Perplexity AI is facing fresh scrutiny over how its systems collect and reuse online content. According to WIRED, Amazon Web Services has opened an investigation into whether the startup violated AWS terms of service by crawling websites that had explicitly tried to block that access.
The dispute sits at the intersection of AI search, publisher rights, cloud platform rules and automated web crawling. It also raises a practical question for media companies: what happens when a publisher says no to bots, but the content still appears to be accessed and reused?
What AWS is examining
WIRED reports that AWS is investigating whether Perplexity violated its terms of service. The central issue is whether Perplexity crawled websites and used their content despite clear prohibitions against that kind of access.
The concern is tied to the Robots Exclusion Protocol, a web standard that lets websites signal that automated bots should not access certain pages. The protocol is not legally binding, but WIRED reports that compliance with robots.txt is included in the terms of service of cloud providers such as Amazon.
That distinction matters. A robots.txt block may not by itself settle every legal question, but it can become important when a service provider’s own rules require respect for those signals. If a cloud customer is accused of ignoring those rules, the issue becomes not only a publisher complaint but also a platform compliance question.
Why publishers are objecting
WIRED says Perplexity’s bot regularly scraped its content even though Condé Nast, WIRED’s publisher, had explicitly blocked the bot. According to the magazine, Perplexity in some cases reproduced WIRED’s content verbatim.
The allegations are especially sensitive because they involve both access and output. One concern is whether blocked content was collected in the first place. Another is whether the resulting AI responses or pages present publisher material too closely to the original work.
The source article also states that Perplexity’s bot sometimes uses special URLs to access paywalled content. A Perplexity spokesperson acknowledged that this occurs, but said it is "very infrequent" and happens only when a user explicitly enters the URL in the prompt. The spokesperson also stated that Perplexity’s bot does not violate AWS’ terms of service.
That explanation leaves the dispute unresolved. Perplexity frames the behavior as limited and user-triggered. Publishers and investigators are focused on whether the underlying access and reuse respect the restrictions those publishers put in place.
Perplexity’s defense
Perplexity CEO Aravind Srinivas has defended the company by saying the criticized crawlers came from a third-party company that provides crawling and indexing services. He said it would be "complicated" to stop these services.
Srinivas declined to name the company Perplexity works with, citing a non-disclosure agreement. That defense shifts part of the attention to the infrastructure behind AI products: even when a startup is the public-facing service, outside vendors may be involved in collecting and indexing the web.
But the practical issue remains the same for publishers. If content is blocked, accessed, and then used in an AI product, the publisher may not care which technical partner performed which part of the process. The visible product is Perplexity, and the complaints are being directed at how Perplexity’s service operates.
Pages brought more attention
The controversy has widened since Perplexity launched "Pages." The product automatically aggregates content from multiple sources, compiles it into a landing page, and has that page indexed by Google to compete with original content.
Following the release of Pages, criticism intensified after Forbes revealed that Perplexity was plagiarizing its content. The source article says Srinivas previously told Forbes that his tool does nothing different from other news sites that quote journalistic primary sources.
That comparison is central to the argument. News organizations quote sources as part of reporting, but publishers are objecting to automated systems that collect, compile and reproduce their work in ways they see as competing with the original content. The release of Pages made that concern more concrete because it created indexed pages built from multiple sources.
Perplexity is reportedly planning to offer publishers a revenue share in the future. The source does not provide details on timing, terms or which publishers might be involved, but the reported plan suggests the company knows publisher relationships will remain important as AI search products expand.
The broader AI search problem
The dispute around Perplexity is not isolated. The source article notes that OpenAI’s ChatGPT faced similar criticism for bypassing paywalls, which led to a temporary shutdown to address the issue. OpenAI is now signing deals with media companies to display their fresh content on ChatGPT.
Perplexity, despite substantial funding and a very high valuation, may struggle to afford similar agreements, according to the source article. That creates a strategic challenge: AI products need current, reliable content, but publishers increasingly want control, credit and compensation.
For AWS, the immediate question is whether its terms of service were violated. For Perplexity, the larger challenge is trust. If publishers believe their blocks are ignored, their paywalls are bypassed, or their work is reproduced too closely, technical explanations may not be enough.
The outcome of the AWS investigation has not been stated in the source article. What is clear is that Perplexity AI is now under pressure from several directions at once: cloud platform scrutiny, publisher complaints, questions about robots.txt compliance, and criticism of products that turn sourced material into new indexed pages.