How to Fix the “Perplexity Wrong URLs” Problem: A Practical Guide

Many users have recently noticed a strange issue with Perplexity. When the model is asked to display full URLs in plain text, the links it generates are often completely wrong. Some are partially correct, but most appear to be entirely fabricated. This behavior only started happening recently and was not present just a couple of weeks ago. If you have run into this, you are definitely not alone.

This guide explains why it happens, how to confirm the issue, and what you can do to work around it while waiting for an official fix.

What Causes Perplexity to Produce Wrong URLs

Perplexity normally does a good job pulling accurate URLs from its research results. The recent shift toward incorrect or spoofed URLs likely comes from internal changes. Several technical factors may be involved:

• A backend update affecting how URLs are passed through the model • Problems with plain text parsing and URL tokenization • Temporary issues in how research mode surfaces its sources • A safety or filtering layer replacing real URLs with fabricated ones • A bug introduced after a recent system update

These issues tend to appear specifically when the model is asked to output URLs in plain text rather than in tag format.

When the Problem Happens

Users commonly encounter the issue in the following situations:

• Asking Perplexity to show the full URL in plain text • Requesting the exact search result links from research mode • Using custom prompts to force the model to embed URLs directly in summaries

The unusual part is that less than two weeks ago, this workflow still produced correct URLs. The shift happened suddenly, which suggests a backend update rather than user error.

It is also important to note that:

• URLs in tag format remain correct • Only plain text links are affected

This difference helps isolate the issue and gives us a path to work around it.

How to Verify the Issue Yourself

If you want to confirm that you are experiencing the same problem, try the following:

• Repeat the same prompt several times and check if the link is consistently incorrect • Compare results between research mode and quick mode • Look at older chat logs to see when the behavior changed • Ask the model to display the same URL in both plain text and tag format

If the tag version is correct while the plain text version is wrong, you’re seeing the same bug many users are reporting.

Workarounds You Can Use Right Now

While waiting for Perplexity to fix the underlying issue, you can still obtain reliable links by adjusting how you request them.

1. Ask for URLs in tag format

This is currently the most reliable method. Tag-style output still returns correct URLs.

2. Force a structured JSON output

A structured format reduces the chances of hallucinated text. However, testing shows that even with JSON structured output, Perplexity still returns incorrect URLs in text format. This method alone is not sufficient to prevent wrong links.

Example:

Return the results in this JSON format:
{ "title": "", "summary": "", "url": "" }

3. Ask for quoted source excerpts rather than URLs

This method has been tested and works. By instructing Perplexity to quote the original text rather than generating a URL, the problem of fabricated links is avoided. This is currently the most effective approach for accurate references in plain text output.

4. Cross-check using another tool

If accuracy is critical, double-check using another search tool such as Google or Bing, or use Perplexity Pro if available.

Using remio as a Practical Alternative

Lately, I’ve been using remio as a backup when Perplexity gives me wrong URLs. Honestly, it’s been a lifesaver for keeping track of sources. What I like about it is that it can automatically collect information from webpages, PDFs, Word files, Slack chats, and Gmail, so you don’t have to hunt down the original content yourself.

A few things that worked well for me:

• Citations you can trust: remio can quote the original text directly, so I don’t have to worry about fake or broken links.

• Flexible prompts: You can set your own instructions for how it summarizes or structures content, which helps me get exactly what I need.

• Keeps everything organized: It’s like having a second brain—my research notes, transcripts, and references are all in one place.

• Privacy-focused: I don’t have to worry about my personal data being misused.

• Accurate URL retrieval: remio can also search the web directly and return real, correct URLs, which makes it even more reliable than just relying on collected documents.

For me, using remio alongside Perplexity made a huge difference. When Perplexity messes up the plain text URLs, I just check the same content in remio. It’s simple, reliable, and saves a lot of frustration.

Prompt Templates to Improve URL Accuracy

Here are three prompts that increase your chances of getting correct URLs.

Template 1: Structured JSON Output

Provide each source with title, summary, and full URL in clear JSON format. Use only real URLs from your research results.

Note: JSON output alone may still produce wrong URLs, so always cross-check or combine with quoted text.

Template 2: Tag Format Output

Show all source URLs using tag-style links to ensure accuracy.

Template 3: Quote Original Text Instead of URL

Provide a summary of each source and quote the original text instead of showing the full URL.

This template has been confirmed effective in tests.

When You Should Contact Support

If any of the following happens, consider reaching out to Perplexity’s support team:

• Tag-format URLs also become incorrect • All modes of the model produce hallucinated URLs • The issue persists for several weeks without improvement • Incorrect URLs impact your professional or research workflow

Clear feedback helps them diagnose and prioritize the fix.

Conclusion

The “Perplexity Wrong URLs” issue appears to stem from recent backend changes that affect how the system outputs plain text links. Although the problem is disruptive, the tool still generates accurate URLs when using tag formatting or quoted original text. Until Perplexity provides a fix, these workarounds offer a reliable path to continue your research with minimal interruption.

Testing confirms that forcing JSON structured output alone is insufficient, while quoting original text instead of URLs works effectively. Using Remio as a complementary tool further improves reliability, especially for managing research and citations. The fact that Remio can also search online and directly return accurate URLs makes it a practical solution for anyone who needs reliable links.