Can We Trust AI-Generated 'Evidence'? The Betrayal of Perplexity Citations

Imagine this: for a presentation this evening, you ask an AI search engine, “What is the AI market growth rate in our country this year?” The AI immediately provides an answer and kindly attaches sources like [1] and [2] at the end of the sentence. We usually feel reassured when we see these sources, thinking, “The AI found and verified this information itself.” But what if those sources were actually pointing to the wrong pages?

Recently, a shocking reality regarding the citations provided by the AI search service Perplexity as evidence for its answers has been exposed. Let’s take a look at how accurate these ‘sources’ we’ve trusted really are, and why AI makes such mistakes.

Why Does It Matter?

Unlike traditional search engines, Perplexity summarizes vast amounts of web data to generate answers itself. This allows users to get answers all at once without having to click through multiple sites individually. Source: Perplexity is a citation engine. In fact, the rate at which users click on citations (the sources indicated by numbers) reaches 18–24%, which is much higher than the 2–4% click-through rate of traditional search engines. Source: How to get cited by Perplexity in 2026.

In other words, we highly trust the sources provided by AI and actually use them to delve deeper into information. However, if this information does not contain the facts, we are at risk of falling into a swamp of fake information.

Understanding It Simply

To put it simply, Perplexity’s operation is similar to a ‘smart assistant searching through numerous books and summarizing them for you.’ While writing an answer, the assistant footnotes it, saying, “This information is on page 5.” However, sometimes the assistant writes the whole text first and then adds the footnotes later, thinking, “Ah, this part must have been around page 5.” Source: Perplexity citation patterns. In this process, the assistant’s memory becomes blurry, leading it to point to the wrong page.

Data analysis revealed that about 34.7% of citations attached to sentences containing figures were linked to pages that did not contain those figures at all. Source: Perplexity citation audit report. Metaphorically, this is like trying to check the answer key at the back of a textbook after solving a math problem, only to find that the answer key contains solutions for completely different problems. Furthermore, when evaluated overall, results showed that about 14.4% of the claims made by Perplexity were not supported by the actual cited sources. Source: Perplexity citation audit report.

Current Situation

Perplexity is so active in citing sources that it specifies them in about 94% of its answers. Source: In 2026, does Perplexity always specify sources?. The problem, however, lies in the way the AI model itself generates an answer and then ‘post-facto’ matches sources to it without verifying if the answer is factual. Source: Perplexity citation patterns.

Of course, sometimes it’s not Perplexity’s fault. There are instances where external apps fail to display Perplexity’s data properly, making it look like the source link has disappeared. Source: Perplexity source non-citation issue. Fundamentally, however, the ‘hallucination’ phenomenon—where the system retrieves sources that do not match the answer content—clearly exists, and this is a limitation that users must be aware of. Source: 2026 Perplexity Review.

What Will Happen Next?

In the future, the standard for competition among AI search services will shift from ‘how many sources are shown’ to ‘how accurately are the sources connected.’ Some research has already pointed out that Perplexity’s citations are about three times more numerous than ChatGPT’s, noting that quantitative expansion does not always guarantee qualitative accuracy. Source: 9 signals of Perplexity citations. As users become smarter, AI platforms that provide incorrect citations will lose trust.

MindTickleBytes’ AI Reporter Perspective

AI search engines are convenient, but one must be wary of baseless confidence. If the content you want is missing when you click the source provided by an AI, it is highly likely that the AI did not understand the content deeply but merely guessed a ‘plausible location.’ In this era, it is necessary to cultivate the habit of verifying content one more time with a ‘critical perspective’ when reading searched answers.

References

1. AthirdofPerplexity’scitationsdon’tcontainthenumberthey’r…

2. [How to GetCitedbyPerplexity: The Tactical Playbook for 2026

   Cintra](https://cintra.run/blog/how-to-get-cited-by-perplexity)

3. [How to Rank inPerplexityAI: What 21CitationsPer Query…

   BlueJar](https://bluejar.ai/blog/how-to-rank-in-perplexity-ai/)

4. [How to GetCitedbyPerplexityAI

   Mentionable](https://mentionable.ai/en/guides/rank-on-perplexity)

5. PerplexityInlineCitations: How [1][2][3] Links Work
6. PerplexitySEO: How to GetCitedin 2026

7. [How to GetCitedbyPerplexity(2026 Playbook)

   MentionAgent](https://mentionagent.ai/blog/how-to-get-cited-by-perplexity/)

8. The 50 Most-CitedWebsites inPerplexity(September 2026)

9. [PerplexityCitations

   Fetchable Sources, Enquire Desk](https://www.worldwidebacklinks.com/ai-backlinks/perplexity-citations/)

10. PerplexitycitesClickUp 6,474 times. Notion gets 741… Why?
11. PerplexityCitationPatterns: What Actually Gets Sourced — b/cited
12. How to earn morecitationsinperplexityai search

13. [How to GetCitedbyPerplexity: 9 Source Signals

    CiteVantage](https://citevantage.com/blog/how-to-get-cited-by-perplexity/)

14. A third of Perplexity’s citations don’t contain the number they’re …
15. Perplexity Not Citing Sources: 8 Fixes 2026
16. Perplexity AI Review 2026: Citations, Limits & Real Failures
17. Does Perplexity Always Cite Sources? 2026 Data Says No
18. How Perplexity Selects Its Citations: What We Know From Testing and …
19. Getting Cited by Perplexity: What It Actually Quotes — Genαi
20. How Perplexity Decides Which Sources to Cite - authoritytech.io