Perplexity: what it is, how it works, and how it picks the sources it cites
Perplexity is the answer engine that always cites its sources. What it is, how to use it for free, how it assesses sites with Government, Academic, and Trusted labels, and what it actually cites: across 450 Italian-language answers measured with Searchable, the top domain is Quotidiano Motori, which still gets only 31 visits a month from Perplexity.
In brief
- Perplexity is an answer engine: it searches the web in real time and answers with numbered citations and links to its sources. The free plan includes practically unlimited basic searches.
- Since August 2026, some sources carry a label: Government, Academic, or Trusted. To assign it, Perplexity reviews the whole site: whether it corrects its mistakes, whether it credits its authors, and whether it keeps news separate from advertising.
- Across 450 Italian-language answers about cars and motorcycles, Perplexity uses 10.6 sources per answer and cites Quotidiano Motori more than any other site, in 30.9% of cases. In the same month, though, Perplexity sent 31 visits, compared with 334 from ChatGPT.
Perplexity is the AI assistant that always shows where its answers come from. On Italian-language questions about cars and motorcycles, it is also the one that most often cites Quotidiano Motori, the Italian online motoring daily we run. We track this every day with Searchable, and to understand what it means for a brand we put together Perplexity’s official pages and data from 450 answers: how it works, how it assesses sources, and how much traffic it actually sends.
Index
What Perplexity is
Perplexity is an answer engine: instead of a list of links, it gives a written answer, built by searching the web at the moment you ask. It interprets the question, searches the internet in real time, and pulls the information together into a summary, using models such as GPT-5 and Claude. Every answer has numbered citations linking to the original sources, and you can ask follow-up questions without repeating the context.
How to use it: free and paid plans
Perplexity can be used for free. The Standard plan offers “practically unlimited” basic searches, search history, and a very limited number of Pro searches, the ones that consult more sources and give fuller answers. Paid options include the Pro and Max plans, plus versions for businesses: Enterprise Pro starts at $40 a month per seat.
For longer tasks there is Deep Research, which according to Perplexity runs “dozens of searches,” reads “hundreds of sources,” and delivers a report in two to four minutes.
How it assesses sources
In general, Perplexity says it gathers information from “authoritative sources like articles, websites, and journals.” Since August 2026 it also shows how it assesses some sites: certain citations carry a shield icon and a Government, Academic, or Trusted label. Government applies to official public agency sites, Academic to scientific sites, and Trusted to sites that appear often in results and publish within their own area of expertise.
The review covers the whole site, not a single page, and starts from specific questions. Does the site correct its mistakes? Does it say who wrote each piece? Does it keep news separate from advertising and opinion? According to Perplexity, its “partnerships, payments, and other business arrangements” do not affect a site’s label.
Most sites have no label, and Perplexity stresses that this is not a negative judgment: the domain has not been reviewed yet, or there is no clear assessment. Labels help readers understand what kind of source they are looking at. Perplexity does not say that they decide which sites make it into answers.
Perplexity vs. ChatGPT: what’s the difference
Both search the web and cite their sources. Perplexity was built as an answer engine and puts citations in every answer; ChatGPT searches the web when it decides it is useful or when asked to. The difference you can measure is how many sources they use: across the questions about cars and motorcycles measured with Searchable, ChatGPT uses an average of 25.4 per answer, Perplexity 10.6.
What it actually cites: data from 450 answers
Over the last 30 days, Perplexity gave 450 answers to the automotive questions we track, with 4,790 sources. The domain that appears most often is Quotidiano Motori, in 30.9% of answers, with 218 citations across 23 different pages. Three news outlets follow, all between 20% and 30%, then institutional websites: ACI, Italy’s automobile club, in 16% of answers, and gov.it domains such as ENAC and the Ministry of Infrastructure and Transport, also in 16%. Wikipedia stops at 12.9%.
The same site, on the same questions, holds very different positions from one assistant to another: in Google AI Mode it ranks fifth, in 11.7% of answers. Perplexity does not explain why. The data do show, however, that visibility on one assistant does not carry over to the others, and it has to be measured on each one.
Cited does not mean visited
In the same month, Perplexity sent Quotidiano Motori 31 visits from people. ChatGPT, which cites the site far less often, sent 334. The difference in audience size between the two services plays a part, but for a brand the point stands: being cited and receiving visits are two different measures.
The bots tell a different story. In 30 days, PerplexityBot, the crawler used to surface and link sites in results, visited the site 17,848 times. Perplexity-User, which opens pages when a user asks a question, did so 1,025 times. Perplexity writes that this second crawler generally ignores robots.txt rules, because the visit is requested by a person.
What a brand can do
The questions Perplexity uses for its labels make a useful checklist for any site: correct mistakes visibly, credit content with the name of whoever wrote it, and keep news, advertising, and opinion separate. They do not guarantee a citation, but they are what Perplexity says it looks at when it reviews a site.
Then there is the technical side. To appear in results, Perplexity recommends allowing PerplexityBot in robots.txt and from the IP addresses it publishes. The firewall needs checking too: Cloudflare’s Bot Fight Mode, Perplexity warns, can still stop it even when a rule allows it.
Finally, it has to be measured: how often the brand is cited, from which pages, and how many visits actually arrive. That is the work of Generative Engine Optimization, and on Perplexity the gap between the two measures can be very wide.
Frequently asked questions
Yes. The Standard plan is free and offers practically unlimited basic searches, search history, and a very limited number of Pro searches. The paid plans, Pro, Max, and the Enterprise versions, add more Pro searches, the most powerful models, and features such as creating files and apps.
Both search the web and cite their sources, but Perplexity was built as an answer engine and puts numbered citations in every answer, while ChatGPT searches the web when it decides it is useful or when asked to. Across the Italian-language questions about cars and motorcycles measured with Searchable, Perplexity uses an average of 10.6 sources per answer and ChatGPT 25.4.
Perplexity recommends allowing PerplexityBot in robots.txt and from the IP addresses it publishes. For its source labels, it reviews the whole site with specific questions: whether it corrects its mistakes, whether it says who wrote each piece, and whether it keeps news, advertising, and opinion separate. According to Perplexity, payments and business arrangements do not affect the labels.
