A convincing bank statement still needs a reviewable source

The difficult application is often the one that looks routine: a familiar bank logo, the right name and regular salary credits. The useful question is whether the submitted evidence supports a particular claim, and what needs independent confirmation before the application moves forward.

Checkr’s vendor-sponsored survey of 1,400 US landlords and property managers, conducted 13–18 August 2026, found 67% perceived fraud as harder to detect. This self-report is not a fraudulent-application rate.

FinCEN’s November 2024 deepfake alert concerns financial institutions and falsified identity evidence. It provides historical context, not rental-market prevalence data.

A practical process turns a vague suspicion into a reproducible observation: which page, which field, which difference, and which follow-up could resolve it.

First establish which documents you may collect

Confirm the local rules and the stage of the application before requesting financial evidence. Asking for more pages is appropriate only when the document itself is permitted and the additional information is necessary. A tool’s ability to read a file does not create a right to collect it.

For French principal-residence applications governed by the 1989 tenancy law, the permitted-document list does not include bank statements. Use the authorised evidence instead; do not request a full transaction history for this workflow.

For an international portfolio, use a document checklist for each market. Define the purpose, authorised reviewers and retention period. Keep unrelated spending and sensitive personal details out of the review. The examples below apply only where receiving and processing the relevant evidence is permitted.

Separate the file, the income claim and the rental decision

Three different questions can hide behind the instruction “verify the statement”. Keep them separate so a narrow technical result cannot turn into a broader assurance.

  • File consistency: do the readable dates, identifiers, transactions and balances agree within the supplied document?
  • Income support: what independently supports the claimed salary, source of funds or relationship between the account and applicant?
  • Application assessment: does the complete application meet the documented, lawful rental criteria? This remains a separate accountable decision.

Begin with the source and record what is missing

Where collection is allowed, an original exported PDF normally gives more context than a cropped screenshot. Retain the relevant statement period and the pages needed for the agreed check. If a page is absent or redacted, describe the resulting limitation rather than silently treating the statement as complete.

Save the received file under controlled access, with its receipt time and your own case reference. Annotate a working copy. If a replacement arrives, preserve the relationship between the two versions so another reviewer can see what changed.

Never ask for banking passwords, one-time codes or access to an applicant’s device. Any independent confirmation should use an appropriate authorised process and a contact channel verified separately from the document under review.

Five checks that produce useful questions

Start with facts you can point to in the file. Before escalating a result, compare extracted values with the visible source: decimal separators, minus signs and debit/credit columns can be read incorrectly.

  • Identity of the record: note the account-holder name, masked account reference, currency and statement period. A matching name alone does not prove account ownership.
  • Page coverage: look for stated page counts, gaps in the period and missing opening or closing information. Record uncertainty about completeness.
  • Balance arithmetic: reconcile opening balance plus credits minus debits only when the necessary entries and their directions are available.
  • Timing: distinguish the period a salary covers from the day it is paid. Monthly, four-weekly and irregular payments should not be compared as identical cycles.
  • Repeated details: compare references and names consistently. A changed spelling, a formatting difference or editing-software metadata needs context before it becomes significant.

Worked example: a €250 difference worth explaining

This invented example illustrates a review question. It is not a customer case, a live result or an accuracy test. Assume the supplied period contains an opening balance of €1,200, credits of €2,800 and debits of €2,100. The stated closing balance is €2,150.

The supplied arithmetic gives €1,200 + €2,800 − €2,100 = €1,900. The stated closing balance is therefore €250 higher. That is a specific discrepancy; it does not establish its cause.

Check whether all relevant pages are present, whether a separate adjustment explains the amount, and whether the extraction reversed a sign or missed an entry. A matching total would establish consistency of those amounts only. Fabricated amounts can reconcile too.

  • Observation: the stated closing balance exceeds the calculation from the supplied entries by €250.
  • Source: record the pages and exact amounts used, together with any unreadable entries.
  • Next step: check extraction and completeness, then request a permitted correction or explanation if the discrepancy remains.

Ask a question the applicant can actually answer

A useful request is neutral and precise: “The closing balance differs from the calculation using the entries provided. Could you explain the difference or supply the missing permitted evidence?” Say which period and amounts you mean.

Avoid a generic demand to “prove this is not fake”. It gives an honest applicant little guidance and leaves your reviewer without a defined resolution. A wrong month, incomplete export or genuine correction may explain the problem.

Record the response, any corrected evidence and the reviewer’s reason for closing or escalating the issue. Keep the original finding visible alongside its resolution. A later explanation should add context rather than rewrite the original file.

Compare related evidence without inventing a match

If the permitted review depends on a payslip and another supplied record, compare equivalent periods and amounts. Gross salary, net pay and a bank credit are different quantities. Reimbursements, deductions, split payments and timing can explain a difference.

For a multi-file question, DeepfakePolicy Cross-check compares supplied evidence in a separate workflow and tariff. It can help organise matches and differences; it does not independently contact the employer or bank.

Keep outcomes that preserve what you actually learned

A clear outcome is more useful than an unqualified “passed”. Another reviewer needs to know the scope completed, the evidence unavailable and the issue still open.

  • No inconsistency found in completed checks: continue with any independent evidence the process requires.
  • Further evidence needed: specify the missing or unreadable information and the permitted way to resolve it.
  • Discrepancy recorded: cite the source, confirm extraction, request clarification and route unresolved findings for human review.
  • Keep a dated decision record with the application criteria, follow-up and responsible reviewer. Apply the same relevant standards consistently.

Choose a document check or API that fits the question

DeepfakePolicy’s single-document service reviews supported fields, metadata and calculations, with findings linked to the submitted evidence. It does not certify that the bank issued the statement or that the applicant owns the account.

Use the document workspace for a file review, or explore the document API for repeated processing in your own application. Start with the contract and fixed sandbox fixtures to test the integration; use representative authorised files to evaluate the live results separately.

Before production use, decide how your system will show an incomplete, unreadable or failed result. A timeout is not a clean document. Keep the report attached to your case, check the cited values and follow the current security and retention terms.

Make the next step clearer before the application moves on

The commercial value of a statement check is a repeatable review at the moment unfamiliar evidence enters your process. A useful report narrows the follow-up: an amount to explain, a page to replace or an independent fact still to confirm.

Start with one relevant, authorised document. Inspect the supported checks and the evidence behind each observation. Then choose a browser review, a multi-file comparison or an API integration to match the question your team actually needs to answer.

FAQ

Frequently asked questions

Can a bank statement checker prove that a statement is fake?

A discrepancy can justify further review, but missing pages, extraction errors or a genuine correction can also explain it. Matching totals do not prove authenticity. Check the source and obtain independent support where the application requires it.

Can an AI image detector verify rental income?

No. Image-generation signals do not establish salary, account ownership or the identity of the bank that issued a statement. Use a document review for supported fields and calculations, and verify the relevant income claim separately.

Should a rental team request every applicant’s full bank history?

No. Check local rules, the application stage and necessity first. In French principal-residence applications covered by the 1989 tenancy law, bank statements are outside the permitted supporting-document list. Use authorised evidence instead.

Which DeepfakePolicy workflow should I choose?

Use a single-document check for supported observations within one file. Use Cross-check when the question depends on comparing supplied files. Explore the document API if you need to integrate repeated reviews into your own application.

Automated results require source and context review.

Continue with independent verification.

Explore bank statement checks
Sources

Primary reading

We use original standards, regulators, public institutions and research papers wherever possible. Sources were last checked on 18 October 2026.