On this page

What Humata does

Humata lets users ask questions about files, request summaries and compare documents. Its product page emphasizes citations that lead back into source files. For academic research, the useful job is navigating material you have chosen, then checking the passage behind each answer. This updated review assesses official documentation, not a hands-on comparison of answer quality or a guarantee that every relevant passage is retrieved.

Choose the Humata workflow for your next task

Start with the file set and handoff you need. These routes point to documented controls; selecting files or seeing a citation does not establish complete retrieval or source support.

Humata workflows by task
Your taskStart withCheck before continuing
Ask a question that should stay within one paperCheck the response mode and begin with GroundedOpen the cited passage and confirm its context. Grounded describes intended source boundaries, not a measured accuracy guarantee. [2]
Compare information across specific papersSelect the intended files or folder before Ask AllKeep the selected set explicit, then verify each comparison against its own source. The selection guide does not establish complete retrieval. [7]
Read a scanned or image-only PDFCheck selectable text and OCR access firstThe pricing table lists OCR separately from monthly pages and does not include it on Free or Expert. [3]
Estimate the cost of a larger reading setCheck included pages, overage and spending controlsThe listed allowances and per-page charges differ by plan; confirm the current account terms before uploading a large set. [3] [4]
Share a file with a classmate or embed a chatReview team roles, folder restrictions and embed settings separatelyAn invitation, file permission and website embed are different access paths; confirm who can view the source and how access can be revoked. [8] [9]

Where it fits, and where it falls short

Useful strengths

  • Document-linked citations provide a route from a response back to its proposed evidence.
  • Different response modes make the boundary between file content and added information explicit.
  • Published page allowances and spending controls help you plan a document-reading budget.

Limits to consider

  • A citation can point to a genuine passage without supporting the exact claim in the answer.
  • Page-based charges mean a large collection may cost more than the headline subscription.
  • Scanned text, sharing permissions and confidential uploads require plan and policy checks.

A suggested Humata workflow

View the workflow diagram
Humata: an illustrated workflow for checking sources and reviewing your work
An editorial illustration of the workflow.
  1. 1

    Choose a small set of permitted papers and define the information you need from each.

  2. 2

    Check the response mode and begin with narrow questions that have verifiable answers.

  3. 3

    Open the cited passage and inspect definitions, numbers and qualifications in context.

  4. 4

    Keep verified notes outside the chat, and check page usage before uploading more material.

Use Humata to inspect a reading set, not define the whole evidence base

A file-based answer is constrained by what you supplied and how you asked the question. Before importing a folder, decide which documents belong in the project and why. Separate a background report from a primary study, and label different versions of the same paper. Otherwise a comparison can mix sources with different purposes or quote an outdated version.

Ask for one checkable field at a time: the research question, population, method or reported limitation. Then decide whether the answer helps you locate a useful passage. For a literature review, keep discovery and screening records separately. A clear summary of your uploaded collection does not show that important studies outside the collection were considered.

Supporting documentation: [1]

Grounded, Balanced and Creative have different source boundaries

Humata's Chat Settings guide describes Grounded as answering from the uploaded documents and declining when the requested information is absent. Balanced can add information outside the files, while Creative uses them more loosely. For a reading note that must represent a particular paper, start by checking that the response setting matches that requirement.

Treat the mode description as an intended behavior, not a verified error rate. Include a question the paper cannot answer in your own trial and inspect the result. An admission that information is missing is useful. A confident extension beyond the paper needs a separate source before it belongs in an academic argument.

Supporting documentation: [2]

Choose the document set before using Ask All

Humata's Ask All guide tells users to tick the boxes beside files or a folder, then choose Ask All. Make that selection part of your research record. A question about two studies should start with the intended papers, rather than an account-wide collection that also contains lecture notes, review articles and earlier versions. The guide describes the selection step; it does not establish that every passage in every selected file will be retrieved.

For the urban-heat exercise below, treat records A and B as the selected set. With real papers, keep relevant supplements identifiable by filename. First ask each source's question separately and verify the source location. Then ask for a comparison across the selected set. If the comparison disagrees with the individual notes, return to the cited files before combining the findings. Naming the files in a prompt is useful context, but should not replace checking which files you actually selected.

Supporting documentation: [7]

Check the supporting passage, not just the citation marker

Follow the cited link into the document and read around the proposed evidence. Check whether a finding belongs to the authors' own analysis or to a previous study mentioned in the introduction. Confirm the sample and measure before turning a short answer into a sentence in your essay. A nearby keyword is not enough to establish support.

Figures and tables need their own inspection. Compare the reported number with the original unit, denominator, caption and notes. Do not interpret a failed extraction as proof that the paper lacks the information you asked for.

Check text extraction separately from the page allowance. Humata's pricing table marks OCR as unavailable on both Free and Expert, so a free page balance does not establish that an image-only scan can be read. Before choosing a plan, inspect whether your permitted PDF has selectable text and check the account's OCR entitlement. If you use a short public sample to evaluate extraction, compare a known sentence and a table entry with the original. This is a suggested check, not a scan we tested; text extraction alone would not verify the answer's interpretation.

Supporting documentation: [1] [3]

A constructed check: compare two urban-heat records

These are complete fictional extracts for this exercise, not published papers or answers obtained from Humata. A1 and B1 are local exercise locators. Read both records before deciding what a comparison can support; no account or document upload is required.

A1 — During one July afternoon in a fictional city, a satellite record gave mean land-surface temperatures of 34°C on one tree-shaded block and 36°C on one unshaded block. No air-temperature measurement was recorded.

B1 — In a different fictional city during one August evening, street-height sensors recorded air temperatures of 27°C on one tree-shaded street and 28°C on one unshaded street. No satellite measurement was recorded.

First identify the measure and observation period in each record. Then decide whether the two differences establish where trees had a stronger cooling effect, and whether either record gives a winter result. Make your own decisions before opening the explanations.

Which measure and difference does each record supply?

A1 supplies a 2°C difference in mean land-surface temperature during one July afternoon. B1 supplies a 1°C difference in street-height air temperature during one August evening. Keep the outcome, location and time with each number; the differences are arithmetic within these fictional records.

Can we rank a causal cooling effect by comparing 2°C with 1°C?

No. The records use different temperature measures, places and observation times, and do not supply a design that isolates a causal effect of trees. A numerical comparison alone does not make the outcomes equivalent or establish which intervention works better.

What is a defensible response to a question about winter temperatures?

Winter temperatures are not supplied in either exercise record. Leave that field unresolved rather than inventing a seasonal value. With an actual paper, distinguish information absent from the passage you inspected from information absent from the whole publication.

Set a page budget before uploading a large collection

Humata's pricing distinguishes included pages from paid additional pages. Ten short papers and ten long reports can consume very different allowances. Count the material you intend to process, including supplements and duplicate versions, rather than estimating cost from file count. Check current usage before starting the next batch.

The current plan table lists one user on Free, up to three on Expert and up to 10 on Team. It also marks GPT-5 support across the displayed tiers. These are plan-table statements, not evidence that a particular model is more accurate or that every answer covers the uploaded files. Check current plan details and account access before estimating a team workflow.

The official spending-control guide places a limit for additional-page spending in the Plan section of Account settings. Set and inspect that control before exceeding your allowance; this review has not tested its enforcement or timing. The homepage's broad unlimited wording should not be read as unlimited free usage. Check what happens at the cap and ask support if the account does not make it clear.

Supporting documentation: [3] [4]

Keep website imports and research-note exports inspectable

Humata’s Website Import guide describes importing from a URL with a site-wide selection. If your reading set includes public policy pages or a course-guide website, retain the original page URLs and the date you inspected them with your notes. The guide does not establish complete page coverage, access to restricted pages or automatic refresh. Inspect the imported set before treating it as a complete or current copy of the website.

Check export claims at the handoff you need. Humata’s API guide documents an authenticated insights-download request that returns a file URL. This developer route does not establish a browser citation export, bibliography file or export of a complete research conversation. The reviewed guide does not specify the downloaded file’s format or preserved fields. Verify the destination file and source locators before depending on it for coursework or a reference manager.

Keep a separate record of the original source, passage and correction even when the interface makes an answer easy to copy. A downloaded file and a usable evidence trail are different things. For a small permitted sample, check whether the export distinguishes source text, generated explanation and your own annotations. This is a suggested handoff check; we did not call the API or test an export.

Supporting documentation: [10] [11]

Imports, deletion and alternatives for the next research step

Humata's privacy policy says connected-service imports are copied into its account, and disconnecting the service does not delete those copies. It also says imported data may be shared with service providers, including cloud hosts and AI model providers such as Google Cloud, OpenAI and Cohere, to operate Humata; it says those providers are bound by confidentiality and data-protection obligations. The policy says connected-service data is not used to train generalized models. Humata's separate security page states that user data is not used to train its models. These are provider statements, not our certification or proof of which provider processes a particular file. Confirm permissions, retention and any institutional agreement before sharing unpublished work.

The connected-service section of the privacy policy distinguishes selecting files from ongoing access. For Google Drive, it describes access only to files chosen in the picker. For OneDrive, Dropbox and Box, it describes showing folder and file names to support selection, then downloading the selected files. It says access tokens serve the requested imports and are not retained for ongoing account access. These statements do not establish continuous synchronization with your storage service; inspect the imported copies and revoke access separately from deleting them in Humata.

The security FAQ distinguishes document data used for the model, which it says is not retained beyond 30 days, from dashboard data that remains until you request otherwise. Do not read that processing statement as automatic deletion of your uploaded library after a month. It also does not establish a deletion schedule for every backup or processor. Confirm the applicable terms before relying on a retention deadline.

Compare SciSummary for section-focused paper reading, Scholarcy for structured reading notes and literature matrices, and SciSpace for a broader discovery workflow. Start with a public document you understand in each candidate. Choose by the quality of recoverable evidence and the work needed to correct it, rather than the fluency or length of the generated response.

Supporting documentation: [5] [6]

Check team access separately from a website embed

Humata's permissions guide describes admin and member roles, teams and folder access restrictions. It says the number of users depends on the subscription. The invitation instructions also say invited users should not already have a Humata account. If your reading group already uses the service, confirm the current invitation or account-transfer path with support before buying seats around that workflow.

Check permissions separately from seat count. The pricing table lists department- and folder-level permissions on Team and Enterprise, but not Free or Expert. Being able to add another user does not establish that the plan provides the folder restrictions your reading group needs.

The security page says Humata now supports SAML 2.0 single sign-on, while the homepage says Okta, Google and SAML SSO is coming soon. The pricing table does not specify SSO availability by plan. These official pages conflict, so confirm current availability and plan entitlement with Humata before procurement. This review has not tested an SSO setup.

The White Labeling guide describes embedding a file or folder through Ask, Share and Embed, and places custom branding on Enterprise. That is a different task from inviting a collaborator into a restricted research folder. The guide does not establish whether every embed requires sign-in, which viewers can open its source documents, or how revocation works. Do not treat custom branding or a hidden webpage address as proof of restricted access.

Before sharing course readings, use a permitted public sample to check the experience as the intended recipient and as a signed-out visitor. Inspect which files and passages are exposed, then check what happens when access is removed. This is a suggested evaluation, not a permission test we performed. Share only material you have permission to distribute, and keep private notes outside the shared set until its audience is clear.

Supporting documentation: [1] [3] [6] [8] [9]

Pricing, free access and plan checks

Humata's current pricing table lists Free at $0 for one user and 60 pages per month; Expert at $9.99 per month for up to three users, 500 pages and $0.02 per additional page; Team at $49 per user per month for up to 10 users, 5,000 pages and $0.01 per additional page; and Enterprise at custom pricing with unlimited users and a custom page allowance. The homepage's Free plan card also says up to 10 answers, while its feature section advertises unlimited questions and the detailed pricing table omits an answer cap. Confirm the effective limit in the account before relying on Free for a course or research workflow. The table marks GPT-5 support in all four tiers; that feature label is not an accuracy or quality benchmark. OCR is not listed on Free or Expert. The FAQ still mentions Student, but the plan table does not show a Student price. Confirm the current bill, seat arrangement and allowances in the account before subscribing.

Confirm billing period, renewal terms, limits and regional availability on the provider’s current plan page before paying.

Earlier writing samples

These preserved PDF samples document the earlier version of the service. They are reference material, and do not establish how the current product performs.

Questions about Humata

Can Humata import a website?

Its official Website Import guide describes importing from a URL with a site-wide selection. The reviewed guide does not establish complete coverage, access to restricted pages or automatic updates. Check the imported material and retain the original source locations.

Can I export Humata citations as a bibliography?

The official API documentation describes an authenticated insights download returning a file URL. It does not establish a browser bibliography export, its file format or preserved citation fields. Confirm the exact handoff you need; this review has not tested an export.

Can I use Humata Free for a scanned PDF?

The pricing table does not include OCR on Free or Expert. Check whether the PDF already contains selectable text and whether the feature your scan needs is included in the chosen plan. A page allowance is not an OCR entitlement.

Can Humata answer questions across multiple PDFs?

The official Ask All guide describes selecting files or a folder with checkboxes, then choosing Ask All. Check that the selected set contains the documents you intend to compare, and verify each important claim in its cited file. Multi-document access does not establish complete retrieval or research coverage.

Is Humata free?

A $0 Free plan is listed with up to 60 pages per month. The homepage's plan card also says up to 10 answers, but its feature section advertises unlimited questions and the detailed pricing table omits the answer cap. Confirm the effective limit in your account before relying on the Free tier.

Does every Humata plan include GPT-5?

The current pricing table marks GPT-5 support in Free, Expert, Team and Enterprise. This is a provider-listed feature, not a benchmark of answer accuracy or a guarantee that the model can read every uploaded file. Check the current plan table and account controls.

Does Humata guarantee that cited answers are correct?

No independent accuracy guarantee is established here. Open the proposed source passage and check that its context, method and qualifications support the answer.

How can I limit additional-page charges?

The official guide describes a spending limit in Account settings under Plan. Review that control and the current overage rate before uploading beyond the included allowance.

Does disconnecting cloud storage delete imported files?

Humata's privacy policy says it does not. Imported files must be deleted within Humata if you want to remove those copies; check applicable retention and organizational requirements.

Does Humata send imported documents to AI model providers?

Humata's current privacy policy says connected-service data may be shared with service providers, including AI model providers, to operate the service; it says those providers have confidentiality and data-protection obligations. It separately says this connected data is not used to train generalized models. This review does not establish which provider processes a particular file. Ask your institution to review its applicable terms before uploading unpublished or identifiable work.

Is an embedded Humata chat the same as restricted team access?

The documentation describes folder restrictions and team membership separately from website embedding. The White Labeling guide does not establish all recipient authentication, source-document access or revocation behavior. Confirm those controls for your intended audience; this review has not tested them.

Sources and research notes

This guide synthesizes official product information with our editorial assessment of academic use. Suggested exercises are examples you can repeat; they are not results of tests we performed. We do not infer accuracy from a provider’s marketing claims.

  1. Humata: document questions, comparisons and source citations
  2. Humata: Grounded, Balanced and Creative response modes
  3. Humata: plans, page allowances, OCR and overage rates
  4. Humata: additional-page spending controls
  5. Humata: privacy policy and imported cloud-file copies
  6. Humata: security and data-use statements
  7. Humata: select files or a folder for Ask All
  8. Humata: member roles, teams, invitations and folder restrictions
  9. Humata: website embeds and Enterprise white labeling
  10. Humata: Website Import guide and site-wide URL selection
  11. Humata: API insights-download request and returned file URL

Found a changed feature or a factual error? Send the exact page and an official source to info@ai-essay.com. Read our editorial policy for how we handle corrections.