Review tables after AI PDF remediation

Use artificial intelligence (AI) remediation to improve the accessibility of Portable Document Format (PDF) files. Improved table-header recognition helps identify headers across the top of a table and down its side.

After remediation, review the report and downloaded document. Check that table headers describe the correct data and that the document remains understandable with assistive technology.

Note: AI remediation doesn't guarantee a fully accessible or conformant PDF. Results vary by document. Automated results don't replace the human review needed for complex or high-risk content.

Before you start

Make sure:

Your organization has access to AI PDF remediation and the required AI entitlement.
You have permission to access the PDF and use the available testing and remediation controls.
The document meets the upload requirements shown in the Platform.
Your organization permits the document to be uploaded and processed through the service.

Keep the original file so you can compare content and structure after remediation. For upload and testing instructions, refer to Get started with the PDF checker.

If the AI remediation option isn't available, ask your administrator or Level Access contact to confirm your permissions and your organization's access.

Run AI remediation

The table improvements run through the existing PDF auto-remediation workflow. You don't need to configure table headers during processing.

Go to the appropriate workspace or digital asset.
Under Find and Fix, select PDF.
Upload the document if needed, then run a PDF test.
Open its test report and review the original results as your baseline.
Select Auto Remediate for the document and wait for processing to finish.
Review the updated results and download the remediated PDF for validation.

Compare the original and remediated results to see which checks changed and which issues remain. Where the report shows Original and Remediated tabs, switch between them to compare the versions.

For report navigation, refer to View the PDF test report. Use the document's download controls to retrieve the remediated version, not just the original file or its report.

Understand the improvements

These refinements improve document structure without adding a manual editing step during the automatic process.

Area	What the refinement addresses
Table headers	Recognizes header columns as well as header rows. Header cells can be marked with scope indicating whether they describe a row or a column.
Document title	Can generate a title from document content instead of using a generic placeholder when a title needs remediation. Generated wording and language still need review.
Page numbers and running content	Improves identification of page numbers and repeating header or footer content as artifacts, so screen readers can skip that content in the main reading order.

In a PDF, an artifact is content marked to be skipped by assistive technology. Meaningful information should remain available in the reading order.

Review table accessibility

A table can look correct while its underlying structure is incomplete. Check both the visual document and its accessibility tags using an appropriate PDF inspection tool, then test the reading experience with assistive technology.

Check	What to review
Column headers	The headings across the table are identified as headers rather than ordinary data cells. They describe the correct columns.
Row headers	Headings down the side of the table are identified as headers when appropriate. They describe the correct rows.
Header relationships	Each data cell has the correct header context. Pay particular attention to merged cells, multiple header levels, and tables with both row and column headers.
Table structure and reading order	Cells remain in the correct rows and columns and follow a logical reading order. Check tables that continue across pages for unintended splits or missing context.
Content accuracy	Data, headings, notes, and captions remain complete and accurate when compared with the original document.

Note: Complex tables can still need manual correction. Header text can be misclassified as body content. Don't assume every header is correct because the remediation job completed or the table looks unchanged.

Review any failed, warning, or unchecked results. For information about the automated checks, refer to PDF testing rules. A passing automated check doesn't confirm every contextual relationship in the table.

Review titles and running content

Alongside table checks, review the document metadata and the information read between pages.

Document title

Check the title in the downloaded PDF's metadata, not only the file name. Confirm that it accurately describes the subject and uses an appropriate language.

A generated title can be too general, omit an important detail, or use the wrong language. Correct it when necessary using your normal PDF editing workflow. Title generation also has a fallback if the AI service is unavailable, so a completed job doesn't guarantee a newly generated title.

Page numbers, headers, and footers

Check that decorative page numbers and repeated running content don't interrupt the main reading order. Also confirm that meaningful information wasn't incorrectly marked as an artifact.

For example, a repeating page number may be skipped, but information needed to understand the document should remain accessible.

Complete remaining remediation

AI remediation is a starting point for improvement, not a certification of accessibility. Before publishing or distributing the result:

Review the updated test report for remaining issues.
Validate the downloaded PDF's tables, reading order, and generated metadata.
Compare it with the original to identify missing or changed information.
Correct remaining issues in a suitable PDF editing tool, or request expert remediation.
Retest and perform the human checks appropriate to the document's purpose.

Complex, critical, or compliance-sensitive documents may require specialist review. If your team needs help, contact Level Access about manual PDF remediation services and any applicable charges.

Keep the original file, remediated file, and validation results according to your organization's documentation process. Don't rely on a fixed percentage of issues being repaired or a guaranteed completion time.

On this page: