Extract data from tax return transcripts
Unstract reads IRS tax return, wage and income, and account transcripts, with the 4506-C or 4506-T that requested them, and returns the IRS's own income figures as checked data in your loan origination system.
| AMOUNT | |
|---|---|
| INCOME | |
| Wages, salaries, tips, etc | $86,240.00 |
| Taxable interest income | $410.00 |
| Business income or loss (Schedule C) | $22,850.00 |
| Total income | $109,500.00 |
| Adjustments to income | $1,614.00 |
| Adjusted gross income | $107,886.00 |
| TAX AND CREDITS | |
| Standard deduction | $31,500.00 |
| Qualified business income deduction | $4,247.00 |
| Taxable income | $72,139.00 |
| Tentative tax | $8,180.00 |
| Self-employment tax | $3,229.00 |
| Total tax liability | $11,409.00 |
| PAYMENTS | |
| Federal income tax withheld | $9,860.00 |
| Estimated tax payments | $2,400.00 |
| Total payments | $12,260.00 |
| Amount you owe / refund | ($851.00) |
AGI ties to 1040. wages 2k off? ask LO
This product contains sensitive taxpayer data.
{ "transcript": "Tax Return Transcript", "tax_period": "Dec. 31, 2025", "tracking_number": "108412736029", "ssn": "XXX-XX-6208", "filing_status": "Married filing joint", "wages": 86240, "business_income": 22850, "agi": 107886, "total_tax": 11409, "confidence": 0.96}
Trusted by tax and accounting teams around the globe
The IRS answers in its own format, and the loan file mixes it with everything else
A transcript is the most reliable income document in a loan file. It rarely reaches the processor as one clean page.
Transcripts the borrower printed from the IRS site and scanned
Two or three tax years, and the lines shift between them
A 4506-C completed in pen, with a tax year struck out
Request form and every transcript type in a single PDF
Wage and income transcripts that list every W-2 and 1099 filed
A processor reads each transcript line by line against the 1040, the W-2s and the 1003, and the loan waits in processing until every year ties out.
Unstract reads the transcripts and the 4506-C, extracts the IRS figures, compares them with the borrower's return and posts the result to the loan file.
Parse, extract and automate every transcript in the loan file
Parsing, extraction and automation are three separate products. Read transcripts with the first today, and add the other two when your pipeline needs them.
Parse tax return transcripts into text that keeps every line and amount together
LLMWhisperer reads IRS transcripts from IVES responses, the borrower's own printouts and scanned copies, and it keeps the layout. Each dollar amount stays beside its line description, the masked SSN and tax period stay in the header, and a wage and income transcript keeps each payer's W-2 or 1099 as its own block.
Pass the text to your own LLM, or make parsing the first step of a larger loan workflow.
Learn more about LLMWhispererGet structured data from tax return transcripts and the 4506-C behind them
Name the transcript lines you need, and Document Studio returns each transcript as the same JSON record for every tax year and every transcript type. It works as a Form 4506-C data extractor too: the borrower's name, SSN, address, the transcript type on line 6 and the tax years requested come back as fields, so a request with a missing signature or the wrong year is caught before it goes out. The figures are compared with the 1040 and the 1003, and every difference comes back flagged.
Deploy it as an API, run it as an ETL pipeline, or call it from a larger orchestrated workflow.
Learn more about Document StudioRun mortgage loan origination with the IRS transcript as the income check
Describe in plain English how your team uses tax return transcripts, and Flow's agents build the workflow. The loan file arrives with the 1003, the 1040s, W-2s, pay stubs and the IVES responses. Each document is indexed to the loan, the transcripts are read with the returns they cover, and the AGI, wages and Schedule C income on each transcript are compared with the borrower's 1040 and the income on the 1003. Clean files land in your loan origination system, such as ICE Encompass or Dark Matter Empower. You approve the workflow once, and it checks every borrower's transcripts the same way.
You set the points where a loan processor takes over: transcript income outside your tolerance against the 1003, a tax year the IRS has no record of, a missing month or statement elsewhere in the file, or a value the model read with low confidence.
Learn more about FlowRead your transcripts today. Automate the loan file next.
Start for free to parse and extract transcripts from your own loan files. Automating the income check across the whole file begins with a consultation.
IRS transcripts, the forms that request them, and the returns they confirm
Documents
- Tax return transcripts
- Wage and income transcripts
- Account transcripts
- Record of account transcripts
- Verification of non-filing letters
- Form 4506-C IVES request for transcript
- 4506-T Request for Transcript of Tax Return
- Form 4506-T-EZ short form request
- Form 8821 tax information authorization
- Form 2848 power of attorney
- Business return transcripts (1120, 1120-S, 1065)
- Form 1040 individual tax returns
- W-2 forms
- 1099 forms
- Schedule C and Schedule E
- IRS notices and letters
Fields captured
Processes these transcripts feed
Built for the teams that check a borrower's income against the IRS
Each IVES response read the day it returns, with the AGI and wages tied to the 1040 before the file moves to underwriting.
Two years of transcript figures beside the return and the 1003, so a condition names the exact line that disagrees.
Account transcripts read for balances, penalties and installment agreements, so a client's IRS history is on one screen.
4506-C requests checked before submission and thousands of transcripts returned as one consistent record for every lender you serve.
Income verified against the IRS without the line-by-line reading
Built for the IRS's own record of a borrower's income
A transcript is the IRS's own record of a person's income, filed with their Social Security number. Lenders may request it only with the taxpayer's signed consent on a Form 4506-C, and a leaked copy gives anyone the facts needed to impersonate that taxpayer to the IRS.
- Encrypted in transit and at rest.
- Borrowers' transcripts are never used to train models.
- Access by role and loan: processors and underwriters see their own files, and every correction is logged against the transcript.
- Retention you control: transcripts and 4506-C forms are deleted on the schedule your compliance team sets.
- Self-hosted options keep transcripts inside your own environment for lenders and verification providers that require it.
What lenders ask about reading IRS transcripts
Unstract gives you an IRS tax transcript API for reading, not for ordering. You request transcripts through IVES or the taxpayer as you do today, then POST each transcript PDF to a Document Studio endpoint and get AGI, wages, total tax and the other lines back as JSON.
Yes. LLMWhisperer reads tax return, wage and income, and account transcripts with their layout kept, and Document Studio returns the lines you name as fields. Your processors review the flagged differences instead of reading every page.
Yes. Name the lines you want, such as adjusted gross income, wages, business income, total tax or the refund, and each transcript comes back with those fields filled for every tax year in the file.
Form 4506-C processing covers checking the signed request, sending it through IVES and reading the transcripts that come back. Unstract reads the 4506-C itself, flags a missing signature or tax year, and reads each returned transcript into your loan file.
LLMWhisperer reads a 4506-C filled in by pen, signed and scanned or faxed back, including the check boxes on line 6 and the tax years on line 8. Extraction then marks a value it could not read with confidence for a processor to confirm.
A good Form 4506-C parser confirms the taxpayer's name and SSN match the loan file, that a transcript type and the right tax years are entered, and that the form is signed and dated within the IRS's window. Unstract returns each as a field you can test in your rules.
Tax return transcripts, wage and income transcripts, account and record of account transcripts, verification of non-filing letters and business return transcripts for 1120, 1120-S and 1065 filers, plus the 4506-C, 4506-T and 4506-T-EZ forms that request them.
Yes. When the 1040 is in the same loan file, the transcript's AGI, wages, Schedule C income and total tax are compared line by line, and any difference beyond your tolerance comes back flagged with both figures.
Yes. Describe the check in plain English and Flow's agents build the workflow: index the file, read the transcripts and returns, compare the figures and raise conditions. You approve it before it runs, and a loan processor sees only the files you told it to stop.
Unstract is SOC 2, ISO 27001, GDPR and HIPAA compliant. Transcripts are encrypted in transit and at rest, never used to train models, and kept only as long as you choose, and self-hosted options are available. The Trust Center explains each safeguard.
Bring a few closed loan files with their transcripts
IVES responses, borrower printouts and signed 4506-C forms. We run them live and show every figure, every comparison with the 1040 and every flag.