Perplexity reportedly failed a simple source-grounding test: when Wired asked it to summarize a webpage containing only “I am a reporter with Wired,” the system produced an unrelated fantasy about Amelia, a girl following glowing mushrooms through a magical forest called Whisper Woods. Futurism’s June 22, 2024 report says Wired’s logs showed that Perplexity never attempted to open the test page.
What the test asked Perplexity to do
Wired created a minimal webpage containing one sentence: “I am a reporter with Wired.” The request was to summarize that page. A faithful answer could have repeated or briefly paraphrased that sentence. Instead, Perplexity reportedly returned “a story about a young girl named Amelia who follows a trail of glowing mushrooms in a magical forest called Whisper Woods.”
As an Amazon Associate I earn from qualifying purchases.
| Test element | Reported detail | What a grounded answer would require |
|---|---|---|
| Source page | A page containing only “I am a reporter with Wired.” | The system must retrieve and read that page. |
| Requested task | Summarize the supplied webpage. | The response should stay within the page’s single factual sentence. |
| Reported response | A fantasy narrative involving Amelia, glowing mushrooms and Whisper Woods. | No invented characters, setting or plot. |
| Retrieval evidence | Futurism says Wired’s logs showed no attempt to visit the page. | Retrieval and answer claims should be independently traceable. |
Did Perplexity actually open the page?
According to Victor Tangermann’s report for Futurism, Wired’s logs indicated that Perplexity never attempted to visit the test webpage. That is a reported finding from the incident, not a general measurement of how Perplexity behaves on every query.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Scan for outdated or missing drivers - takes under a minute3Clear out junk files and repair common Windows errorsThe distinction matters because Perplexity describes its product as one that “searches the internet to give you an accessible, conversational, and verifiable answer.” If the reported logs are accurate, this answer was generated without retrieving the source the user had supplied. The result therefore was not merely a poor paraphrase; it was an answer detached from the requested evidence.
#1 Best Overall
Why the Amelia story is an AI hallucination
In this context, “hallucination” means that the system generated plausible-sounding content that was not supported by the source. The prose may sound coherent, but coherence is not evidence that the model found, read or understood the page.
Fluency concealed the failure
The invented narrative had specific names, imagery and a setting. Those details can make an answer feel researched even when none came from the supplied text. A detailed response is therefore not proof of browsing, citation accuracy or source comprehension.
Rank #2
The failure was testable
Because the page contained only one sentence, the mismatch was easy to establish. There was no ambiguous passage to interpret and no long document to summarize selectively. The central question was whether the answer reflected the source at all.
Windows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallOutdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchWhat this case does—and does not—prove
The incident demonstrates a source-grounding failure in one reported test. It does not establish how often this exact behavior occurs, whether every Perplexity answer has the same problem, or how competing answer engines would perform under identical conditions. The available report provides qualitative evidence, not an incident-rate statistic or a controlled product comparison.
Tangermann placed the episode alongside other reported concerns about AI-generated information, including an Associated Press finding involving fake quotes attributed to real people and Forbes reporting about weak attribution of published reporting. Those are related allegations discussed in the article, not findings independently established by this account.
The report also says Perplexity general counsel MariaRosa Cartolano used the phrase “willful infringement” in a letter obtained by Axios. That statement concerns a separate dispute and should not be treated as an explanation for the Amelia output.
How to verify an AI-generated webpage summary
Use the answer as a starting point, not as proof that the source was retrieved. These checks can expose an invented or distorted summary:
- Open the cited page yourself. Confirm that the URL resolves and read the relevant passage rather than relying on the assistant’s description.
- Compare every concrete detail. Check names, dates, figures, quotations and causal claims against the source. Details absent from the page require separate evidence.
- Ask for a verbatim quotation. Request the exact sentence supporting each major claim. A system that cannot provide a matching quotation may not have grounded its answer.
- Test a one-sentence source. Give the assistant a deliberately minimal page or passage and ask for a summary. Any new plot, character or fact is an immediate warning sign.
- Inspect citations individually. Follow each link and make sure it supports the sentence beside it. A citation that leads to a related topic is not support for the claim made.
- Separate retrieval from interpretation. Ask what the page says first, then ask for analysis. Keeping those tasks separate makes unsupported additions easier to spot.
- Seek confirmation for consequential claims. For legal, medical, financial or professional decisions, verify the underlying documents and consult an appropriate qualified source.
How to judge an AI search answer
Five checks are more useful than a confident tone:
- Source retrieval: Did the system actually fetch the supplied page?
- Faithfulness: Does each material statement follow from the source?
- Citation traceability: Can a reader reach the evidence and match it to the claim?
- Uncertainty disclosure: Does the answer say when access, evidence or interpretation is limited?
- Independent verification: Can a reader reproduce the conclusion without trusting the chatbot?
Perplexity’s reported performance in this test raises direct concerns about the first two checks. The report does not provide a controlled scorecard for the other products or establish a comparative ranking.
Best Value
Why this matters for everyday users
People often treat an answer engine’s polished explanation as evidence that it performed the search it describes. This case shows why that assumption is unsafe. A system can produce a specific, readable narrative while failing to access the only source it was asked to summarize.
The practical safeguard is simple: verify the source, quotations and citations before accepting the summary. When the source is short, compare it line by line. When the stakes are high, treat the chatbot as an index or drafting aid rather than the final authority.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
Recommended Free Tools
