Perplexity (Free, Pro and Max plans, iOS app 26.40.1)

This is a reading of Perplexity, the answer service that shows numbered sources beside every reply, on its Free, Pro and Max plans as the pricing page listed them on 13 October 2026, with the iPhone app at version 26.40.1. We read 11 sources that day. Six are review articles dated March to 9 October 2026, one is a study of news citations reported by the Columbia Journalism Review on 6 March 2025, one is a Hacker News thread from 2 September 2026, and three are Perplexity's own pricing and help pages and its App Store listing.

The owner tier is thin. We read no buyer reviews, only the App Store's headline figures, and one forum thread. The review articles are mostly from small review sites, and three of the six repeat the same 2025 study, so they are less independent than six reviews would suggest.

Consensus

Reviewers agree on the shape of it. Perplexity makes its sources easy to open, and they say the reader should open them. Four of the six review articles praise the citations as a way to check an answer, and five of the six tell readers to click through to the cited page before relying on it. Three of the six call it a good first pass or a source-finder.

Across the reviews we read, this is a reading of what writers say, not a measurement of how often a citation holds. Our confidence is mixed: strong that reviewers hold this view, thin on how often the problem occurs, because only one source measures any rate and it measures one task. The reading is of the plans and app version named above, on 13 October 2026.

Recurring strengths

Four of the six review articles, in the editorial tier, praise the citations themselves. They describe numbered links on every answer, a one-click route to the cited page, and more cited sentences than rival tools carry. Perplexity's own help page makes the same claim about itself: each answer carries numbered citations so a reader can check it. That is the maker's claim, and the review articles agree it is how the product behaves.

Two of the six, in the editorial tier, say mainstream and well-indexed topics are handled well, and one adds that most factual claims it checked cited strong sources. One review article, Zapier's, says the Max plan draws on paywalled research databases and finds the sources more authoritative than a rival chatbot's. The other five do not make that comparison.

The pricing page lists Free at $0 a month, Pro at $17 a month billed annually and Max at $167 a month billed annually. The Free plan lists citations in every answer. Four of the review articles quote $20 a month or $200 a year for Pro, which is close to the annual rate once divided by twelve.

Recurring complaints

Three of the six review articles, in the editorial tier, say a cited page can fail to support the claim beside it. One reviewer says they caught this several times when they clicked through. The study reported by the Columbia Journalism Review measured it on one task: asked to identify the article behind an excerpt, the free version of Perplexity answered 37% of 200 queries incorrectly. That is one task on news excerpts in March 2025, and it is not a rate for ordinary use. The study also reports that the paid version answered more questions correctly and also had a higher error rate, because it gave confident wrong answers where the free one sometimes declined.

Three of the six review articles say weak, stale or copied pages get cited. They describe lower-quality or AI-generated pages cited as if authoritative, cached pages presented as current, and syndicated copies cited in place of the original. The study found the paid version cited syndicated copies in three of ten queries for one publisher.

Two of the six, and the study, say paywalled or blocked pages turn up. The study reports that the free version identified every excerpt from one publisher that blocks its crawler. A Hacker News thread of 124 points and 48 comments carried the same worry about AI search in general. Its title came from an audit we did not open, and some commenters doubted who had published it.

These are what reviewers say, not facts about the company's conduct, and a count of complaints is not a failure rate.

Where reviewers split

Reviewers split on source quality. Zapier finds the sources more authoritative than a rival chatbot's. Three other review articles, Toolchase's, AllAble's and FutureAGI's, and the study describe weak, stale or copied pages being cited. We read both as true: one compares tools, the others judge each citation.

The study and the review articles also pull apart on size. The Tow Center study, reported by the Columbia Journalism Review, puts one task at 37% incorrect. Toolchase says most factual claims in its 20-query check cited strong sources, with weak ones on niche topics. The two used different tasks, so neither number settles the other.

On Hacker News, commenters split between sharing citation mismatches they had met in AI search and doubting the audit behind the thread's title. One thread is one source, and we count it once. Four of the six review articles say they used the product themselves. The other two summarise studies and report no checks of their own.

The review articles are dated from March to October 2026, and the study is from March 2025. Pricing and plan names may have moved in that time, and the March article predates the other five.

Who it suits

You want links you can open in one click and mean to open them. Four of the six review articles praise exactly that, and five of the six say to check.

You want a quick first pass on a mainstream topic. Three of the six describe it as a good starting point or source-finder.

Who should pass

You would quote an answer without opening the cited page. Five of the six review articles say to check, and three say a cited page can fail to support its claim.

You work on niche technical topics or fresh prices. One review article says citation quality drops on niche topics, and another says stale pricing from cached pages turned up.

You want a figure for how often the citations hold. No source we read gives one for ordinary use.

One verdict a week: the most useful Master Review we finished, the complaint that kept appearing, and who should skip it.

Sources

• Perplexity pricing page, official, read 13 October 2026.
• Perplexity help centre, how Perplexity works, official, read 13 October 2026.
• Perplexity on the App Store, buyers, headline figures read 13 October 2026.
• Zapier, Perplexity vs ChatGPT, editorial, revised March 2026.
• TechSifted review, editorial, updated 4 April 2026.
• FutureAGI, where its citations hold up and where they break, editorial, 9 August 2026.
• Toolchase review, editorial, updated September 2026.
• GLBGPT, Perplexity citation accuracy, editorial, updated 25 September 2026.
• AllAble review, editorial, updated 9 October 2026.
• Columbia Journalism Review, Tow Center study of eight AI search tools, news, 6 March 2025.
• Hacker News thread, a third of Perplexity's citations, forum, 2 September 2026.

On 13 October 2026 we read 6 review articles from March to 9 October 2026, 1 study reported by the Columbia Journalism Review on 6 March 2025, 1 Hacker News thread, 2 Perplexity pages and 1 App Store listing, by web search and plain page fetch, and synthesised them with AI; we collected no buyer reviews and tested nothing.