Does Perplexity Send Traffic to Publishers? What the Data Shows

Yes, Perplexity sends referral traffic to publishers, and unlike ChatGPT it is reliably trackable in standard analytics because Perplexity consistently passes a perplexity.ai referrer in browser sessions. The catch is that the volume is small in both absolute and relative terms, and it does not come close to replacing the search clicks that AI is eroding. AI search engines as a whole send roughly 96% less referral traffic to news sites and blogs than traditional Google search, and the deeper problem is the imbalance between how much Perplexity crawls and how little it refers back. This article walks through what the data actually shows, how to find Perplexity referrals in your own GA4, how it compares to ChatGPT, and whether any of it is worth optimising for.

Does Perplexity send traffic to publishers?

Yes. When a Perplexity answer cites a source and a user clicks through to that source, the resulting visit lands on the publisher's site with a perplexity.ai referrer attached. That makes it a genuine, measurable referral channel. Perplexity is, in fact, the cleanest AI referral source to track, because it passes that referrer consistently in browser sessions rather than stripping it the way many AI tools do.

The problem is scale. AI-driven referral traffic is a rounding error for most sites today. AI platforms drive only about 1% of overall web traffic on average across ten major industries, according to Conductor data reported by Digiday. Separately, the TollBit analysis of 160 news, tech and shopping sites found that AI search engines sent roughly 96% less referral traffic than traditional Google search in Q4 2024. Within the slice of AI referral traffic that does exist, ChatGPT dominates and Perplexity is a smaller share.

Growth off that small base has been rapid, though. Per Similarweb estimates, Forbes received 236,300 visits from Perplexity in August 2024, up from 10,800 a year earlier; The New York Times went from 11,600 to 96,600, and The Guardian from 6,700 to 76,800 over the same period. Those are large multiples on tiny starting numbers. It is worth being clear that these are modelled Similarweb estimates, not first-party analytics, so they should be treated as directional. For any individual publisher, the first-party perplexity.ai / referral figure in your own GA4 is the ground truth.

The honest read: Perplexity referral traffic is real and worth watching, but it is currently modest and is not a replacement for the search clicks AI is taking away.

How do I find Perplexity referrals in GA4?

This is the easy part, and it is a big reason Perplexity is worth tracking at all. Because Perplexity passes a perplexity.ai referrer, its visits appear natively in GA4 without any tracking script, UTM tag or custom setup. To find them:

1. Open Reports > Acquisition > Traffic acquisition.
2. Change the primary dimension to Session source / medium.
3. Search or filter for perplexity.ai / referral.

That single row is your Perplexity referral traffic. The how-to from Rankshift confirms Perplexity is the most reliably trackable AI source in GA4 for exactly this reason.

To monitor all AI referrers together over time, build a custom channel group with a regex that matches the major AI hostnames, for example:

perplexity\.ai|chatgpt\.com|chat\.openai\.com|claude\.ai|gemini\.google\.com|copilot\.microsoft\.com

That isolates AI referrals into a single reportable channel so you can watch the trend rather than hunting through source and medium rows each time.

One caveat worth knowing: Perplexity's Comet browser and its native app may attribute in-app visits differently, often as direct / none rather than a clean referral. So the perplexity.ai / referral row is a floor on Perplexity-driven visits, not necessarily the full picture. Even so, it is the most trustworthy number you have, because it comes from your own analytics rather than a third-party model.

If you want a running read of AI referrals across engines without building regex in GA4 by hand, that is exactly the kind of view a clean GA4 dashboard surfaces for you. Try Ramprt free on the live demo to see AI referrals, AI crawlers and citations pulled from your existing GA4 with no new tracking.

Does Perplexity link out or just summarise?

Perplexity does link out. Its answers carry inline citations, and clicking one of those citations is what generates the referral visit you see in GA4. In that sense it behaves more like a search engine with an answer layer than a pure summariser that keeps users on-page with no way out.

But the balance between crawling and referring is heavily skewed, and this is the number publishers should really care about. The relevant metric is the crawl-to-refer ratio: the number of HTML pages a platform crawls for every referral visit it sends back. Cloudflare measures it as HTML requests from a platform's crawler user-agents against HTML requests whose referrer header points back to that platform. A high ratio means heavy crawling and little traffic in return.

Cloudflare's network-level data shows Perplexity's ratio getting worse through 2025. It rose from about 54:1 in January 2025 to about 195:1 in July 2025, a 256.7% increase. In plain terms, by July Perplexity was crawling roughly 195 pages for every single visitor it sent back. For comparison, here is how the major platforms stacked up over the same window:

Two things stand out. First, Google's search crawler remains in a completely different league, taking only around 4 to 5 pages per referral, which is the ratio you would expect from a channel that actually sends traffic. Second, Perplexity's ratio moving the wrong way, from 54:1 to 195:1, means it referred proportionally less over the year, not more. So Perplexity does link out, but it takes a great deal of content for each click it returns.

The training-focused crawlers make even that look generous. Cloudflare's Radar data found that in a single week, 19 to 26 June 2025, Anthropic's Claude made nearly 71,000 HTML page requests for every referral, a ratio of 70,900:1. The broader TollBit report captured the same dynamic from the publisher side, finding that AI companies scraped the 160 analysed sites around 2 million times on average in Q4 2024, at roughly seven scrapes per page.

What about Perplexity's two crawlers?

Perplexity runs two distinct declared crawlers, and the difference matters if you are trying to control access:

• PerplexityBot indexes content for Perplexity's search results. Per Perplexity's own documentation it respects robots.txt and is not used for model training. Its user-agent contains the token PerplexityBot/1.0.
• Perplexity-User fetches a live page in response to a specific user question. By Perplexity's documentation it generally does not follow robots.txt, on the basis that a human initiated the request. Its user-agent contains the token Perplexity-User/1.0.

The practical implication is that a robots.txt disallow will not stop the user-initiated fetch, and, as covered below, it may not reliably stop the rest either.

How does Perplexity traffic compare to ChatGPT traffic?

ChatGPT is the bigger referrer by a wide margin. It accounts for roughly 87% of AI referral traffic across ten key industries, per Conductor data reported by Digiday, leaving Perplexity and everyone else to share the remainder. So if you are ranking AI referral sources by raw volume, ChatGPT sits at the top and Perplexity is a smaller, secondary source.

The comparison flips when you look at trackability. ChatGPT referral traffic is notoriously messy in analytics because it often does not pass a clean, consistent referrer, so a meaningful share of ChatGPT-driven visits gets misattributed as direct traffic. Perplexity, by contrast, passes the perplexity.ai referrer reliably, which is why it is the AI source you can actually measure with confidence in GA4. In short: ChatGPT sends more, Perplexity is easier to see.

On the crawl-to-refer axis the two also differ. From the Cloudflare figures above, OpenAI's ratio was around 1,217:1 in January and 1,091:1 in July 2025, far heavier than Perplexity's 54:1 to 195:1. OpenAI crawls proportionally much more per referral, though its absolute referral volume is still higher because its user base is so large. The table below sums up the trade-offs.

If you want a fuller breakdown of the ChatGPT side of this, see Does ChatGPT send traffic to publishers?

Is Perplexity referral traffic worth optimising for?

Be realistic about the size of the prize. On the current data, Perplexity referral traffic is real, measurable and modest. It is not a channel you can build a business on, and it is not a substitute for the organic search clicks that AI is eroding. Chasing it hard, at the expense of the fundamentals, would be a mistake.

The bigger picture is what makes this urgent, and it is happening on the Google side too. Google AI Overviews are compounding the click loss. Pew Research found that users clicked a traditional search-result link on 8% of visits when an AI summary appeared, versus 15% without one, and clicked a source link inside the AI summary itself on just 1% of visits. Sessions ended with no further click on 26% of pages with an AI summary, versus 16% without. That study drew on 68,879 unique searches from 900 US adults. The sting is that this loss is largely invisible in Search Console, which has no dedicated AI Overview segment, so it usually surfaces only as impressions holding steady or rising while clicks and click-through rate fall.

So the sensible posture is to treat Perplexity as a signal to monitor, not a growth channel to over-invest in, and to do three things:

• Measure it as a signal. Watch the perplexity.ai / referral row and your wider AI referral channel over time. If it grows, good, but judge it against your total traffic, not in isolation.
• Watch the crawl-to-refer imbalance. If a platform is taking a great deal of content and returning very little, that is a commercial fact you should factor into any access decisions.
• Enforce crawler controls properly. If you decide to limit access, do it at the WAF by verified IP range and behaviour, not robots.txt alone (see next section).

There is also a nascent revenue answer taking shape. Perplexity announced Comet Plus in late August 2025, a $5 per month subscription tier funded by an initial $42.5 million pool. Perplexity keeps 20% and pays 80% to participating publishers, splitting revenue across three engagement types: direct visits, citations in on-screen answers, and content used inside AI agent tasks. Launch partners included CNN, Condé Nast, The Washington Post and the Los Angeles Times. This arrives against a backdrop of copyright lawsuits from Dow Jones and the New York Post, and separately The New York Times, among others, so whether it becomes a meaningful line item or a goodwill gesture is still an open question.

Why can't I just block Perplexity in robots.txt?

Because robots.txt alone is not a reliable block. Two mechanisms undermine it. First, as noted above, the Perplexity-User agent generally does not follow robots.txt, on the basis that a human initiated the request. Second, and more seriously, Cloudflare accused Perplexity in August 2025 of using stealth, undeclared crawlers to reach content on sites that had explicitly blocked its declared bots via robots.txt and WAF rules. Cloudflare removed Perplexity from its list of verified bots as a result.

Cloudflare's investigation described an undeclared crawler using a generic Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 ... Chrome/124.0.0.0 Safari/537.36 user-agent, making 3 to 6 million daily requests from IPs outside Perplexity's published ranges, alongside the declared Perplexity-User agent at 20 to 25 million daily requests. In tests, that stealth traffic reached domains that had explicitly blocked Perplexity, and the tool still returned detailed information about content it should not have been able to access.

The takeaway for publishers is that enforcement should not trust the user-agent string on its own. Perplexity publishes IP-range verification data and recommends combining user-agent matching with source-IP verification at the WAF, whether Cloudflare or AWS. Practically, that means:

• Match on user-agent and verify the request comes from a published Perplexity IP range.
• Use WAF rules and behavioural detection rather than robots.txt as your control layer.
• Accept that a determined, undeclared crawler may still get through, and monitor your logs for anomalous crawl patterns.

None of this is a reason to panic. It is a reason to treat crawler control as a security and infrastructure problem, not a text file you edit once and forget.

The bottom line

Perplexity does send traffic to publishers, that traffic is the most trackable of any AI source thanks to the perplexity.ai referrer, and it is currently small. The number that should shape your thinking is not the referral count on its own but the crawl-to-refer ratio, which for Perplexity worsened from about 54:1 to about 195:1 over 2025 while Google's search crawler stayed near 5:1. Measure the perplexity.ai / referral row in GA4, watch it in the context of the wider AI-versus-search click erosion that Pew and TollBit have documented, and enforce any crawler controls by verified IP range rather than robots.txt alone. Perplexity is a signal worth watching. It is not, yet, a channel worth betting on.