Do backlinks still matter for visibility in Perplexity?

What does the research say about backlinks and Perplexity?

One study measured it directly, and links came out as the strongest single signal. Sharma (opens in a new tab) picked 112 startups from the top 500 on the 2025 Product Hunt leaderboard and ran 2,240 queries to Perplexity and to an OpenAI model without web search. The queries ran through each company’s developer connection between December 15 and 20, 2025.

Sharma then compared each startup’s visibility with its link data, Product Hunt results and community presence. For Perplexity, referring domains had the strongest link with visibility across the full sample, at 0.319. The share of links that pass ranking credit, known as followed links, also mattered, at 0.238.

The scale runs from 0 (no link) to 1 (a perfect one), so these are moderate links, not strong ones. The Reddit figure is higher, but it rests on fewer products: Sharma removed 46% of the sample because their generic names, such as “Cursor”, matched unrelated Reddit posts.

Why would links matter to an AI engine?

Because Perplexity searches the web as it answers, and search results reward links. Sharma’s explanation is simple: the signals that help a page rank in traditional search also make it more likely to be found and cited by an engine that searches live.

The contrast with a model that does not search supports this. Perplexity named the startups in 8.29% of discovery questions, and it named 31 of the 112 at least once (27.7%). For the OpenAI model without search, Sharma found no signal that predicted visibility at all. Perplexity showed seven signals with a clear link, which suggests an engine that searches live gives marketers levers that a model answering from memory does not. We look at how much assistants still answer from training data in a separate guide.

Our own data links Perplexity closely to Google rankings. In our study of AI citations and Google rankings, Perplexity cited 33.0% of the pages in Google’s top 10 for each question, far more than ChatGPT, Gemini or Claude. It cited 40.8% of pages ranked 1 to 3 and 28.8% of pages ranked 4 to 10.

Does ranking explain everything Perplexity cites?

No. Most of what Perplexity cites does not rank on Google for the question at all. In the same study, 66.8% of Perplexity’s citations were not in Google’s top 100 for the question. Perplexity lists many sources, a median of 20 per answer, and its first three citations ranked far more often than the rest, 23.8% against 14.2% overall. The wider gap is covered in how AI sources differ from Google’s.

Perplexity also reaches pages by routes other than its own crawler. In our crawler blocking study, 40.4% of the pages Perplexity cited on top sites were closed to PerplexityBot by the site’s own robots.txt file.

Third-party coverage carries a lot of weight. Chen and colleagues (opens in a new tab) found 67.4% of Perplexity’s sources about well-known brands came from earned media, independent publications and reviews. In a soft drink test, 67.9% of Perplexity’s brand mentions were major brands. Coverage on other sites and links from them often come together, so this fits with the link finding without proving it.

How reliable is a single Perplexity check?

Not very. Perplexity’s sources change from run to run, so one check can mislead. Sielinski (opens in a new tab), who works for an AI visibility measurement company, ran the same queries repeatedly on Perplexity, OpenAI’s search model and Gemini over nine days. Repeated Perplexity runs shared only about half of their cited domains, an overlap of 0.50.

Perplexity also cites more sources than its rivals in other data. In a public dataset of 602 test prompts analyzed by Zhang, He and Yao (opens in a new tab), Perplexity cited 16.35 sources per prompt on average, against 6.88 for ChatGPT. The authors argue that classic signals such as backlinks help a page enter the pool of sources, but do not show whether the answer actually uses it. Signals on the page itself are weighed in which on-page signals track with AI citations.

What should you do about it?

Keep building links and third-party coverage, and treat them as an input to Perplexity rather than a guarantee. In practice:

1. Earn links from relevant, reputable websites. In the one direct test, referring domains were the strongest signal for Perplexity visibility.
2. Pursue coverage in publications and review sites. Perplexity draws most of its sources from that kind of earned media.
3. Encourage genuine discussion in relevant communities such as Reddit, not planted posts.
4. Rank for the narrower questions buyers ask, since Perplexity cites top-ranked pages more often.
5. Check that robots.txt and your firewall do not block Perplexity’s crawlers by accident.
6. Measure visibility over many runs and weeks, not once.

If you want help prioritizing these, our generative engine optimization service starts with that measurement.

What does the research not tell us yet?

The link between backlinks and Perplexity rests on thin evidence.

• Only one study measures backlinks against Perplexity visibility directly: 112 startups, mostly developer, productivity and AI tools, queried over one week.
• It shows a link, not a cause. Better-known products tend to have more links, more coverage and more discussion all at once.
• The Reddit result rests on 60 products after cleaning, and the study tested Perplexity through its developer connection rather than the consumer app.
• No study has added links to a site and then measured the change in Perplexity answers.
• The findings may not carry over to established brands, other industries, or other AI engines such as ChatGPT with search.

Frequently asked questions

Do backlinks help you appear in Perplexity answers?

Probably. In a December 2025 test of 112 startups, referring domains were the strongest predictor of Perplexity visibility, with a moderate link of 0.319, though no study has shown cause.

Does Perplexity use Google rankings?

It overlaps with them more than other assistants but does not copy them. Perplexity cited 33.0% of Google’s top-10 pages in our test, yet 66.8% of its citations were outside Google’s top 100 for the question.

Are backlinks more useful than AI-optimized content for Perplexity?

In the one study that compared them, yes. Referring domains showed a link with Perplexity visibility, while an AI-friendly content score showed none (−0.102).

How often does Perplexity recommend a new startup?

Rarely. Perplexity named the startups in 8.29% of discovery questions in Sharma’s test.

Sources

• Sharma (2026), The Discovery Gap: How Product Hunt Startups Vanish in LLM Organic Discovery Queries (opens in a new tab), arXiv:2601.00912.
• Chen, Wang, Chen and Koudas (2025), Generative Engine Optimization: How to Dominate AI Search (opens in a new tab), arXiv:2509.08919.
• Sielinski (2026), Quantifying Uncertainty in AI Visibility: A Statistical Framework for Generative Search Measurement (opens in a new tab), arXiv:2603.08924.
• Zhang, He and Yao (2026), From Citation Selection to Citation Absorption: A Measurement Framework for Generative Engine Optimization Across AI Search Platforms (opens in a new tab), arXiv:2604.25707.
• Underneath (2026), Do ChatGPT, Gemini, Perplexity and Claude cite pages that rank?
• Underneath (2026), Which AI crawlers do top websites block? 10,000 sites, 2026