Is Perplexity accurate?
Perplexity was the best citer in two independent news tests, but it repeated false news claims 46.67% of the time in NewsGuard's August 2025 audit, second worst of 10.
Short answer
Perplexity is the most accurate AI search tool for citing news in published tests, but not for avoiding false stories. The Tow Center found it wrong 37% of the time, the best of eight tools. NewsGuard found it repeated false news claims 46.67% of the time.
Results
- Independent study
- News audit
- Vendor claim
| Test | What was measured | Result | Date | Source |
|---|---|---|---|---|
| EBU/BBC News Integrity in AI Assistants | Share of Perplexity news answers with at least one significant issue (681 responses) | 30% | Source 1Independent study | |
| Tow Center AI search citation test | Wrong answers when free Perplexity was asked to name the source of a news excerpt (200 queries) | 37% | Source 2Independent study | |
| NewsGuard AI False Claims Monitor | Rate of repeating provably false news claims | 46.67% | Source 3News audit | |
| SimpleQA (with web search)Vendor's own figure, not independently checked | Accuracy of Perplexity Deep Research on short factual questions, run by Perplexity | 93.9% | Source 4Vendor claim |
What these results mean
Perplexity's design helps it on sourcing. It searches the web and shows links next to its answer. In the two independent tests that judged sourcing, it came first. The EBU/BBC study had journalists review news answers, and Perplexity had the lowest rate of significant issues of four assistants, though still nearly one in three (EBU and BBC, 2025-10-21). The Tow Center asked eight AI search tools to identify the source of 200 news excerpts, and Perplexity had the lowest error rate (CJR, 2025-03-06).
The same web search hurts it on false claims. In August 2024, Perplexity scored 0% in NewsGuard's audit, a perfect result. A year later it repeated false news claims 46.67% of the time, second worst of 10 chatbots (NewsGuard, 2025-09-04). NewsGuard found that live web search made chatbots answer more and repeat more falsehoods, because they pull in whatever the web is saying.
Perplexity's own figure is a vendor claim. It reported 93.9% accuracy on SimpleQA for Deep Research, with web search, in its own run (Perplexity, 2025-02-14). That score is not comparable to scores from tests run without browsing.
A cited answer is only as good as the page it cites, and Perplexity will cite a bad page.
Where Perplexity goes wrong
Repeating what the web says
When false stories circulate online, Perplexity can find and repeat them. Its NewsGuard rate rose from 0% in August 2024 to 46.67% in August 2025.
Citations that do not support the claim
Being the best citer still meant 37% wrong answers in the Tow Center test. The collective error rate across the eight tools was over 60%.
One in three news answers flawed
Even with the best EBU/BBC score, 30% of Perplexity's news answers had at least one significant issue.
Paid tiers are not safer by default
The Tow Center tested both Perplexity and Perplexity Pro, and found premium tiers gave more confidently wrong answers than free ones.
The common thread is that Perplexity trusts the pages it finds. When the web is right, its answers are well sourced. When a false story is spreading, Perplexity can cite a weak page with the same confidence as a strong one.
How it compares
Each table uses one test only.
Tow Center AI search citation test (2025-03-06): wrong answers when naming a news source
| Tool | Result |
|---|---|
| Perplexity (free) | 37% |
| ChatGPT Search | 134 of 200 (67%) |
| Grok 3 | 94% |
Source: CJR, Tow Center. Perplexity had the lowest error rate of eight tools. See ChatGPT and Grok.
EBU/BBC News Integrity in AI Assistants (2025-10-21): news answers with a significant issue
| Assistant | Result |
|---|---|
| PerplexityThis page | 30% |
 ChatGPT | 36% |
 Microsoft Copilot | 37% |
 Google Gemini | 76% |
Source: EBU and BBC. Perplexity placed first of four.
NewsGuard AI False Claims Monitor (2025-09-04): rate of repeating false news claims
| Chatbot | Result |
|---|---|
 Claude | 10% |
| ChatGPT | 40% |
| PerplexityThis page | 46.67% |
| Inflection Pi | 56.67% |
Source: NewsGuard. Perplexity placed ninth of 10. Claude placed first. The overall order is on the most accurate AI assistants.
How to check Perplexity's answers
Click every numbered citation
Confirm the page exists and contains the claim, not just the topic.
Look at who the source is
A citation to an unknown site or a forum post is weaker than one to the original publisher or record.
Search for a debunk
For any surprising news claim, search the claim plus "fact check". Perplexity may have found the rumour before the correction.
Check quotes and numbers in the original
Summaries change wording and round figures.
Do not treat more citations as more proof
Several links can repeat one bad original.
The full routine is on how to fact-check an AI answer. How sources are weighed is on how we score AI accuracy.
Sources
- Independent study
- Independent study
- News audit
- Vendor claim
Questions
Is Perplexity always right?
No. Perplexity is the strongest of the tested AI tools at citing news, but it is still often wrong. In the EBU/BBC study of 2025-10-21, 30% of its news answers had a significant issue. In NewsGuard's August 2025 audit it repeated false news claims 46.67% of the time, second worst of 10 chatbots. Open its citations before relying on an answer.
Is Perplexity reliable for research?
Perplexity is one of the better AI tools for finding sources, which makes it a good research starting point. The Tow Center found it named the wrong source 37% of the time, the lowest rate of eight AI search tools in March 2025. That still means checking every citation. Cite the original pages, not Perplexity's summary of them.
Does Perplexity make up sources?
Perplexity mostly links to real pages, but the page does not always support the claim. In the Tow Center test it gave wrong answers to 37% of source queries, the best of eight tools, while the group as a whole was wrong over 60% of the time. Click each citation and confirm the exact claim appears there.
Is Perplexity Pro more accurate?
Not reliably. The Tow Center's March 2025 test included Perplexity Pro and found premium tiers gave more confidently wrong answers than free versions. Perplexity reports 93.9% accuracy on SimpleQA for Deep Research, but that is its own run, with web search. Check paid answers exactly as you would check free ones.
How do I check a Perplexity answer?
Click every numbered citation and confirm the page contains the exact claim. Check who published each source and prefer the original publisher. Search the claim plus "fact check" to catch rumours Perplexity may have picked up. Compare quotes and figures with the original. Several citations that trace back to one bad source still count as one.
