What Is Elicit?Elicit 是什么？
Elicit is a browser-based AI research assistant built for scientific literature rather than general web answers. Ask a research question and it returns relevant papers with one-sentence abstract summaries, ranked by relevance rather than keyword match alone; it searches a corpus of 138M+ papers (some 2026 coverage cites 125M+ or 128M — the corpus grows, so treat the exact count as approximate) plus PubMed keyword search and uploaded PDFs. Its defining workflow is the systematic review: set the question and search strategy, choose Fast or Thorough screening depth, and let Elicit gather papers, screen them against AI-generated inclusion and exclusion criteria that you can edit, and extract data into columns — population, intervention, outcomes, sample sizes — with each value linked back to supporting text. A research report synthesises the screened set, and on higher tiers a PRISMA-style diagram and logged methodology are produced. Screening ceilings are the stable numbers across sources: Pro 5,000 papers, Scale 20,000, Enterprise 40,000; extraction columns run 20 on Pro, 30 on Scale and 40 on Enterprise, and final reports draw on up to 135 sources on Pro and 200 on Scale. Figure extraction and live team collaboration arrive on Scale; API access starts on Pro; Zotero import is on every plan. For a multilingual search-to-output pipeline on this site, see Felo AI Search — Elicit is the extraction specialist, Felo the broader research-to-slides pipeline.Elicit 是面向学术文献的浏览器端 AI 研究助手，不做通用网页问答。输入研究问题，它返回按相关性排序的论文与一句话摘要，不只按关键词匹配；语料库为 138M+ 篇论文（部分 2026 报道写 125M+ 或 128M——语料在增长，精确数字视为约数），另支持 PubMed 关键词检索与上传 PDF。它的标志性工作流是系统综述：定好问题与检索策略，选 Fast 或 Thorough 筛查深度，Elicit 负责聚拢论文、按可编辑的 AI 生成纳入/排除标准筛查、再把数据抽成列——人群、干预、结局、样本量——每个值都链回原文佐证。研究报告综合筛查后的文献集，高档位可出 PRISMA 式流程图与方法学记录。筛查上限是各家一致的稳定数字：Pro 5000 篇、Scale 20000 篇、企业版 40000 篇；提取列 Pro 20 列、Scale 30 列、企业版 40 列；终稿报告 Pro 最多引用 135 个来源、Scale 200 个。图表提取与团队实时协作在 Scale 档；API 从 Pro 档起；Zotero 导入全档位都有。本站多语言“检索到成稿”管线见 Felo AI Search——Elicit 是提取专精，Felo 是更宽的检索到幻灯片管线。
Verdict: buy Elicit when the deliverable is an evidence table, not an answer. If the job is a quick literature skim or a first draft, a general chatbot or a broader research tool is cheaper and faster. Elicit earns its seat when hundreds of papers must be screened against explicit criteria and turned into auditable columns — the systematic review, the meta-analysis appendix, the policy evidence brief. The buying discipline is the ceiling: pick the tier whose screening count and column count survive your largest review, and never cite an extracted value without opening the source paper. Its recall gap is the reason it is a second reviewer, not a replacement for a registered database search.
实在的结论：Elicit 只有在交付物是一张证据表、而不是一个答案时才值。如果你只是快速扫文献或起草初稿，通用聊天机器人或更宽的研究工具更便宜更快。Elicit 的席位价值在必须按明确标准筛查数百篇论文、并转成可审计的列时——系统综述、meta 分析附录、政策证据简报。购买纪律是看上限：选筛查数与列数能扛住你最大一篇综述的档位；提取值不打开原文核对就不引用。它的查全率缺口决定了它是第二筛查员，不是注册数据库检索的替代品。
— 中国吧 AI Tools Hub Editorial Team
Who Should Use It?谁适合用？
✓ Good Fit For适合以下场景
- Systematic and scoping reviews with explicit screening criteria有明确筛查标准的系统综述与范围综述
- Research teams building evidence tables across hundreds of papers跨数百篇论文建证据表的研究团队
- Policy, health and R&D teams preparing evidence briefs做证据简报的政策、医疗与研发团队
✕ Not Ideal For不适合以下场景
- Anyone who needs one quick answer with citations只要一个带引用快答案的人
- Registered reviews that need a full audit trail as the primary search把 Elicit 当主检索做注册综述、要求完整审计链的人
- Humanities topics and very recent or paywalled work as the sole source只靠它做人文选题、最新或付费墙内文献的人
Getting Started快速开始
1. Start free: one real question, check whether Elicit finds the right literature1. 先免费：拿一个真实问题，验 Elicit 是否找得到对的文献2. Test extraction on 10 papers: columns, values, then open each source to verify2. 拿 10 篇论文试提取：列、值，然后逐篇打开原文核对3. Pick the tier by screening ceiling and columns — then confirm the checkout total3. 按筛查上限与列数选档——再确认结算价Key Features核心功能
- Semantic Search Over 138M+ Papers138M+ 论文语义检索 — Relevance-ranked papers with one-sentence abstract summaries, plus PubMed keyword search and PDF upload.按相关性排序的论文与一句话摘要，另支持 PubMed 关键词检索与 PDF 上传。
- Systematic-Review Screening Workflow系统综述筛查工作流 — Fast or Thorough screening against editable inclusion and exclusion criteria, with supporting quotes per decision.Fast 或 Thorough 筛查深度，按可编辑的纳入/排除标准筛查，每项决定附佐证引用。
- Column-Based Structured Extraction列式结构化提取 — Extract population, intervention, outcomes and custom fields into tables; 20 columns on Pro, 30 on Scale, 40 on Enterprise.把人群、干预、结局与自定义字段抽成表；Pro 20 列、Scale 30 列、企业版 40 列。
- Research Reports With Sentence-Level Evidence研究报告：证据到句 — Reports synthesise the screened set with claims linked to exact sentences in source papers; up to 135 sources on Pro, 200 on Scale.报告综合筛查后的文献集，结论链到原文精确句子；Pro 最多 135 个来源、Scale 200 个。
- Alerts, Library, Zotero and API提醒、资料库、Zotero 与 API — Personalised research alerts, a source library, Zotero import on every plan and API access from Pro.个性化研究提醒、来源资料库、全档位 Zotero 导入、Pro 档起 API 接入。
Pros & Cons优缺点
✓ Pros优点
- The only widely cited tool with a true systematic-review screening pipeline少有真正系统综述筛查管线的工具，引用广泛
- Sentence-level citations make AI output fast to audit句级引用让 AI 输出核查很快
- Free Basic tier genuinely covers exploration: unlimited search and summaries免费 Basic 档确实能探索：搜索与摘要不限量
- Screening ceilings and column counts are stable and tiered clearly筛查上限与列数稳定、分档清晰
✕ Cons缺点
- Pricing page shows multiple ladders — $39 vs $49 Pro, $89 vs $169 Scale定价页多套梯队并存——Pro $39 对 $49、Scale $89 对 $169
- Recall gap: about 40% of relevant studies in one independent evaluation查全率缺口：一次独立评估仅约 40% 相关文献
- Automated extraction can misread tables and nuance — verify before citing自动提取可能误读表格与细微处——引用前须核对
- Figure extraction and live collaboration sit on Scale, not Pro图表提取与实时协作在 Scale 档、不是 Pro 档
Use Cases应用场景
Popular ways people use it.常见用法：
🧪 A Systematic Review, Screened in Days几天筛完一篇系统综述
Gather, screen and extract across thousands of papers against explicit criteria, then build the evidence table.按明确标准聚拢、筛查、提取数千篇论文，再建成证据表。
📊 Meta-Analysis Evidence TablesMeta 分析证据表
Pull sample sizes, interventions and outcomes into columns that feed a meta-analysis appendix.把样本量、干预与结局抽成列，喂给 meta 分析附录。
🏛️ Policy Evidence Briefs政策证据简报
Policy, health and R&D teams turn a question into a cited brief with traceable sources.政策、医疗与研发团队把一个问题转成来源可追溯的证据简报。
🎓 Thesis Literature Mapping学位论文文献摸底
Map a field fast on the free tier, then pay only when extraction volume demands it.先用免费档快速摸清领域全貌，提取量上来再掏钱。
Similar AI Tools相似 AI 工具
Related pages on this site: Felo AI Search for the multilingual research-to-output pipeline — search, slides, documents and mind maps from the same research. Elicit is the narrower, deeper pick when the output must be an auditable evidence table; Felo is the broader pick when the same research must also become slides or a document the same day.本站相关页：Felo AI Search（多语言检索到成稿管线——同一批研究出幻灯片、文档与思维导图）。当交付物必须是一张可审计的证据表时，Elicit 是更窄、更深的选择；当同一批研究当天还要变成幻灯片或文档时，Felo 是更宽的选择。