What Adjusters Miss in Altered Claim PDFs

This article is a snapshot – content was accurate as of April 2026. The product evolves actively; specific counts, examples, and detection rules may have changed since publication – see the changelog for the current state.
Insurance fraud in the United States costs an estimated $308.6 billion per year, according to the Coalition Against Insurance Fraud. That figure spans every line – auto, property, health, workers’ compensation, life. Altered PDF documents submitted as supporting evidence are one way that money is lost.
The alterations themselves are usually small: one total, one date, one treatment count. This article walks through that claim by claim – the four documents adjusters see most, what typically gets changed in each, what content review already catches, and what a file-level check adds on top.
It is written for the adjuster or intake supervisor on property, contents, liability and health claims, who has the documents in front of them and decides what to do next. Motor claims have their own document set – crash reports bought from a portal, tow invoices, total-loss valuations – and are covered in altered estimates and reports in auto claims. Engineers wiring verdicts into a claims system want routing altered claim PDFs to SIU, and the workers’ compensation variant is in altered medical and wage PDFs.
Anatomy of an Altered Claim, Document by Document
Every scenario below follows the same shape: a real document exists, someone changes one or two values, and the result is submitted as the original. The scenarios are illustrative composites, not individual cases.
Repair Estimate
What gets edited. The grand total, labour hours, part prices, sometimes the date of loss on the header so the estimate falls inside the policy period.
Scenario. A shop prepares a $4,200 estimate for hail damage. The claimant changes line-item quantities so the total reads $8,600 and submits it. The logo, technician name and VIN are all genuine, and the arithmetic is internally consistent because the totals were recalculated.
What content review catches. Prices out of line with regional labour and parts norms, a scope that does not match the photos, a total that does not match the shop’s own invoice once it arrives. This is the adjuster’s core skill and no file check replaces it.
What the file check adds. A signal on the estimate itself – whether the PDF shows evidence of being modified after the shop produced it – before anyone spends time arguing about the numbers. See repair estimate fraud detection for this document type in detail.
Medical Bill or Treatment Summary
What gets edited. Session counts, dates of service, billed amounts, diagnosis wording that supports a larger injury claim.
Scenario. A claimant with a soft-tissue injury has a clinic summary listing three physiotherapy sessions. The submitted copy lists twelve, with dates that fit neatly inside the claim window.
What content review catches. Treatment frequency that does not fit the injury, dates that collide with other records, a provider whose billing history does not match. Medical review and bill audit remain the deciding step.
What the file check adds. Whether the copy in the file is the one the clinic’s system produced or has been changed since. See medical bill tamper detection; for pharmacy and prescription records, medical and pharmacy document fraud.
Receipts and Proof of Loss
What gets edited. Item prices, quantities, purchase dates, merchant names on hotel, electronics and contractor receipts.
Scenario. A $340 hotel receipt for additional living expenses becomes $1,340. A $200 electronics receipt becomes $2,000. Receipts are processed in bulk and each one is small enough to skip close scrutiny, so a single changed figure rarely draws attention.
What content review catches. Totals that do not reconcile with tax, items implausible for the loss, merchants that do not exist at the stated address.
What the file check adds. A per-receipt answer at intake volume, including a separate outcome for receipts that do not come from an institutional system: when no modification is found and the file was made in consumer software, the result is inconclusive, and the practical next step is asking the merchant for the original. See fake receipt detection.
Police and Incident Reports
What gets edited. Date and time of the incident, location, the parties’ details, fault attribution, narrative wording.
Scenario. A claimant submits a collision report as a PDF. The incident time has been moved to fall within the policy period, or a fault line has been changed in the claimant’s favour. The rest of the report reads normally.
What content review catches. Mismatch with the claimant’s own statement, with weather and traffic records, with the other party’s account. When the report matters to liability, the strongest control is to request it directly from the agency.
What the file check adds. A report submitted as a file by the claimant is a copy. The check tells you whether that copy shows evidence of modification, which affects how urgent the agency request is. It does not replace that request: where the report bears on liability, obtain it from the agency whatever the result, since the check cannot tell you whether the agency’s own record is accurate.
One Claim, Four Documents: A Triage Example
Claims rarely arrive as a single document, and the checks do not have to give the same answer on each. This is an illustrative water-damage claim on a home, with invented amounts.
| Document | Source of the file | File check | What the adjuster does |
|---|---|---|---|
| Contractor estimate, $18,400 | Emailed by the claimant | modified | Holds the figure. Asks the contractor directly for the estimate as issued |
| Plumber’s invoice, $950 | Plumber’s invoicing system, forwarded by the claimant | intact | Normal handling. Content review as usual |
| Two hotel receipts for additional living expenses | One portal PDF, one photo saved as PDF | intact for the portal file, inconclusive for the photo (scanned origin) | Pays the portal receipt in the normal course. Asks the claimant for the hotel’s own folio for the other |
| Fire-brigade attendance letter | Photographed and saved as PDF | inconclusive (scanned origin) | Requests the letter from the brigade, since the file cannot answer the question |
The contractor then replies with its own copy of the estimate. There are two outcomes, and the file check does not choose between them:
- The contractor’s copy reads $11,800 and the line items differ from the submitted file. The case goes to SIU with the two files side by side and the neutral referral wording below.
- The contractor says two line items were revised after a second inspection and sends a supplement that explains the difference. The supplement is checked in turn and the claim proceeds on the supplemented figure.
Notice what did not happen. No document was labelled forged on the strength of the file alone, three of the four items moved on a routine request for the source, and only one figure was held. That is what the triage is for: it spends the investigative effort on the one document the file itself questions.
Questions to Put to the Claimant When a File Is Flagged
A flagged file is a reason to ask, and the tone of the first message decides whether the answer is useful. Keep it factual and neutral:
- ‘Can you send the document in the form you received it – the email attachment, or the file from the provider’s portal?’
- ‘Was the file opened, edited, annotated or re-saved before you sent it, for example to reduce its size?’
- ‘Who issued it, and can we ask them to send us a copy directly?’
- ‘Do you have the provider’s own invoice, folio or payment record for the same item?’
Answers to the second question are often innocent: people compress files, merge pages and sign forms. An evasive answer to the third, or a provider who does not recognise the document, is more informative than anything in the file.
An Intake Flow for Adjusters
The order matters. The file check is independent of the claim story, so it runs first. Content review – is this claim plausible? – stays with the adjuster and runs second.
- Receive the document as a file. Keep the PDF as submitted. A screenshot, photo or re-export of it cannot be checked the way the original file can; ask for the file.
- Run the file check before reading the content. One call per document through the API, or an upload through the web interface. Reading the numbers first anchors you on the story the document tells.
- Read the verdict as a statement about the file, not the claim.
intactmeans no structural evidence of modification was found. It does not mean the document is genuine or the claim is valid.modifiedmeans the file shows evidence of change after creation.inconclusivemeans the file’s origin does not allow a conclusion – for example, a scan, an online editor or consumer software – and the result carries a reason. - Run content review regardless. A file check does not judge whether a claim is plausible. An
intactrepair estimate can still be priced out of line; aninconclusivereceipt can be perfectly real. - Route by combination.
intactplus consistent content: normal handling.intactplus content that does not fit the claim: content review decides, whatever the file says.inconclusive: request the original from the issuer, then re-check what they send.modified: send the document to an adjuster for review and request the original from the issuing party. Whether to hold payment or refer to SIU is the reviewer’s decision, not the verdict’s – see the wording below.
The response fields, verdicts and status_reason values are covered in the API documentation; the checks themselves are listed on the how it works page.
SIU Referral Wording: Fact of Edit, Not Intent
A file check establishes that a document was changed after it was created. It does not establish who changed it, why, or whether the change was dishonest. Clinics re-save records, brokers re-export attachments, and claimants sometimes crop or annotate a document innocently. Referral notes should say only what was observed.
Wording that holds up:
Document received [date] as file [name]. Automated file analysis returned status ‘modified’ with markers [IDs]. This indicates the file was changed after its original creation. The analysis does not identify who made the change or why. Requested: original document from [issuer], and claimant’s account of how the submitted copy was produced.
Wording to avoid: ‘The claimant forged the estimate’, ‘fraudulent document’, ‘tampered with intent’. These are conclusions for the SIU investigation and, where relevant, counsel – not for the intake note. A referral that states the fact of an edit survives scrutiny; one that asserts intent can become the weak point in a denial.
For inconclusive results, the note is shorter: the file’s origin does not allow a conclusion, and the original was requested from the issuer.
What a File Check Does Not Cover
- Documents fabricated from scratch. A document that was never an original has no edit to find. When no modification is found and the file comes from consumer software, the result is
inconclusivewith a reason, which is a prompt to ask the issuer – not proof of anything. - Scans and photos. The result reports a scanned origin; the content has to be verified with the issuer.
- Content truth. Whether a repair price is reasonable, a treatment was medically necessary, or an incident happened as described is outside the file check entirely.
- Legitimate intermediaries. Third-party administrators, e-signature platforms and records systems rewrite files in normal workflows. Review early
modifiedresults by hand and record outcomes before automating any routing.
Structural analysis works as one layer next to predictive models, network analysis, adjuster judgement and SIU investigation. For the broader controls around it, see PDF fraud prevention best practices, and for checking a file when no original is available, PDF forensics without the original file. Receipt-heavy claims, such as expense and additional-living-cost reimbursement, are covered in altered receipts in expense claims.
The Cost Arithmetic
Insurance fraud adds an estimated $400–$700 per year to the average American family’s premiums, according to the Insurance Information Institute. One inflated estimate caught before payment – the gap between $4,200 and $8,600 in the scenario above – is worth more than the checks run on that claim; see current pricing. For context on the scale of PDF-based fraud across industries, see the document fraud statistics for 2026.
For the overview of how file-level checking fits into insurance claims operations, see the insurance claims fraud detection use case.