loops
Crawlers can’t reliably tell human from machine, so training sets get contaminated and models learn from their own outputs. Retrieval systems begin indexing chatbot answers and later quote them back, reinforcing errors. Autogenerated code and documentation spread subtle bugs and insecure patterns. On dashboards, benchmarks may still tick upward, but real user trust and satisfaction erode.
Knock-on Effects for the World Wide Web Ecosystem.
- Ad markets and SEO arms races intensify, squeezing smaller publishers who can’t win on volume.
- Platforms spend more on moderation and still risk over‑ or under‑filtering, distorting what remains.
- Journalists, researchers, and educators lose time to verification because source quality is uncertain.
How to Prevent “Digital BSE” Without the Doom!
This trend isn’t inevitable. Platforms and organizations can require content provenance (signatures/C2PA) and down‑rank unlabeled AI pages; block re‑ingesting their own outputs; maintain human‑verified “clean room” datasets; and track drift with hidden human‑only evaluations, self‑citation detectors, and novelty metrics.
Wait, What are they?
- C2PA (Coalition for Content Provenance and Authenticity): is an open technical standard that allows creators and publishers to attach tamper-evident metadata to digital content (images, videos, audio, documents).
- Digital Signatures: At its core, C2PA uses cryptographic signatures to verify that the provenance information hasn't been altered. It's like a digital seal that breaks if someone tries to change the history.
(like that cryptex thingamajig in the Da Vinci Code)
How they work:
Capture: When content is created (e.g., a photo taken on a phone, an article written in a word processor, or an image generated by AI), information about its origin is captured. This includes:
- Creator: Who made it (e.g., "Chris Spencer," "Acme AI").
- Tool: What software or device was used (e.g., "iPhone 15 Pro," "Midjourney v6," "Photoshop").
- Actions: What edits were made (e.g., "cropped," "color adjusted," "AI-generated background," "text translated by AI").
- Date/Time: When these actions occurred.
Embed: This provenance information is then securely embedded directly into the content file itself. It's not stored separately, making it harder to lose or detach.
Sign: The embedded information is cryptographically signed. This signature acts as a tamper-evident seal. So if anyone tries to alter the content or its provenance data, the signature will no longer match, indicating that the content's history is no longer trustworthy.
(laughs like Skeletor)
Verify: Viewers or systems can then use a C2PA-compatible tool (like a browser plugin, a dedicated app, or built-in features in software) to inspect this metadata and see the content's history.
Why They're Crucial for "Digital Mad Cow Disease":
- Transparency: They provide a clear, verifiable record of whether content was human-made, AI-generated, or edited by AI. This directly combats "synthetic media saturation without provenance."
- Trust: By knowing the origin, users and AI models can make informed decisions about whether to trust or use a piece of content.
- Filtering: AI models can be trained to prioritise content with strong, human-verified provenance and down-weight or ignore content with unknown or AI-generated provenance. This helps prevent models from "eating their own output."
- Accountability: If misinformation or harmful content spreads, C2PA can help trace its origin, making creators and platforms more accountable.
As users and creators, we can support primary sources, link to evidence rather than summaries of summaries, and clearly mark AI‑assisted work. With provenance, curation, and ranking signals that reward originality and source transparency, the web can stay a place where new human knowledge surfaces instead of it becoming a quagmire of Human AI Centipede created microwaved marketing that gets worse in quality and greater in quality every year.
So use AI but do it as sparingly as possible so as not to create Artificial Intelligence Spongiform Encephalopathy - or Digital Mad Cow Disease.
Subscribe to Our Moosletter
Don’t worry it's spam and mad cow free.
It's definitely not microwaved brain rot for doom scrolling either. Only the good and interesting stuff from us.
