The output#1
Retab
It returned a clean nested bank-statement JSON with correct account, balance, disclaimer, and transaction-type fields, and the result was immediately usable. The main blemish was the summary count, which overstated the total number of transactions.
7335c86f94a340cf8443685a9fca91ce.png
The output#2JSONresearch-media-unstract-bank-statement-output-f0bd422ffa9e.jsonopen raw ↗ Unstract
It extracted the full bank statement structure and all 51 transactions, but the summary transaction count was wrong, so the result was strong overall with one notable derived-field mistake.
research-media-unstract-bank-statement-output-f0bd422ffa9e.json
The output#3
Landing AI
The bank statement run produced clean nested JSON, kept all 51 transactions as separate rows, and added useful transaction types; the main weakness was that it assigned new sequential transaction IDs instead of preserving the source IDs.
research-media-landing-ai-extracted-bank-statement-c4910d90b9f3.png
The output#4
Nanonets
It accepted the statement and returned a valid JSON export with the right top-level structure and accurate header values, but the transaction section was unreliable: several rows were merged, dates and transaction types were missing, and the transaction count was off.
a6afda6d46f44a9892ef7792ed6fb1f4.png
The output#5
LlamaParse
It produced a well-structured bank statement with the right major sections, but some transaction metadata was left blank, the transaction count was inconsistent, and the derived transaction IDs and types were not filled in.
research-media-llamaparse-bank-statement-extracted-deta-1c069bb9dc7a.png
The output#6
Reducto
It returned a usable nested bank-statement JSON with citations and several correct header fields, but the transaction table drifted in count, key transaction fields were missing, and the summary was not reliable for reconciliation.
8bb08fbfe42e417db6ccbeb6262e87f7.png
The output#7
Datalab
It produced nested bank-statement JSON with citations, but the transaction layer was shaky: it overcounted rows, merged adjacent entries, misdated at least one transaction, and left transaction IDs blank.
cef4de5010ba43ee8f3a32c9127421a6.png
The output#8
Extend AI
It produced a solid structured bank statement with balances, rewards, and transaction rows, but the totals were off, one transaction identifier stayed blank, and the section order was not preserved.
research-media-extend-ai-bank-statement-summary-a4a38bf7978f.png
Comments (0)