Remember when you could type a question into Google and get an answer? Not a "Featured Snippet" scraped from a blog post written by a language model trained on other blog posts written by earlier language models? An actual, human-authored, experience-informed answer? That was nice. That was a different internet. Pour one out.
Today, searching Google for something like "best running shoes for flat feet" returns approximately eleven thousand words of content that no human being with flat feet has ever read, let alone written. You'll scroll past a listicle that ranks products the author has never touched, a comparison table assembled by a content farm operating out of a shared Google Sheet, and at least three articles that open with the phrase "In today's fast-paced world." None of them will tell you anything useful. All of them will rank above the podiatrist who wrote three paragraphs about it in 2019.
This is not an accident. This is the business model.
The Algorithm That Ate Itself
Here's the uncomfortable truth about how Google's search index became a landfill: the platform spent years telling content creators exactly what the algorithm wanted, and content creators — being rational economic actors — gave it exactly that. Keyword density. Header tags. Internal links. Word count minimums. "Comprehensive" guides that cover every conceivable angle of a topic without actually saying anything about any of them.
For a while, this produced mediocre content written by humans trying to game a system. Annoying, sure, but at least there was a person somewhere in the chain who had theoretically eaten a sandwich or owned a dog or experienced the thing they were describing.
Then AI content generation became cheap. Then it became free. Then it became something you could run on a laptop overnight and wake up to find you'd produced four hundred articles about kitchen appliances, none of which required a kitchen.
The content farms didn't hesitate. Why would they? Google had already trained them to optimize for signals rather than substance. AI just removed the last remaining bottleneck: the human who had to actually type the words.
What "Helpful Content" Actually Means
Google has been running its so-called "Helpful Content" updates for a couple of years now, loudly promising to demote AI slop and surface genuine expertise. The results have been… mixed, which is a diplomatic way of saying the slop is still very much present and doing great, thanks.
Independent researchers and SEO watchers have documented site after site — actual experts, actual practitioners, people with actual credentials and actual readers — getting torched in algorithm updates while thin, templated content from large content networks sailed right through. The pattern is consistent enough that it stopped being a coincidence and started being a feature.
The dirty secret is that Google's quality signals were always proxies. Domain authority. Backlinks. Click-through rates. Time on page. None of these measure whether the content is true, useful, or written by someone who knows what they're talking about. They measure whether the content looks like content that people clicked on before. AI-generated articles, optimized correctly, can fake every single one of those signals. And they do.
The AI Answers Are Also Wrong
Just when you thought it couldn't get more chaotic, Google decided the solution to bad search results was to generate the answer directly, skipping the results entirely. Enter AI Overviews: the feature that confidently tells you to put glue on your pizza and eat one small rock per day for digestive health.
Those examples became memes because they were funny. They're less funny when you realize the same confident wrongness is being applied to medical questions, legal questions, financial questions, and anything else a person might type into a search bar during a moment of genuine need. The AI doesn't know it's wrong. It has no mechanism for knowing. It produces fluent, authoritative-sounding text because that's what it was built to do, and fluent, authoritative-sounding text is exactly what a frightened person searching for answers wants to see.
The result is a search engine that has become, at scale, a misinformation delivery system dressed in the visual language of a reference tool. It looks like an answer. It feels like an answer. It is, with some regularity, completely made up.
Who Actually Wins Here
Follow the money, as they say, and you end up somewhere uncomfortable.
Content farms that produce AI slop at volume need somewhere to host it, which means ad revenue for platforms. Google sells ads against search results, including the terrible ones. AI companies need training data, and the web — even the bad web — is full of it. The worse the content ecosystem gets, the more valuable "premium" information sources become, which creates a market for paywalled databases and subscription research tools that used to be freely available.
The people who lose are the ones who just wanted to know whether ibuprofen is okay to take with their blood pressure medication, or which local plumber won't charge them four hundred dollars to look at a drain. Regular people. The people search engines were allegedly built to serve.
Instead, those people get twelve paragraphs of AI-generated hedging that ends with "consult a professional," which is technically correct and completely useless, followed by four sponsored listings for services they didn't ask about.
The Indie Web Is Still Out There (You Just Can't Find It)
Here's the genuinely maddening part: the good content still exists. Human experts still write things. Enthusiast communities still maintain wikis and forums packed with hard-won, specific, accurate information. People who actually own flat feet and have actually tried seventeen pairs of running shoes still post about it somewhere.
You just can't find any of it on the first page of Google anymore. Or the second. Sometimes it surfaces on Reddit, which is why "reddit" has become the most common word people append to their searches — a manual override, a desperate signal to the algorithm that they want to hear from a human being for once.
The fact that users have developed workarounds to escape the search engine's own results is not a sign of a healthy product. It's a sign that the product failed and nobody in charge has had to feel the consequences yet.
Google's search page used to be a window. Now it's a mirror — reflecting back an AI's best guess at what an answer should look like, optimized for engagement, monetized at every layer, and largely disconnected from the actual information that actual people actually need.
The internet is still out there somewhere, broken and beautiful and full of genuine human weirdness. Google just stopped helping you find it.
We noticed.