AI is Eating Itself

Google made answers easy. AI is making us source-blind.

“Ask the chatbot and get an answer.” That sounds good until you ask where the answer came from, who created it, who verified it, who updated it.

And if you don’t verify the answer, you are rolling the dice if the answer is right, wrong, or incredibly misinterpreted.

AI did not create the original open web. People did. Writers did. Researchers did. Practitioners did. Teachers did. Scientists did. Curators did. Website owners did. The open web was built by people spending time, money, experience, and judgment to make information available, shared, and improved upon.

Now AI systems scrape that information, compress it, summarize it, and hand the answer to the user before the user ever reaches the source.

Search used to work like this: Search → click → read → compare → decide.

AI changes that to this: Ask → answer.

The old way was slower. It was also better for the mind.

Before Google, and honestly before the Internet became the default answer machine, finding answers required physical and mental work.

You went to the original sources. Books. Reports. Manuals. Case files. Journals. People with direct experience. You took notes. You compared one source against another. You looked for conflicts. You noticed what was missing. Then you had to make sense of it. You had to interpret it.

That process took hours or days. Sometimes weeks. It was inefficient by modern standards. But it did something valuable that is becoming lost. It transformed the researcher in a way that the Internet and chatbots never will.

The person who did the work remembered the material differently. Not because they memorized every line, but because they fought with it. They saw the contradictions and tracked the sources. They knew which fact came from which place and why one source was stronger than another. Sometimes they threw the books across the room in frustration, only to pick them up and apologize.

That kind of learning is not just information retrieval. It is judgment formation. It is comprehension at a deep level. And it can last a lifetime as the principles learned can be applied throughout life.

A chatbot user gets a different experience. They ask. They receive. They move on. Subconsciously, they often treat the chatbot as the source. They do not remember the underlying source because they never saw it. They may not even remember the answer because they did not work for it. The answer appeared too easily, so the mind assigns it less weight. Useful in the moment to get a quick answer and tossed aside. That which is free or easy or cheap is simply used and neglected. That which is earned is carried.

I’m a GenXer, which means I grew up through the last stretch of the analog world and then watched the digital world eat it piece by piece. I lived through card catalogs, rotary phones, analog television, handwritten notes, paper maps, Walkmans, boomboxes, VHS tapes, floppy disks, command lines, MS-DOS, Windows, dial-up Internet, early web browsers, search engines, USB drives, BlackBerry phones, smartphones, GPS, social media, cloud storage, and now AI chatbots. Some of those things arrived before I was old enough to understand them. Some arrived when I was already working. But I lived through the shift.

I remember when knowledge had weight and it lived in books, binders, tapes, disks, folders, devices, and people. You had to go find it. You had to touch the source, compare it, copy it, label it, store it, and remember where it came from. Now the answer appears in seconds, stripped from the trail of sources that produced it.

I wrote about this topic in 2018, (https://brettshavers.com/brett-s-blog/entry/i-don-t-need-to-learn-just-give-me-the-answer) before AI chatbots became the default answer machine. Back then, Google was the shortcut. My warning was that being handed an answer is not the same as learning how to find one. The answer solves the moment but it is the search builds the mind. The journey to the answer is what matters. The failed attempts, the source checking, the comparison, and the eventual “I found it!” moment are what make knowledge stick. AI did not create this problem. It industrialized it. Google made answers easier to find. AI makes the source easier to ignore.

The chatbot becomes the source.

This is where the damage gets worse. At first, AI systems summarize sources. Then users stop visiting the sources.

Then publishers, researchers, curators, and independent site owners stop producing public work because there is no return, their research is plagiarized and misinterpreted by AI algorithms. Then the web fills with outdated pages, AI-written summaries, recycled listicles, copied summaries of copied summaries, and search-optimized sludge.

LinkedIn, in particular, is flush with “articles” written fully or almost fully by AI. I have personally had my writings and work scraped by someone who used AI to turn it into an article that was completely wrong. That is the problem in miniature: human research gets scraped, AI distorts it, someone posts it as content, and another human has to clean up the mess. Worse is when the inaccurate information becomes the source. This is not limited to people who do not know what they are talking about. It is also being done by people who should know better (this example was exactly that).

We ask chatbots to write for us and it pulls from sources online and training data. Then we post that back onto the Internet. To which the cycle begins with AI producing content from AI created content. Then future AI systems train on that. AI trained on AI created content, two, three, and more steps away from a human.

This is the content death spiral.

1. AI takes from the source and changes it.
2. AI becomes the source.
3. The original source disappears.
4. Then AI feeds on what is left.

Bad sources become undetectable mistakes.

There is another problem. When a chatbot gives an answer, the user often cannot see the evidence chain. If the original source was wrong, the chatbot may repeat or magnify the error. If the source was incomplete, the chatbot may fill the gap. If multiple sources disagree, the chatbot may try to merge disagreements into a believable answer, even if it is wrong.

Truth becomes a luxury

Here is where this goes if nothing changes. The open web becomes a graveyard and pages will be outdated or even abandoned. “New” content will be AI-filler based on outdated Internet content and worse (AI-slop created from AI-slop created from AI-slop created from…).

The cheap and easy answer will be everywhere and the verified answer will be harder, if not impossible, to find. The human-curated answer will move behind gates, memberships, private communities, paid databases, licensing deals, newsletters, courses, and closed networks.

That means truth becomes expensive in both time and money. The people who want the real answer will pay for it. The people who cannot pay, or will not pay, will get the synthetic version. Raw facts become cheap while the verified truth becomes premium.

AI is a photocopy of a photocopy.

The best analogy is still the old one. A photocopy of an original can be useful. A photocopy of a photocopy is worse. Keep copying the copy and the edges blur and eventually the whole image turns into gray mush.

That is what happens when AI eats AI.

So what is the answer?

The answer is not to stop using AI. It is here to stay and will become more powerful because it is useful and fast. Refuse to use it completely and you fall behind. Rely on it completely and you will fall behind.

In my opinion, toss the old ways that are outdated. But in the same vein, keep the old ways that are effective.

Here is a personal example: I attended a seminar by an OG in a field who sadly passed away not long after the seminar. I listened intently to his human experience, and spoke with him during a break. That one day stays with me more than having read a book or watched a video. That was a human experience. Every person I meet in this field (and in life generally), is like that for me, intentionally. I listen to the source’s lived experience and connect. We can’t do that with a chatbot.

So now you know why I will wait in line to meet every speaker in every conference I’ve ever attended, and you know that I remember everything you said in that 30 seconds. Because it was important to hear.

The last person feeding the alligator still gets eaten like everyone else.

AI is eating itself. But before it eats itself, it eats the people feeding it. It eats the original sources, the job roles, the researchers, the curators, the writers, and the innovators. When the original work dries up, AI will eat the leftovers. AI doesn’t remember or care who fed it last.

Harlan Carvey inspired this post (which is usually the case for me…..) with a LinkedIn post and the Blade Runner scene he picked. Take a read and when you watch the short video, consider the replicant as representing all the OG DFIR practitioners who didn’t document their work, or publish and share their findings.

Who will own the future?

The future will not belong to the people with the fastest answer. Everyone will be able to do to spit out an essay or answer in 30 seconds having not an idea of the source of the essay or answer, let alone even understand it. They’ll forget what the chatbot said 33 seconds after the prompt.

The future will also not belong to the “prompt engineer.” The hype we see that this is the biggest skill to have is a trap. It is a trap because the best prompt engineer is no better than anyone else if they can’t comprehend and verify the answer.

For those in the DFIR field, if you know bit-level forensics, you are most likely a rare source of knowledge. And if you also have a critical thinking mind…no, that’s not accurate..if you have an investigative mindset, you are not only a rare source of knowledge, but a finder and verifier of any and all sources.

The future will belong to the people who can prove where the answer came from, what it means, and why it matters. That will soon become the most expensive thing on the Internet.

So…for the employers

I find it odd when management replaces humans with AI (chatbots, agents, or otherwise) don’t realize that they are simply scheduling their own replacement for later, just like the last guy feeding the alligator. Service and products become worse in quality, customer service because chatbot frustration service, and in the end, a computer is running what is left of an organization.

For everyone

Technology has enabled us to do so much more than ever in history. We “connect” instantly to virtually anyone in the world via typed instant messages or real-time video. With the advent of AI, we have instant answers to hundreds of years of information. Yea!

But with that, we have lost true connections to people by sending messages instead of a call, an email instead of meeting face-to-face, an emoji instead of a hug or handshake. We are accepting a chatbot as the source of information, the answer that we want through biased prompts and biased programming. 

I hope we get through this evolution of technology intact as one people.