Symptoms
Totals and invoice numbers become inaccurate text after OCR though the image is readable. This guide addresses that particular contrast, not every fault that can occur with the product or service.
- Observable failure: Totals and invoice numbers become inaccurate text after OCR though the image is readable.
- Control test: Compare 5 selected numbers in the scan against recognized text.
- A confirmed correction must meet this outcome: Test receipt totals are searchable and match the image character for character.
A working state in another app or device is a useful control only when the account, source material and operating conditions are comparable. If they differ, that comparison may produce a misleading conclusion.
Why It Happens
A PDF can contain page images, a searchable text layer, annotations and form fields that behave independently. Visual correctness in one viewer never guarantees that hidden text, signatures or export layers are correct.
The symptom alone cannot distinguish low scan resolution from incorrect recognition language. Evidence related to glare or compression artifacts changes the next action, so record the comparison before applying a fix.
Three competing explanations
- Low scan resolution.
- Incorrect recognition language.
- Glare or compression artifacts.
These are hypotheses. Do not infer a hardware defect, account restriction or network outage from one status indicator alone. A result that supports low scan resolution does not automatically exclude incorrect recognition language; the comparison needs to separate them.
Before Making Changes
Work on a duplicate PDF and retain the unmodified original, especially when text recognition, signatures or redaction are involved. Never trust OCR totals for financial reporting without review.
Preserve the current state: note the affected product or account, current mode and time of the failed attempt. If settings will be changed, photograph or write down their original values. Keep copies of any unique files or evidence before overwriting anything.
Use a non-sensitive sample whenever possible. If the issue involves a shared device, building system, workplace account or regulated equipment, obtain the appropriate owner or administrator permission first. Stop if the test would compromise security, electrical protection, access or an animal's welfare.
Problem-Specific Diagnosis
Run this discriminating test before restarting or resetting the entire system: Compare 5 selected numbers in the scan against recognized text. Compare the observed result against the three paths below.
| Observation or suspected layer | Interpretation | Next controlled check |
|---|---|---|
| Low scan resolution | If the failing state changes under the comparison, low scan resolution is plausible. This is not yet proof. | Repeat with that one factor changed, then return to baseline. |
| Incorrect recognition language | If the first change has no effect but a supported configuration differs, incorrect recognition language moves higher on the list. | Inspect the relevant option or documented behavior without clearing unrelated settings. |
| Glare or compression artifacts | If both earlier paths are contradicted, investigate whether glare or compression artifacts accounts for the timing or device-specific result. | Collect an event timestamp or status report for qualified support. |
Now apply the least-disruptive change that corresponds to the evidence: Rescan one receipt with better contrast and set the appropriate recognition language. Avoid treating all three hypotheses at once; simultaneous changes erase the diagnostic value of the test.
Step-by-Step Troubleshooting
Begin with reversible comparisons. Use Advanced only if the evidence points to a deeper problem.
Capture the current behavior
Totals and invoice numbers become inaccurate text after OCR though the image is readable. Write down where it happens, its most recent time and the state displayed by the product. Do not treat a nearby working feature as proof the whole path is healthy.
Run the safe comparison
Compare 5 selected numbers in the scan against recognized text. Use the same device, account and operating conditions where practical. Note what remains unchanged so the comparison is meaningful.
Correct the supported setting
Rescan one receipt with better contrast and set the appropriate recognition language. Make one adjustment at a time and preserve the previous value so it can be restored if the result becomes worse.
Confirm the real outcome
Test receipt totals are searchable and match the image character for character. Repeat the complete user-facing operation and then make sure unrelated protections and functions still work. If the result is inconsistent, restore the previous setting and continue diagnosis.
Separate the three suspected layers
Assess evidence for low scan resolution, incorrect recognition language and glare or compression artifacts separately. A positive result for one path is not evidence for every path. Preserve before-and-after observations, including timestamps.
Review event-level evidence
Review suspect text in Acrobat's correction workflow and verify numeric columns manually. Gather only logs and supported diagnostic views that apply to this exact scenario; do not change service-wide policy or hardware wiring.
Isolate a second independent variable
Revisit the comparison: Compare 5 selected numbers in the scan against recognized text. If the first suspected cause is not supported, compare the behavior predicted by incorrect recognition language with the behavior predicted by glare or compression artifacts. Keep the first correction reversible.
Commit or escalate with evidence
Confirm: Test receipt totals are searchable and match the image character for character. If that cannot be reproduced, restore the original safe configuration and send sanitized observations to the document originator or PDF software vendor. Safety boundary: Never trust OCR totals for financial reporting without review.
Solution Checklist
Solution Checklist
Common Mistakes to Avoid
- Assuming low scan resolution is proven simply because the visible symptom matches this page.
- Changing multiple accounts, rules, connections or device settings before running the control: Compare 5 selected numbers in the scan against recognized text.
- Skipping this safety boundary: Never trust OCR totals for financial reporting without review.
- Resetting, formatting, deleting credentials or replacing hardware without a confirmed backup, a supported recovery path and evidence that the step is necessary.
A setting that makes one symptom disappear is not always the correct permanent configuration. If it disables another essential function, revert it and investigate the underlying dependency rather than leaving an unsafe workaround in place.
How to Verify the Fix
The primary pass condition is specific: Test receipt totals are searchable and match the image character for character. Repeat the original workflow, not a loosely related shortcut, and confirm both the previously failing behavior and an unaffected comparison still work.
- Record the before-and-after state in the same location, account and operating mode.
- Repeat the original comparison: Compare 5 selected numbers in the scan against recognized text.
- Check the stated pass result: Test receipt totals are searchable and match the image character for character.
- Confirm that normal security, notification, storage or physical safety protections remain enabled.
If the result works only once or disappears after reopening the app, reconnecting the accessory or waiting for the next scheduled event, treat the issue as intermittent instead of fixed. Keep the event time and what changed between attempts.
If the Problem Continues
If the safe correction does not work, move to a focused deeper investigation: Review suspect text in Acrobat's correction workflow and verify numeric columns manually. This is a different path from the first settings check; use it only if the device or service supports the required diagnostics.
When escalating to the document originator or PDF software vendor, provide the exact symptom, result of the control test, changes already tried, time of occurrence and any non-sensitive error or status data. Explain whether low scan resolution, incorrect recognition language and glare or compression artifacts were supported or contradicted by evidence.
Do not share account recovery codes, private documents, live access tokens, sensitive location histories or full recordings merely to prove the problem. Redact or summarize identifiers before opening a public support request.
Prevent Recurrence
Keep a short reference note with the working configuration and the successful check: Test receipt totals are searchable and match the image character for character. After an app update, firmware change, network migration or scheduled clock change, run the same short test before depending on the function again.
Watch for the condition that originally made the failure visible: Totals and invoice numbers become inaccurate text after OCR though the image is readable. Where practical, retain an alternative approved way to perform the essential task so that one device, cloud service or interface does not become a single point of failure.
Official Resources
- Adobe Acrobat: Fix text recognition errors in scanned PDFs – vendor documentation specifically relevant to this workflow.
FAQ
How can I tell whether low scan resolution is responsible?
Run the controlled comparison: Compare 5 selected numbers in the scan against recognized text. A change that specifically affects the failing layer supports this hypothesis, but repeat it before changing permanent settings.
What if the real cause is incorrect recognition language rather than low scan resolution?
Do not repeat the initial correction blindly. Check the supported diagnostic path: Review suspect text in Acrobat's correction workflow and verify numeric columns manually. Keep the earlier working configuration recorded so you can reverse the test.
When is this particular issue resolved?
The decisive result is: Test receipt totals are searchable and match the image character for character. If it fails after the normal workflow is repeated, document whether glare or compression artifacts still explains the remaining behavior and escalate with non-sensitive observations.