Petal
Petal is an AI document-analysis platform for researchers, corporate R&D teams and subject-matter experts. It combines a reference manager, a citation generator and a chat that answers only from the documents you upload, with sources.
What is Petal?
Petal is an AI document-analysis platform published by Paladin Max, Inc., a Delaware corporation headquartered in Walnut Creek, California. The product grew out of academia — it was previously called KaleidoGlobe, and the About page traces its founders' work from 2013 to 2018 — and it still addresses researchers first.
Three blocks sit side by side. The Petal Reference Manager is a cloud library where PDFs are uploaded, sorted into collections, tagged, deduplicated and searched full-text and semantically, with metadata extracted automatically. The Citation Generator searches books and articles, imports records by DOI, PMID, ArXivID, ISBN, ISSN or EAN, captures metadata from a URL and exports to BibTeX or Word in more than 10,000 styles. On top of both sits the AI layer: single-document chat, multi-document chat, an AI Table that queries and compares several documents against criteria you define, and AI Create, a notebook for assisted writing. Reading tools translate into more than ten languages, summarize, explain a selected passage, suggest questions and pull out key points.
The stated promise is context-aware generative AI: answers are drawn from the user's own files rather than from a general-purpose model, a contrast the publisher makes explicitly with ChatGPT and its pre-2021 corpus. Technically it is a cloud SaaS running on Amazon Web Services with servers in Oregon, United States, calling OpenAI's GPT-3.5 and GPT-4; consumption is metered in monthly AI credits, from 400 to 2,000 depending on the plan. The surfaces are the web application, a Web Importer extension for Chrome, Firefox and Safari, and a Microsoft Word add-in.
Three use cases are published — Academia, Corporate R&D and Industry Experts — supported by shared workspaces, guests, annotations, comments, shareable links and granular access controls. The company claims 20,000 to 30,000 researchers, teachers and experts, displays logos including MIT, Caltech, UT Austin, University of Hawaii Manoa, Mira Geoscience, Indeed and Intuit, names an advisory board (Karen Willcox, Zeynep Noelle Koller, Marcus Koller) and is referenced by MIT Libraries. A white-label offer and an enterprise license extend the product to research institutes, universities and large organizations. The publisher states that it is owned by no scientific publisher and trains no model on customer data.
What it does
- Ask questions about your own documents and get answers backed by the source passages
- Compare several documents at once in an AI table built around criteria you define
- Summarize, translate, explain a selected passage, suggest follow-up questions and extract key points
- Build a research library: import files, automatic metadata, deduplication, collections and tags
- Generate citations and bibliographies in more than 10,000 styles
- Annotate, highlight, comment and share documents with collaborators
- Capture PDFs and web pages from the browser into the library, then draft in the built-in AI Create notebook
When to use Petal / When not to
A quick filter to help you decide if Petal is the right fit.
When to use Petal
- Academic researchers and doctoral students running a literature review, who need to reduce hundreds of papers to the few that match criteria they define themselves
- University librarians, lecturers and teaching staff looking for a low-cost reference manager with citation and bibliography generation in more than 10,000 styles
- Corporate R&D teams that need to compare similar studies side by side in a multi-document AI table and share the result in a common workspace
- Subject-matter experts building a private technical library, who need full-text semantic search, deduplication and granular access controls over their own corpus
- Undergraduate and high-school students who want a passage explained, translated or reduced to its key points while they read a PDF, on a permanent free plan
When not to use Petal
- Developers and product teams that need a public API: the standard petal.org product lists API access as not available, and a customizable API exists only under the white-label agreement
- Organizations with European data-residency or GDPR documentation requirements: hosting is exclusively in Oregon, United States, and there is no data processing agreement and no Article 27 representative
- Teams handling confidential, unpublished or patient-related documents: the publisher states that stored files are not encrypted end to end and that its engineers can access them for debugging, under NDA
- Large departments working outside an enterprise license: every individual plan carries a single seat and three guests, with monthly AI credits capped and not carried over
- Non-English-speaking users and anyone who depends on native PDF annotations: the interface and support are in English only, and existing PDF annotations are not preserved on import or export
How to use Petal
A typical end-to-end flow, from setup to results.
- Create a free account on cite.petal.org/signup, or sign in with a Google account
- Confirm the verification email within 24 hours, checking the spam folder if it does not arrive
- Upload PDFs to the cloud library, or install the Web Importer for Chrome, Firefox or Safari to capture PDFs and web pages while browsing
- Sort documents into collections and tags: metadata is extracted automatically and duplicate files are detected
- Open a document in the viewer, select a passage and ask the AI to explain it, translate it or extract its key points
- Chat with a single document on any plan; multi-document chat and AI Create unlock from the Advanced plan
- Use the AI Table to query and compare several documents against criteria you define
- Generate citations and bibliographies from the citation generator (cite.petal.org/citation/generator/home)
- Install the Microsoft Word add-in to insert references and build a bibliography while writing
- Invite collaborators into a shared workspace, annotate and comment, then produce a shareable link; support is reached at help@petal.org
Pros & Cons
Pros
- Answers are built from the user's own corpus and shown with their sources, which narrows the hallucination risk of a general-purpose assistant
- A permanent free plan, backed by a public commitment from the publisher to always keep a generous free tier
- Prices calibrated for academia, with a reduced .edu rate and a 15% annual discount
- Three tools in one place — library, citations and AI — instead of three separate products to keep in sync
- More than 10,000 citation styles, imports by scholarly identifier (DOI, PMID, ArXivID), extensions for the three main browsers and a Microsoft Word add-in
- Unusually candid security FAQ: the absence of end-to-end encryption and the engineers' access to documents are stated outright, alongside a declared commitment not to train or fine-tune language models on customer data
- Ownership independent of any scientific publisher, and a published shutdown policy promising at least six months' notice and a shift of effort toward open source
Cons
- No mention of the GDPR anywhere on the site, no data processing agreement and no Article 27 EU representative
- Hosting exclusively in the United States (Oregon), with no European or customer-selected region
- No end-to-end encryption at rest, engineers able to read uploaded documents even under NDA, and no formal list of subprocessors published
- ISO 27001 certification described only as in progress, in wording unchanged since the July 2022 privacy policy
- No public API on the standard product: API access is listed as not available and exists only in the white-label offer
- A single seat on each individual plan, monthly AI credits capped without carry-over, and native PDF annotations lost on import and export
- Signs of an aging site: 2022 copyright, a privacy policy from July 2022, features still announced for late 2022, a misspelled contact address in the policy, a pricing page that contradicts itself, an English-only interface and no demo page to evaluate without signing up
Pricing & Plans
A permanent free plan is available at no cost. The lowest paid entry point advertised is the Plus plan at USD 2.55 per month, a figure that already includes an early-bird reduction and the 15% annual discount and applies to a .edu account; Advanced is advertised at USD 8.49 per month and Premium at USD 25.49 per month. Rates differ for accounts without a .edu address, and the pricing page is inconsistent on this point: the plan cards quote USD 2.99 (.edu) and USD 4.99 (other) for Plus and USD 9.99 / USD 13.99 for Advanced, whereas the comparison table quotes USD 2.55 / USD 4.25 and USD 8.49 / USD 11.89. The figures below should therefore be confirmed with the publisher. Enterprise licenses and white-label agreements are quoted on request, and add-ons may be purchased to exceed the monthly AI credit cap.
- 1 GB of storage
- 1 seat
- 3 guests
- standard support
- 400 non-renewing AI credits (cap 400)
- unlimited citation lists without export
- 2 collections
- 3 annotations per document
- 2 GB of storage
- 1 seat
- 3 guests
- standard support
- 400 AI credits per month (cap 400)
- unlimited citations with export
- unlimited collections and annotations
- single-document chat and AI Table
- 10 GB of storage
- 1 seat
- 3 guests
- priority support
- 1
- 200 AI credits per month (cap 1
- 200)
- unlimited citations with export
- 25 GB of storage
- 1 seat
- 3 guests
- priority support
- 2
- 000 AI credits per month (cap 2
- 000)
- all features included
- for research institutes
- universities and large organizations
- with the features and usage level chosen by the customer
- storage on the partner's own servers
- computation on AWS managed by Petal
- customizable API
- customizable access controls and quotas
- embeddable AI chat and document viewer
- and custom branding
Data, GDPR & hosting
A consolidated view of how Petal handles your data.
GDPR overview
No GDPR claim is made anywhere. The words GDPR and General Data Protection Regulation appear nowhere on the site, and there is no data processing agreement, no Article 27 EU representative and no named data protection officer. The privacy policy, effective July 15, 2022, grants rights in California-style wording: to know, to access in a portable format, to correct, to delete, to restrict processing, to withdraw permissions and to close the account, exercised by verifiable request sent from the account address to help@petal.org. Acknowledgment is promised within 10 business days and a response within 45 days, access and portability being limited to two requests per 12-month period, with possible fees for manifestly unfounded or excessive requests and a non-discrimination guarantee. Transfers to servers in the United States are acknowledged explicitly, with a warning that local law may be less protective. EEA residents are carved out of the arbitration clause.
Who owns the data?
The terms of service split ownership. Paladin Max, Inc. owns everything it creates and makes available through the service, defined as Company Content, while the user retains ownership of everything submitted or transmitted through it, defined as Your Content. The terms then introduce Content Derivatives — portions extracted from uploaded content and their associated metadata — and reserve the publisher's right to use those derivatives for its own internal purposes, such as improving the service or its other products. Closing an account deletes or anonymizes the stored content, irreversibly. Under the white-label agreement the partner is described as keeping complete ownership and control of its customer data, held on its own servers.
Reuse rights
Users may reuse their own uploaded content freely: nothing in the terms restricts what they do with their documents, or with the citations, bibliographies and notes produced through the service, and no permission has to be requested. On the publisher's side, the privacy policy declares processing for delivering the service, improving customer records, legitimate interests such as fraud prevention, security and audits, legal obligations and statistical analysis. The categories collected include identifiers, network activity, professional information, inferences, generated profiles and content, device, software and browser data, cookies and web beacons, publicly available information from sources such as Google Scholar and ArXiv, and third-party sources including YouTube, Vimeo and Twitter. Petal states that it will never sell user data and that it does not train or fine-tune large language models on it. Unmarked chunks of text taken from uploads are, however, sent to OpenAI's GPT-3.5 and GPT-4 as context when an answer is generated; the publisher undertakes never to send a complete document to any other organization. Its own engineers can access uploaded documents for debugging and user-experience work, under signed NDAs. Do Not Track browser signals are not honored.
Data retention & training
Hosting summary
Petal is delivered as software as a service: the application and everything uploaded to it live online. Storage and computation run on Amazon Web Services, and the publisher states that its servers are located in Oregon, United States. The privacy policy repeats that user information is stored on servers primarily located in Oregon and warns that personal information submitted from outside the United States may be transferred to and stored there, under data protection laws that might not be as comprehensive as those of the user's own country. No alternative region is offered on the standard product: there is no European, Asian or customer-selected hosting option, and no formal list of subprocessors is published, although OpenAI is named as the language model provider and AWS as the infrastructure provider. Under the white-label agreement the arrangement changes: customer data is stored on the partner's own servers while computation stays on AWS infrastructure managed by Petal, and the partner keeps complete ownership and control of that data. ISO 27001 certification is presented as in progress, with a claimed alignment on the 18 CIS Critical Security Controls.
Where Petal works
Country-level availability.
Not available in
Things to keep in mind
Risks and trade-offs to weigh before adopting Petal.
- Data leaves the European Union: everything is stored on servers in Oregon, United States, the transfer is acknowledged in the privacy policy, and no GDPR commitment, data processing agreement or Article 27 representative is published
- Uploaded documents are not encrypted end to end, the publisher's engineers can read them for debugging under NDA, and no formal subprocessor list exists — think twice before uploading confidential, unpublished or patient-related material
- Fragments of every document queried are sent to OpenAI as generation context; a complete file is never sent, but the exchange itself cannot be switched off
- A sourced answer is still a model's answer: a citation shown next to a sentence proves the passage exists, not that the model read it correctly. Verify quotations and figures before they reach a manuscript or a regulatory submission
- Leaning on summaries and key-point extraction can quietly replace reading; for students in particular, the risk is a literature review whose sources were never actually read
- The terms let the publisher use Content Derivatives — portions extracted from your documents and their metadata — for its own internal purposes, including improving its other products
- Practical traps: ISO 27001 announced as in progress since 2022 and never confirmed, a self-contradictory pricing page, AI credits capped without carry-over, a misspelled contact address in the privacy policy, and California law with AAA arbitration and a class-action waiver outside the EEA
Setup & Integrations
Technical difficulty
Low. No installation is required for the main product, which runs entirely in a browser. Sign-up is free, either with an email address — the verification message must be confirmed within 24 hours — or with a Google account. Optional additions are the Web Importer extension for Chrome, Firefox or Safari and the Microsoft Word add-in, each installed from its own store in a few clicks. No server configuration, key management or developer work is needed on the standard product. Under the white-label agreement, installation and configuration are handled by the Petal team.
Deployment
Apps stores
Integrations
Supported languages
Behind Petal
Social
Resources
All the official URLs gathered for verification and reference.
Alternatives
Tools that compete with or complement Petal.
Frequently asked questions
Does Petal train AI models on my documents?
Does any of my content leave Petal for third parties?
Are my documents encrypted end to end?
Where is my data hosted?
Is there a permanent free plan, and what does the first paid plan cost?
Is Petal open source, and what happens if the company shuts down?
Is Petal owned by a scientific publisher?
Is there an API?
Which browsers and add-ins are supported, and how many citation styles are available?
How do I contact support, and how fast is the answer?
Should you pick Petal?
Petal comes from academia and shows it. The reference manager and the citation generator are the mature part of the product: imports by DOI, PMID or ArXivID, automatic metadata, deduplication, more than 10,000 citation styles, browser extensions and a Word add-in cover the everyday mechanics of a literature review. The AI layer is grafted onto that library rather than onto the open web, and that is its central argument against general-purpose assistants: answers arrive with the passages they rest on, taken from a corpus the user controls.
The pricing matches the intended audience. A free tier the company commits publicly to keeping, a .edu rate, a 15% annual discount and a paid entry point advertised at USD 2.55 per month put the tool within reach of a student or a laboratory budget, while the multi-document AI Table gives corporate R&D teams a genuine comparison workflow.
The reservations are mostly legal. The GDPR is never mentioned, no data processing agreement or Article 27 representative exists, all data sits on AWS servers in Oregon with no European option, documents are not encrypted end to end, and the publisher's engineers can read them under NDA. Petal deserves credit for stating these facts plainly instead of burying them, but a European organization handling personal or sensitive research data will not find the paperwork it needs. A secondary reservation: parts of the site look frozen — a 2022 copyright, a privacy policy effective July 2022, features announced for late 2022, and a pricing page that contradicts itself between its cards and its comparison table.
Petal is a sound choice for individual academic work or for an R&D team that is relaxed about where its data lives. Buyers with data-residency or GDPR documentation requirements should obtain those commitments in writing first.
- Choosing a selection results in a full page refresh.
- Opens in a new window.