Report Uncovers Widespread AI-Generated ‘Best Software’ Pages Cited by Perplexity

A manufactured source is a low-quality, often AI-generated webpage created specifically to manipulate search engines or AI models into citing it as an authoritative reference. Reported evidence suggests a growing ecosystem of these pages is now polluting the “software recommendation” space by deceiving AI search engines like Perplexity into promoting specific tools.

We have all grown accustomed to the convenience of AI search. Instead of scrolling through ten different websites to find the “best video editing software,” you ask an AI like Perplexity, and it gives you a neat, summarized list with citations. But what happens when the very citations meant to provide trust are actually smoke and mirrors? A disturbing new trend in the tech world suggests that the “verified” sources your AI is reading might just be other AI models talking to themselves in a digital echo chamber.

According to a report originally highlighted on Hacker News, there is now a widespread operation involving AI-generated “best software” pages. These sites are not built for humans to read; they are built for AI models to find. By mimicking the structure of a legitimate review site, these manufactured sources are successfully infiltrating the knowledge base of major AI search tools. This creates a loop of misinformation where artificial intelligence is essentially being trained and informed by its own synthetic output, often at the expense of accuracy and consumer safety.

What are manufactured sources in AI search?

Manufactured sources are essentially “content farms” on steroids. While the old-school content farms of the 2010s relied on low-paid writers to churn out SEO-optimized fluff, today’s version uses large language models to generate thousands of pages of “software reviews” in seconds. The goal is simple: capture the attention of crawlers from companies like Perplexity, OpenAI, and Google to ensure specific software products are recommended to users.

When you ask a search engine for a recommendation, it looks for consensus across the web. If a hundred different sites all claim that “Software X” is the best tool for accounting, the AI will likely repeat that claim. The problem is that these hundred sites are often owned by the same entity, using the same AI prompt to generate slightly different versions of the same praise. This is not just a technology problem; it is a fundamental shift in how search works in 2026, where the volume of content now outweighs the quality of the insight.

This tactic is particularly effective in the software niche because recommendations are highly profitable. Affiliate commissions for enterprise software can reach hundreds of dollars per signup. By creating these manufactured sources, operators can direct lucrative traffic toward their preferred partners while the AI search engine unknowingly acts as the middleman, providing a veneer of objective authority to what is essentially a hidden advertisement.

How AI-generated pages deceive Perplexity and other search engines

Search engines like Perplexity use Retrieval-Augmented Generation (RAG) to find relevant documents and summarize them. In theory, this makes the AI more accurate because it is grounded in real-world data. However, the report indicates that these AI-generated pages are specifically engineered to satisfy the RAG process. They use headers that look like common search queries, such as “Is Software Y worth it in 2026?” and “Pros and cons of Software Z,” which makes them highly “slurpable” by AI crawlers.

In practice, these pages often contain no original research. Look at a traditional review site, and you will see screenshots, specific performance benchmarks, and nuanced critiques. In contrast, the manufactured sources cited by Perplexity frequently use generic language like “powerful features” or “user-friendly interface” without ever explaining why or how. Despite this lack of depth, the sheer SEO optimization of these pages allows them to sit at the top of the digital pile when an AI engine goes looking for answers.

This isn’t the first time we’ve seen tech giants struggle with data integrity. We recently covered how global tech leaders are calling for AI safety regulations to address exactly these kinds of systemic vulnerabilities. The risk here is that if the primary “evidence” for a software recommendation is a synthetic hallucination, the user ends up with a product that might not even work as described, or worse, contains security flaws.

The impact of misinformation on software recommendations

When a user asks for the “best secure messaging app” and receives a list cited from manufactured sources, the stakes go beyond a simple bad purchase. If the AI cites a page that was generated to promote a specific tool with known vulnerabilities, the user’s data privacy is at risk. This mirrors concerns we’ve seen in other sectors, such as the FBI investigation into massive data breaches, where the integrity of information systems is compromised by bad actors.

The report found that many of these cited pages were “circular.” A page might cite another AI-generated page as its source, creating a chain of misinformation that is incredibly difficult for a standard algorithm to break. For the average consumer, a citation from Perplexity feels like a seal of approval. You see a little footnote, you think it has been fact-checked, and you move on. The reality is that the “source” could be a domain that didn’t even exist three months ago.

Think about it this way: if you were buying probiotic supplements for gut health, you would want to know the recommendation came from a lab or a nutritionist, not a bot. Software is no different. Whether it is a VPN, a password manager, or a simple photo editor, the recommendation should be based on human experience, not a statistical probability of word sequences.

Why AI content farms are winning the SEO war

The speed at which content farms can produce material is their greatest weapon. A human editorial team might take a week to properly test and review a piece of software. An AI-driven operation can “review” 500 products in an afternoon. Because Perplexity and other search tools prioritize “freshness,” these newly minted AI pages often leapfrog older, more reputable human-written content that hasn’t been updated in a few months.

Furthermore, these sites are designed to be “citation-friendly.” They use bulleted lists, clear pros/cons tables, and bolded conclusions. This structure is exactly what AI models are trained to look for when they need to extract a quick answer. The truth is that we have built an information ecosystem that rewards the appearance of organization over the substance of the facts. It is a classic case of the “GIGO” principle: Garbage In, Garbage Out.

1. Scale: Thousands of pages are generated for every possible long-tail keyword.
2. Structure: Content is formatted as a direct answer to common AI prompts.
3. Authority Spoofing: Domains are often expired sites with existing “backlink juice” that are repurposed for AI content.
4. Affiliate Poisoning: Recommendations are biased toward products that pay the highest commission, rather than the highest quality.

How can you tell if a source is manufactured?

Identifying these pages requires a skeptical eye. One major red flag is the “Author” profile. Many of these sites feature AI-generated headshots and bios of “experts” who don’t exist on LinkedIn or anywhere else on the web. Another tell-tale sign is the lack of specific details. If a review of a video editor doesn’t mention how long it took to render a 4K file or how the interface handled a specific plugin, it likely wasn’t written by someone who actually used the software.

You should also look at the publication date and the site’s history. Many manufactured sources are hosted on domains that used to be about completely different topics. A blog that was about “sustainable gardening” in 2023 suddenly becoming an authority on “enterprise cloud security” in 2026 is a massive indicator of a content farm operation. We see this often in the lifestyle space, where a site might pivot from eco-conscious home essentials to high-tech software reviews simply because the latter pays better.

The response from Perplexity and the AI industry

Perplexity has marketed itself as the “answer engine” that provides transparency through citations. This report puts that value proposition under the microscope. While the company has not yet released a specific patch to address “manufactured sources,” they have previously stated that they are constantly refining their ranking algorithms to prioritize high-authority domains.

However, the technical challenge is immense. Differentiating between a well-written AI summary and a human-written review is becoming nearly impossible for automated systems. As artificial intelligence becomes more sophisticated, the “tells” that used to give it away are disappearing. The industry is now in an arms race where search engines must build better “AI detectors” just to keep their results from becoming a hall of mirrors.

Some experts argue that the only way forward is a “white-list” approach, where search engines only cite a pre-approved list of trusted journalistic and academic sources. But this goes against the open nature of the internet. It would mean that a new, legitimate blog would never get cited because it isn’t on the “trusted” list. It’s a difficult trade-off between the diversity of information and the reliability of that information.

What is the danger of AI search engines citing manufactured software reviews?

The primary danger is the degradation of truth and the promotion of potentially harmful products. When an AI search engine cites a manufactured source, it validates biased or incorrect software recommendations as objective facts. This can lead users to download software that contains malware, lacks promised features, or utilizes predatory subscription models, all while the user believes they are following a “verified” expert recommendation. Furthermore, it creates a financial incentive for bad actors to flood the internet with low-quality AI content, eventually drowning out legitimate journalism and expert analysis.

We are entering an era where “proof” is a relative term. If an AI can be tricked into citing a fake source, and then that AI’s answer is used as a source for another AI, we lose the thread of reality. It’s a digital version of the “telephone game,” but with much higher stakes for the economy and personal security. Ensuring that your software choices are based on verified, human-tested data is no longer just a good idea; it is a necessary defense against a fragmenting internet.

Steps to protect yourself from AI misinformation

Until AI search engines get better at filtering out these manufactured sources, the burden of proof falls on you, the user. Don’t take a single AI summary at face value. If an AI recommends a piece of software, click the citation link. Does the website look like a real publication? Does the author have a history of writing about technology? If the site is filled with generic stock photos and broken English, close the tab.

Another tactic is to use “traditional” search alongside AI. Go to a trusted forum like Reddit or a dedicated tech news site. Look for consensus among actual users, not just “content” pages. Often, a quick search for “Software Name + Reddit” will give you a much more honest picture of a tool’s flaws than any AI-generated pros and cons list ever will.

Finally, support human-led journalism. The reason these AI content farms are thriving is that the business model for real, deep-dive reporting is under threat. When we choose to read and share content from real experts, we help maintain the infrastructure of truth that AI models need to function correctly in the first place.

The discovery of manufactured sources behind AI recommendations is a wake-up call for anyone who relies on technology to make informed decisions. As we move deeper into a world where artificial intelligence mediates our access to information, we must remain vigilant about the quality of the data we consume. By understanding how these content farms operate and taking steps to verify citations manually, we can navigate the digital landscape without falling for the synthetic illusions of the SEO war.