Pay stub and tax form extraction software: 8 tools for teams that check income
For lenders, loan processors, housing compliance and income verification teams. See which tax forms each tool reads besides pay stubs, which ones check the forms against each other, and what each one costs.

Key takeaways
- Pay stub and tax form extraction software reads the pay stubs, W-2s, 1099s, tax returns and K-1s that applicants send. It turns each box and line into data your team can check.
- The tools differ most on tax forms. Google's ready-made parsers now cover only pay stubs and W-2s, not 1099s, 1040s or K-1s. Azure and Amazon list no K-1 model, while Docsumo, Sensible and LlamaParse read K-1s.
- Many tools read the forms, but fewer run checks between forms. Docsumo runs them by rules you set, on its Enterprise plan. Amazon, Azure and Google don't describe such checks on their sites, so your engineers would build them.
- The cloud APIs and developer tools publish prices. Amazon Textract charges $0.07 a page for lending documents, and Sensible starts at $449 a month. Docsumo, Ocrolus and Informed.IQ quote theirs.
- Test each tool with a recent pay stub and a full tax return from your own files, not a single clean W-2.
On this page
- What pay stub and tax form extraction software does
- One couple's tax packet, form by form
- The 8 pay stub and tax form extraction tools
- Pay stub and tax form extraction software compared
- Checks to ask every vendor to show
- If you check income for affordable housing
- Match the tool to the forms your applicants send
- Frequently asked questions
Pay stub and tax form extraction software reads the pay stubs and tax forms that applicants send. The tax forms include W-2s, 1099s, tax returns and K-1s. It turns each one into data your team can check. The tools differ on two things. One is which tax forms they read. The other is whether they check the forms against each other.
We compared 8 tools for lenders, housing compliance teams and other teams that verify income. Docsumo is our product. Every other vendor fact comes from the vendor's own website, checked on October 30, 2026. We didn't test the tools on our own files.
Which team are you?
What pay stub and tax form extraction software does
Applicants rarely send one document at a time. A mortgage or rental application often arrives as one PDF with two pay stubs, last year's W-2 and a full tax return. The software has to split that file into its forms before it can read any of them.
- Pay stubs
- W-2s
- 1099s
- Form 1040
- K-1s
- 01Split the file by form
- 02Read each box and line
- 03Check the forms against each other
- 04Send exceptions to a person
Every tool on this list reads documents. The tools differ on which forms they read and on what happens next.
Pay stubs are hard because each payroll provider lays them out its own way. Lenders also care about the date. Fannie Mae says the most recent pay stub must be dated no earlier than 30 days before the initial loan application date. It must also include all year-to-date earnings. So the software has to read the pay date and the year-to-date column, not just this period's pay.
Tax forms are the opposite problem. The IRS sets each layout, but there are many forms. A self-employed applicant's return can hold several schedules and a K-1 from each business they own a share of.
One couple's tax packet, form by form
The figure shows how Docsumo reads one married couple's 2025 tax packet. Pick a form to see its fields and the checks run on it.
Individual and wage forms
Business returns
Form W-2: Wage and Tax Statement
Employers send one to each employee every January.✓ 3 passedWages one employer paid one employee in the tax year, and the taxes withheld from them.
What Docsumo reads
- Employee, box e
- Elena M. Ruiz
- Employee's SSN, box a
- XXX-XX-4127
- Employer, box c
- Front Range Dental Group PC
- Employer EIN, box b
- 84-3150792
- Wages, box 1
- 136,500.00
- Federal tax withheld, box 2
- 24,180.00
- Social security wages, box 3
- 160,000.00
- 401(k) deferrals, box 12a code D
- 23,500.00
Checked against
- Form 1040Box 1 against wages on line 1a✓ 136,500 on both
- Form 1040Box 2 against withholding on line 25a✓ 24,180 on both
- ItselfBox 3 against box 1 plus the 401(k) deferrals✓ Adds up
SSNs shown truncated. Field names and the output schema are configurable.
Elena Ruiz draws a salary from Front Range Dental Group PC, an S corporation, and owns 25% of it. Her W-2 shows $136,500 in box 1. The IRS instructions say line 1a of Form 1040 is the total from box 1 of the W-2s. The couple's 1040 shows $136,500 on line 1a, so that check passes.
Her share of the company's profit takes three forms to trace. The company's Form 1120-S reports $412,000 of ordinary business income on line 22. Her K-1 gives her 25% of that, $103,000, in box 1. The IRS tells shareholders to report box 1 income on Schedule E, line 28. The couple's Schedule E shows $103,000 there too.
Her spouse, David, runs a design studio on his own. Two businesses paid him on Forms 1099-NEC, for $28,500 and $18,300. That's $46,800, but his Schedule C reports $41,300 of gross receipts on line 1. The IRS asks anyone whose 1099-NEC totals are higher than line 1 to attach a statement explaining the difference. So the check should ask the applicant for that statement, not mark the return as wrong.
Every tool below can read the boxes and lines on the forms it supports and send them back as data. Fewer tools run checks like these, which compare one form with another. Docsumo calls these checks cross-document validation, and they come with the Enterprise plan.
The 8 pay stub and tax form extraction tools
Docsumo is our product, so it comes first, with a plain line on what it doesn't do. The others follow by kind. Lending platforms come first, then developer tools, then the three cloud APIs.
1. Docsumo
Lending platformOur product- Pay stubs
- Weekly, biweekly, semi-monthly and monthly pay stubs, as PDFs or images, printed or handwritten
- Tax forms
- The W-2 and the 1099-NEC, MISC, R, G, INT and DIV. The 1040 with Schedules 1, A, C, D and E. Forms 1065, 1120 and 1120-S, and the K-1s from the 1065 and 1120-S. Also the W-9, the 4506-C, tax return transcripts and state returns. See tax form extraction
- Checks
- You set up checks inside one document as workflow steps, such as gross pay minus deductions equals net pay. Checks between forms, and against the application, are cross-document validation on the Enterprise plan
- Review
- When Docsumo is less sure of a field than the level you set, that field goes to your own reviewers. Clicking a field highlights the line it came from, and each correction improves the model
- Output
- API and webhooks, or Excel. One upload can be up to 500 pages, and Docsumo runs in the cloud only
- Pricing
- A free 14-day trial for up to 1,000 pages; Business and Enterprise plans are quoted. Splitting a file by form is on the Business plan
- 99%field-level accuracy on 250+ document types
- 95%+of documents AI Approved, so no one had to touch them
- <5 minper document, down from 2+ hours
2. Ocrolus
Lending platform- Pay stubs
- Yes, in its Income product
- Tax forms
- W-2s, 1099s, 1040s and supporting schedules. Its product docs don't name K-1s
- Checks
- Income calculations. For mortgages, its Inspect product compares borrower documents with the loan application data in Encompass
- Review
- Ocrolus's own data professionals handle the unusual cases its software can't resolve
- Pricing
- Not published. It charges per document processed and offers a free trial of up to 100 pages
3. Informed.IQ
Lending platform- Pay stubs
- Yes
- Tax forms
- W-2s and tax returns. Its home page doesn't name 1099s or K-1s
- Checks
- Income under each lender's own policy, and stipulations. Informed says it calculates and clears stipulations in under 30 seconds
- Output
- It connects your loan origination system (LOS) with the sources your applications come from. It offers an API and a standalone portal
- Pricing
- Not published
4. Sensible
Developer platform- Pay stubs
- Yes, including layouts from ADP, Paychex, Gusto, QuickBooks, Paylocity and Rippling
- Tax forms
- The W-2; the 1099-MISC, NEC, INT, DIV and R; the 1040 with Schedules C and E; and Schedule K-1. Its tax page also names the 1120
- Checks
- Rules inside each document, such as gross pay minus deductions equals net pay. If Sensible is less sure of a field than the level you set, a person reviews it
- Output
- JSON that matches your schema, with each value linked to its page and position
- Pricing
- Growth is $449 a month, billed annually, for 750 documents a month. Scale is $1,349 a month for 3,200. Each plan has a 14-day free trial
5. LlamaParse
Developer platform- Pay stubs
- Yes, with earnings, deductions and year-to-date figures
- Tax forms
- W-2s and 1099s, naming the 1099-NEC and 1099-MISC. K-1s, on LlamaIndex's own K-1 list
- Checks
- Its pay stub page describes math checks from gross to net pay. Its W-2 page mentions flagging mismatches against pay stubs, but it doesn't describe W-2-to-1040 or K-1 checks
- Output
- JSON, with the page and position of each value
- Pricing
- 10,000 free credits to start. 1,000 credits cost $1.25, and paid plans start at $50 a month
6. Amazon Textract Analyze Lending
Cloud API- Pay stubs
- Yes, as payslips
- Tax forms
- The W-2 and W-9, and the 1040 with Schedules C, D and E. Seven kinds of 1099, from the 1099-NEC to the SSA-1099. Forms 1065, 1120 and 1120-S. Schedule K-1 isn't on its list
- Checks
- Its site doesn't describe checks between forms. Your team would build them
- Review
- Amazon's human review service, A2I, is closed to new customers
- Pricing
- AWS's pricing example for its US West (Oregon) region shows $0.07 a page for the first million pages a month, then $0.055 a page. New AWS customers get 2,000 free pages a month for 3 months
7. Azure Document Intelligence
Cloud API- Pay stubs
- Yes, US pay stubs in English
- Tax forms
- The W-2, W-4, 1098, 1095-A and 1095-C, and more than 20 kinds of 1099. The 1040 with its schedules, including C, E and SE. Its tax page names no Schedule K-1 model
- Checks
- Its site doesn't describe checks between forms. Your team would build them
- Pricing
- $10 per 1,000 pages for ready-made models. A free tier covers 500 pages a month, but it reads only the first two pages of each PDF
8. Google Document AI
Cloud API- Pay stubs
- Yes, but the Pay Slip Parser reads only the first valid pay slip in a multi-page file
- Tax forms
- The W-2 only. The W2 Parser lists the 2018, 2019 and 2020 forms as its supported versions
- Checks
- Its site doesn't describe checks between forms. Your team would build them
- Pricing
- $0.30 for each pay slip or W-2 the parsers process
We also looked at three tools that came up in our research. Veryfi has W-2 and W-9 APIs, but its site shows no pay stub product. Doxis AI.dp, the product on Klippa's site, reads W-2s, but its payslip page doesn't mention US pay stubs. Nanonets has 1099 and tax statement pages, but no pay stub page.
Two other groups do a different job. K1x and Additive read K-1s for tax preparers, who need the figures for a return rather than an income check. Argyle, Truv and Plaid mainly get income data from a link to the applicant's payroll or bank account. Each also accepts uploaded documents.
Pay stub and tax form extraction software compared
The table follows each vendor's own site, checked October 30, 2026. "Not described" means we couldn't find it on the vendor's pages.
| Tool | Pay stubs | Tax forms | Checks between forms | Pricing |
|---|---|---|---|---|
| Lending platforms3 tools | ||||
| DocsumoOur product | Yes | Rules you set, on the Enterprise plan | Free trial; quoted plans | |
| Ocrolus | Yes | Income calculation; mortgage file checks | Not published; free trial | |
| Informed.IQ | Yes | Income under your policy; stipulations | Not published | |
| Developer platforms2 tools | ||||
| Sensible | Yes | Inside each document; cross-form checks not described | From $449 a month | |
| LlamaParse | Yes | Briefly, on its W-2 page | Free credits; from $50 a month | |
| Cloud APIs3 tools | ||||
| Amazon Textract | Yes | Not described | $0.07 a page | |
| Azure Document Intelligence | Yes, US only | Not described | $10 per 1,000 pages | |
| Google Document AI | First stub per file | Not described | $0.30 a document | |
Checks to ask every vendor to show
Every tool reads the boxes. Whether your team still has to open the file depends on which checks the tool runs. Ask each vendor which of these it runs and which ones your engineers would have to build.
- The pay stub is recent enoughFannie Mae wants the most recent stub dated no earlier than 30 days before the application. Your own program may set a different limit.
- The pay stub adds upGross pay minus deductions equals net pay, and this period's pay rolls into the year-to-date total.
- Pay to date fits last year's W-2Compare year-to-date pay with last year's W-2 pay for the same number of months. If it's much lower, ask the applicant why.
- W-2 wages match the returnBox 1 of each W-2 adds up to line 1a of Form 1040.
- K-1 income reaches Schedule EBox 1 of each K-1 matches line 28 of Schedule E. It should also match the owner's share of the income on the business return.
- 1099s match Schedule CIf the 1099-NEC totals are higher than line 1, the return should carry a statement that explains why.
- Names and tax ID numbers agreeThe employer's name and EIN are the same on the pay stub and the W-2, and the applicant's name matches on every form.
Our take. We'd start with the pay stub, because nearly every wage earner's file has one and it changes every pay period. Once the software reads and checks stubs without a person fixing them, add W-2s. Then add the full returns and K-1s that self-employed applicants send. Their forms vary the most.
If you check income for affordable housing
Housing compliance teams read the same forms for a different decision. They use them to work out a household's annual income for a tenant income certification. Our tenant income certification guide walks through the form itself.
Docsumo sorts a tenant file into its documents and reads each one. These include pay stubs, W-2s, tax returns, benefit letters and bank statements. On the Enterprise plan, it checks the documents against the application and each other, by rules you set. A calculation step can annualize income and compare the total with the limit you set. Your specialist reviews and signs the certification. If an adult's W-2 is missing, the Document Collection and Email agents can email the applicant for it. Every override is kept in the audit log.
Match the tool to the forms your applicants send
If most of your applicants are wage earners, most tools here can read their pay stubs and W-2s. In that case, choose by price and by who builds the checks, your team or the vendor. If self-employed applicants send full returns with schedules and K-1s, the list gets shorter. Docsumo, Sensible and LlamaParse read K-1s. If you want the forms checked against each other without building the checks yourself, Docsumo does that on its Enterprise plan. Ocrolus and Informed.IQ calculate income for you. For a whole loan file reviewed by AI agents, see our list of AI loan file review software.
Book a demo and bring a recent pay stub and a full tax return, with the personal details masked. You can also start a free trial and run your own checks.
Frequently asked questions
How do lenders verify pay stubs?
They check that the stub is recent and that its math adds up. They also check that its pay agrees with the W-2, the tax return and the bank deposits. Fannie Mae requires the most recent pay stub to be dated no earlier than 30 days before the application, with year-to-date earnings. Software reads the stub and runs the checks, and a person reviews what fails.
How do landlords and housing compliance teams verify pay stubs?
They read each adult's pay stubs and compare the pay with the application and the other income documents. Affordable housing teams also turn the pay into an annual figure for the tenant income certification. Docsumo reads the whole tenant file and checks it by rules you set, on the Enterprise plan.
What software extracts data from K-1s?
Among the tools here, Docsumo, Sensible and LlamaParse say they read Schedule K-1. Azure Document Intelligence and Amazon Textract don't list a K-1 model. Tax preparers also use K1x and Additive, which are built for preparing returns rather than checking income.
Can Google Document AI still read 1099s?
Not with a ready-made parser. Google said its ready-made 1099, W-9 and 1120 processors would be discontinued on June 30, 2026. For pay stubs and tax forms, its current list has only the Pay Slip and W2 parsers. Other forms need its Custom Extractor.
How much does pay stub and tax form extraction software cost?
Amazon charges $0.07 a page for Analyze Lending, and Azure charges $10 per 1,000 pages for its ready-made models. Google charges $0.30 for each pay slip or W-2, and Sensible starts at $449 a month, billed annually. Docsumo, Ocrolus and Informed.IQ quote their plans, and Docsumo has a free 14-day trial for up to 1,000 pages.
Sources
- Docsumo: pricing (free trial, Business and Enterprise plans)
- Docsumo support: All supported documents (income, employment and tax forms)
- Docsumo support: USA payslips
- Docsumo: pay stub extraction
- Docsumo: tax form data extraction (the example tax packet)
- Fannie Mae Selling Guide B3-3.2-01, Standards for Employment and Income Documentation
- IRS: 2025 Instructions for Form 1040 (line 1a, total from Form(s) W-2, box 1)
- IRS: 2025 Shareholder's Instructions for Schedule K-1 (Form 1120-S), box 1
- IRS: 2025 Instructions for Schedule C, line 1 (Forms 1099-NEC)
- Ocrolus docs: product SKUs (Income and Tax Forms)
- Ocrolus: mortgage (Inspect)
- Ocrolus: platform (use cases and human verification of edge cases)
- Ocrolus: FAQ (pricing and free trial)
- Ocrolus: home page (AI workflow and analytics platform for lenders)
- Nanonets: Form 1099 data extraction
- Plaid: Income docs (bank, payroll and document income)
- Informed.IQ: home page
- Sensible: pay stub extraction
- Sensible: tax form extraction
- Sensible: pricing
- LlamaIndex: pay stub verification (LlamaParse)
- LlamaIndex: W-2 form OCR (LlamaParse)
- LlamaIndex: 1099 form OCR (LlamaParse)
- LlamaIndex: best AI for K-1 forms (May 28, 2026)
- LlamaIndex: pricing
- AWS: Analyze Lending document types
- AWS: Amazon Textract pricing
- AWS: Amazon A2I is no longer open to new customers
- Microsoft Learn: Document Intelligence payStub model
- Microsoft Learn: Document Intelligence US tax document models
- Microsoft: Azure Document Intelligence pricing
- Google Cloud: Document AI processor list (updated October 26, 2026)
- Google Cloud: Document AI release notes (February 17, 2026)
- Google Cloud: Document AI pricing
- Veryfi: W-2 OCR API
- Doxis AI.dp (formerly Klippa): payslip OCR
- Doxis AI.dp (formerly Klippa): W-2 OCR
- K1x: home page
- Additive: home page
- Argyle: home page
- Truv: home page
- Plaid: income verification
First published .