Why you should rethink using Google in the age of AI-generated content

In May 2024, people searching for basic cooking help on Google were told to put non-toxic glue in their pizza sauce. Others were advised to eat a small rock every day. These weren’t jokes or memes. They were confident, AI-generated answers surfaced by Google’s new search experience.

It’s easy to laugh at examples like this, but they point to something more serious: a web increasingly flooded by AI-generated content, and a search engine that’s starting to train on its own output. The result is what many are calling a “dead internet” — and it’s reshaping how we find information online.

From search engine to AI answer box

For most of its history, Google worked in a simple way: you typed a query, and it returned a list of links ordered by relevance. You chose where to click, compared sources, and decided what to trust.

Today, that experience is very different. Above the familiar list of blue links, Google now shows an AI-generated summary that tries to answer your question directly. The traditional search results are still there, but pushed down the page and easy to miss.

Those glue and rock recommendations came from this new AI layer. In one case, the model pulled a decade-old troll comment from Reddit. In another, it confidently cited a satirical Onion article as if it were real science. The problem isn’t just that the AI is wrong sometimes — it’s that it presents fiction as fact, with a design that encourages you not to question it.

The bot takeover: when half the internet isn’t human

Behind these weird answers is a bigger shift: a huge portion of the web is now created, scraped, and manipulated by bots. Recent reports estimate that more than 50% of all web traffic comes from automated systems rather than real people.

These bots aren’t browsing for fun. They’re:

• Generating mass content to rank in search results
• Scraping existing sites to train models or copy ideas
• Buying expired domains and turning them into AI content farms

Search engines like Google are still trying to filter the web into useful results. But now, a large and growing share of that web is synthetic — written by AI, optimized for algorithms, and designed to game ranking systems. Then Google’s own AI systems train on that synthetic content, creating a closed loop where AI increasingly learns from AI instead of humans.

Google’s March 2024 update: a cleanup that broke the web

In March 2024, Google rolled out a major search update that was supposed to fix low-quality content. It aimed to demote:

• Expired domains reused for spam
• Mass-generated pages
• Third-party sites built only to boost other websites

On paper, this sounded like a win for users. In practice, it hit many independent, human-run sites extremely hard. Studies tracking niche websites found that over 80% of them lost traffic between December 2023 and August 2024. Some review sites, like HouseFresh, reported losing over 90% of their Google traffic almost overnight.

For small publishers, it felt like a digital version of a giant billboard being placed in front of their shop. Their content still existed, but hardly anyone could see it anymore.

Parasite SEO and the rise of AI content farms

While many independent sites were struggling, SEO professionals and large companies quickly adapted. One of the most effective tactics is now known as “Parasite SEO.”

Instead of building up their own domains over years, marketers publish content on already trusted platforms like Medium, LinkedIn, and Reddit to ride on their authority. Google often doesn’t distinguish between a thoughtful, human-written post and a mass-produced AI article on those platforms — it just sees a strong domain and relevant keywords.

That’s where AI comes in. Companies can now generate hundreds or thousands of articles in minutes, each lightly tailored to a keyword. The content doesn’t need to be good or original; it just needs to exist in the right place with the right structure.

Some SEO experts have openly bragged about using AI to clone competitors’ entire site structures, spin up thousands of articles, and siphon millions of organic visits. Expired domains, once abandoned, are snapped up by bots, filled with AI content, and pointed at rival brands. It’s a full-blown “search heist” economy.

Zero-click search: when Google keeps you on Google

All of this is happening alongside another big shift: Google doesn’t really want you to leave Google anymore.

Over the years, search results have evolved from simple blue links to:

• Rich snippets (short text excerpts from pages)
• Knowledge panels (fact boxes, often powered by sources like Wikipedia)
• Featured answers and now AI summaries

These are all forms of “zero-click” results — you get your answer directly on the search page, without visiting any external site. According to data from SparkToro, nearly 60% of Google searches now end without a click.

For users, this can feel convenient. For websites, it’s devastating. Their content is scraped, summarized, and displayed in Google’s interface while they lose traffic and ad revenue. And for Google, it’s perfect: more time on their pages, more ad impressions, and more data to feed back into their models.

Reddit: unexpected winner… and new target

One of the few big winners in this new landscape has been Reddit. Between 2023 and 2024, Reddit’s organic Google traffic reportedly jumped from around 132 million to over 500 million visits. Why?

Because people still trust Reddit to feel “human.” It’s mostly text-based, easy to post on, and many communities are still moderated by real people. When you add “reddit” to a Google search, you’re often trying to bypass SEO spam and find real experiences and opinions.

Google noticed. In early 2024, it struck a deal reportedly worth around $60 million per year for access to Reddit’s data, allowing it to train AI models and enhance search results using Reddit posts. That same year, Reddit posted its first-ever quarterly profit as a public company, largely helped by this surge in search traffic.

But this success has a dark side. As Reddit’s data becomes more valuable, it also becomes a bigger target for bots. Users are already spotting repetitive, formulaic posts that look suspiciously AI-generated. Communities are trying to fight back with downvotes and moderation, but the incentives now favor more automation, not less.

Model collapse: when AI feeds on itself

AI skeptics have been warning about a phenomenon called “model collapse.” It happens when AI systems are increasingly trained on synthetic data — content produced by other AI models instead of by humans.

In early stages, this can cause minority or niche information to disappear. The model learns to favor the most common patterns and filters out anything that doesn’t match, making it harder to get specific or unusual answers.

In later stages, the model’s performance starts to degrade more visibly. It repeats errors, becomes less accurate, and is harder to correct. Researchers compare it to inbreeding: defects get amplified over generations because the gene pool (or data pool) keeps shrinking and recycling itself.

Google is in a particularly risky position here. It handles around 90% of global search queries, and its only real competitor in query volume is YouTube — which it also owns. As more of the web’s content becomes AI-generated, and as Google’s own AI systems train on that content, the risk of a self-reinforcing feedback loop grows.

The glue and rock answers aren’t just funny mistakes. They’re early warning signs of what happens when the world’s biggest information system starts to eat its own tail.

The “dead internet” problem

Europol has warned that up to 90% of online content could be AI-generated by the end of 2026. That number might sound extreme, but you can already see the trend:

• Social feeds filled with AI-written posts designed to provoke reactions
• Ghost websites publishing thousands of AI articles a day that no human ever reads
• SEO spam flooding every niche topic with low-quality, automated content

These pages don’t need human visitors to be useful. Their real audience is Google and other AI systems. As long as they get crawled, indexed, and used as training data, they’re doing their job.

That’s what people mean by a “dead internet”: a web where the majority of content is synthetic, optimized for machines, and increasingly disconnected from real human experience.

Can we build a human-first internet again?

Despite all this, not everyone is ready to accept an AI-saturated web as inevitable. A small but growing group of users is actively seeking out tools and spaces that prioritize human content and transparent search.

New search engines like Kagi and Marginalia focus on:

• Highlighting small, independent sites
• De-emphasizing commercial and SEO-heavy content
• Offering optional AI help instead of forcing it by default

They’re niche, and often paid, which makes them a harder sell compared to “free” Google. But they point toward an alternative: an internet where you’re not just training data, and where discovery isn’t entirely controlled by one company’s opaque algorithms.

Similarly, more people are experimenting with AI tools in a controlled way — using them as assistants rather than as the final source of truth. For example, if you’re interested in building your own AI-powered apps or workflows, tools like Google AI Studio or Perplexity Computer can be powerful, as long as you stay in the habit of verifying outputs and checking original sources.

Practical steps you can take right now

You don’t have to quit Google completely to protect yourself from the worst effects of AI-driven search. But it helps to change how you use it.

Here are some simple habits:

1. Scroll past the AI answer.
Don’t treat the AI summary as the final word. Scroll down, scan multiple sources, and compare what they say.

2. Add human filters to your searches.
Include terms like “reddit,” “forum,” or “blog” in queries when you want real experiences instead of SEO content.

3. Favor original sources.
When possible, go to primary sources: official documentation, research papers, or well-known experts in a field.

4. Try alternative search engines.
Experiment with human-first search tools for certain topics, especially research-heavy or niche questions.

5. Be skeptical of confident answers.
Whether it’s Google’s AI, a chatbot, or a random site, treat every answer as a starting point, not the end of your research.

The fork in the road

The internet is at a crossroads. One path leads to a web dominated by AI-generated content, optimized for algorithms and advertising, where users mostly see what a handful of systems decide to show them. The other path is messier and less convenient, but more human: smaller sites, independent creators, and tools that respect your ability to think and choose for yourself.

Google’s current direction — from AI summaries in search to AI suggestions in email — makes it clear which path it prefers. The question is whether users will quietly accept that, or start demanding and building something different.

For now, the most powerful thing you can do is simple: slow down, question what you’re shown, and deliberately seek out human voices. The more we rely blindly on AI-filtered answers, the faster the internet becomes a hall of mirrors. The more we push back, the more room there is to keep the web alive, weird, and genuinely useful.