← Research

PerplexityPerplexity

Perplexity AI

免费+订阅Free + subscription 带来源问答Sourced Q&A Comet浏览器Comet browser 多模型路由Multi-model routing 需网络环境Network access required partial lifecycle: active 核验Checked 2026-10-22

Perplexity 是「答案带来源角标」的 AI 搜索旗舰：多模型路由+自研线，Comet 浏览器把检索延伸成 agent 浏览；「带来源」是招牌，但角标背后的原文仍要自己核。

The flagship of "answers with source footnotes" AI search: multi-model routing plus an in-house line; the Comet browser extends search into agentic browsing — but the original behind each footnote still needs your own check.

适合与不适合When to use it, when not to

适合：带来源事实问答、快速研究入口、英文向检索。不适合：大陆直连（000）、不核原文就引用、中文深度检索主打。

Sourced factual Q&A, quick research entry points, English-first retrieval quality. Not for: mainland direct access (measured 000), treating answers as source-free conclusions (click footnotes and check originals), Chinese deep retrieval as the main axis.

中文生态China availability 国产 / 国内可直接用 / 备案状态 · 快照domestic / direct access / CAC filing · snapshot 2026-10-22

出品属地Origin: 国外出品foreign-made　Perplexity AI（美国旧金山）。Perplexity AI (San Francisco, US).
国内使用Direct access: 直连不可达direct blocked　perplexity.ai 大陆直连实测 000（2026-10-22，代理下可达）——直连不可达，需网络环境。perplexity.ai measured 000 direct from mainland (2026-10-22; reachable via proxy) —unreachable without a network path.
备案状态CAC filing: 未见于备案名单not in filing list：未见于网信办公示名单（2026-10-22 检索）；该产品未面向大陆公众提供服务，备案不适用。not in the CAC public registry (searched 2026-10-22); the product does not serve the mainland public, so filing does not apply.

备案信息以网信办《生成式人工智能服务已备案信息》公示名单为准（按批次滚动更新，本站只抄编号不解读）；「大陆可直连」是本站从大陆网络直连实测的结果，不是厂商承诺。GitHub 类站点当日实测有波动，口径写在依据里。Filing status is quoted from the CAC public registry (updated batch by batch; this site copies the number and does not interpret it). "Direct access" is this site's own reachability measurement from a mainland network, not a vendor promise. GitHub-hosted sites fluctuated on the measurement day — the note says so.

固定坐标系Fixed coordinate system

8 个维度dimensions, 与同分区其他对象逐项可比each comparable item by item against others in the same section.

模型与开放条件Model & openness: 多模型路由+自研线：界面可选多家前沿模型（官方产品面），另有自研模型线（官方叙事）；具体当前清单未逐条核验。API 平台（perplexity.ai/api）存在，细节未逐条核验。
Multi-model routing + an in-house line: several frontier models selectable in the UI (official product surface), plus a self-developed model line (official narrative); the current concrete list not verified item by item. An API platform (perplexity.ai/api) exists, details not verified item by item.
运行位置Where it runs: 网页+移动 App 主入口；Comet 浏览器（官方产品线：AI 浏览器，检索+agent 浏览）；检索与生成在厂商云端。
Web + mobile App as main entries; the Comet browser (official product line: an AI browser, retrieval + agentic browsing); retrieval and generation on the vendor's cloud.
本地文件Local files: 线程/收藏历史存账户侧云端；未见官方声明的本机数据面或离线模式（未逐条核验）。
Threads/bookmarked history stored on the account-side cloud; no officially stated local data surface or offline mode seen (not verified item by item).
关机后的任务Tasks after shutdown: 问答即时；Deep Research 线（多步检索报告，官方产品面）；无「关机续跑」概念。
Q&A is instant; the Deep Research line (multi-step retrieval reports, official product surface); no "keep-running-after-shutdown" concept.
工具与扩展Tools & extensions: 答案带来源角标（产品招牌）、聚焦模式（Focus：学术/社交等检索域切换）、Spaces（知识集合）、 Pages（发布）、文件上传问答；Comet 把检索延伸到浏览操作面。
Answers carry source markers (the product's signature), Focus modes (switching retrieval domains: academic/social etc.), Spaces (knowledge collections), Pages (publishing), file-upload Q&A; Comet extends retrieval into browsing operations.
上下文与记忆Context & memory: 以线程/Spaces 组织，历史可回看；跨会话记忆面存在（官方叙事），细节未逐条核验。
Organized by threads/Spaces, history reviewable; a cross-session memory surface exists (official narrative), details not verified item by item.
权限与限制Permissions & limits: 账号+订阅体系（D9：价格不采）；训练用途开关在账户设置（官方叙事），默认态未逐条核验；出版方合作/分成叙事（官方公告线）——内容授权是它的行业角色之一。
Account + subscription system (D9: pricing not collected); the training-use switch is in account settings (official narrative), default state not verified item by item; publisher-partnership/revenue-share narrative (official announcement line) — content licensing is one of its industry roles.
适合什么任务Best for: 适合：带来源的事实问答、快速研究入口、英文向检索质量优先。不适合：大陆直连（实测 000）、把答案当无源结论（必须点角标核原文）、中文深度检索主打。
Fits: sourced factual Q&A, quick research entry, English-leaning retrieval quality first. Does not fit: mainland direct connection (measured 000), treating answers as source-free conclusions (you must click the markers and check the original), Chinese deep retrieval as the mainstay.

头号误解Top misconception

「带来源」≠「来源可靠」——角标要点开核原文，聚合搜索也会聚合到差源"Has sources" is not "sources are reliable" — click the footnotes and check the original; aggregation-style search also aggregates bad sources
Comet 是浏览器不是搜索附属——agent 浏览能力意味着它可能代你操作网页，权限面比搜索大Comet is a browser, not a search add-on — its agentic browsing means it may act on web pages for you, a bigger permission surface than search
perplexity.ai 大陆直连 000（2026-10-22，代理可达）；官网对 curl 常 403（CF）——核验方法受限如实标注perplexity.ai measured 000 direct from the mainland (2026-10-22, reachable via proxy); the site often returns 403 to curl (Cloudflare) — verification-method limits noted honestly

未知项清单Known unknowns

当前可选模型清单（未逐条核验）
ToS 条款细节（CF 403 未能直抓——不代表不存在）
训练开关默认态（未逐条核验）
出版方分成的具体机制（官方叙事，细节未逐条核验）

The current selectable-model list (not verified item by item)
ToS clause details (CF 403 blocked direct fetch — does not mean they do not exist)
Training-switch default state (not verified item by item)
The concrete mechanism of publisher revenue share (official narrative, details not verified item by item)

证据来源Evidence sources

判断可回到以下一手源复核。本站核验日Every judgement can be rechecked against the primary sources below. Verified on 2026-10-22，内容更新日content updated on 2026-10-22。

类型Type	名称Name	链接Link
official	Perplexity · 官网（产品与能力叙事）Perplexity · official site (product and capability narrative)	https://www.perplexity.ai
official	Perplexity · ToS（官方条款，直抓 CF 403，2026-10-22）Perplexity · ToS (official terms; direct fetch hit CF 403, 2026-10-22)	https://www.perplexity.ai/hub/legal/terms-of-service
changelog	Perplexity · Blog（官方更新线，changelog；直抓 CF 403）Perplexity · Blog (official update line, changelog; direct fetch hit CF 403)	https://www.perplexity.ai/hub/blog

实测记录Measured runs

本档未跑检索质量评测；结论只覆盖产品能力与来源披露。

实测建议（三条，都指向官方可读页没给的东西）：

来源角标与正文论断的对应准确度（抽样核对）
Comet 的 agent 操作权限边界（它会替你点什么）
Deep Research 报告与手动多轮搜索的质量差

This site has not completed measured runs. The coding-oriented test protocol is at protocols/coding/_protocol.md (the coding task suite does not apply to creation products).

Suggested measurements (three, all pointing at things the official readable pages do not give):

How accurately source markers map to the claims in the answer text (sampling check)
The agentic-operation permission boundary of Comet (what it will click on your behalf)
Quality gap: Deep Research report vs manual multi-round searching