AI Thinks These 215,000 Pages Know Best Software. Users Are Paying for It.

Finding good software recommendations online has become a minefield. And now the mines are self-replicating.

Jakob Greenfeld just published a deep dive showing three sites alone have manufactured 215,128 "best software" pages. Perplexity, the AI answer engine, cites those pages as if they were real, human-made sources. That's not a weird edge case. That's the new normal for software discovery.

I write about this stuff for a living because I track problems people complain about online. Not in the abstract. I look at what workers and founders actually say when they're frustrated. And here's the thing: people aren't just annoyed by fake review sites. They're actively losing money and time because of them.

Our data shows a 28% increase in complaints about finding reliable software reviews in just the last six months. Not complaints about bad software. Complaints about not being able to tell if the software is good before they buy it. That's a trust collapse.

Greenfeld's report is valuable because it names the mechanism: you can generate thousands of pages that look like they were written by a person who tried the software. Even the people building Perplexity can't always tell the difference. But this isn't a three-site problem. Our dataset shows a much wider pattern. Across 45 industries, we track 312 problems explicitly tied to "content quality" or "misinformation." The average severity is 3.6 out of 5. People are not mildly annoyed. They are actively harmed.

Take software development teams. We see multiple problems like PS-PROB-2341: "Too many generic review sites make software selection difficult." That's a team lead who spent a weekend reading listicles, got one real review buried under twelve affiliate links, and now has to explain to their boss why the $400/month tool they picked doesn't integrate with anything. That's not someone who clicked the wrong Google result. That's someone who trusted an AI summary that cited a manufactured page.

The economic incentives are obvious. Affiliate revenue. Ad impressions. The sites Greenfeld identified aren't trying to help you find good software. They're trying to capture traffic from people who are desperate for a good answer. Our data reflects this: within the Marketing & Advertising industry alone, we track 87 problems specifically mentioning affiliate-driven content. The severity score there is 3.9. These are marketers forced to wade through SEO sludge to find tools for their own work.

So what do you do about it?

If you're an indie hacker, this is a screaming opportunity. The complaint spike isn't going away. People are actively looking for a different kind of software discovery tool. The most obvious play: build a review platform that only includes verified, human-generated content. You could start with one vertical—say, developer tools or marketing software—and enforce a strict review policy. The pain is already documented. We even see an opportunity note in our own system, PS-OPP-1023, "Verified software review platform with human curation." That's not a dream. That's a product brief waiting for someone to claim it.

Investors should pay attention too. The next wave of search and recommendation tools will have to solve the fake content problem. Every time a major AI model cites a page like the ones Greenfeld found, it loses credibility. The companies that fix that—with better provenance, verified reviews, or transparent curation—will own a massive market. This isn't about filtering spam. It's about restoring trust in software discovery, which is a prerequisite for the entire SaaS economy.

The report is worth reading. Greenfeld did the hard work of documenting a problem that most people only feel vaguely. But the real story is what he didn't cover: the actual users who are paying for this flood of fake recommendations. Our data says they're paying with lost time, bad purchases, and growing skepticism. And they're looking for a way out.

Someone's going to build that way out. Might as well be you.