Data or Header Cells Must Not Be Used as a Data Table Caption

In one paragraph

Data or header cells must not be used to give a caption to a data table means a visual title may have been encoded as ordinary table data instead of an appropriate table label. Review its actual function before changing it: a spanning header can be legitimate. The supplied WebAuditOS configuration does not run this experimental detector, so manually assess table naming and relationships.

Why it matters

A caption can identify a data table and help users decide what it contains before navigating its cells. A merged title row can instead be treated as data or a broad header, which may not express the intended table-level relationship.

Do not assume every merged first row is a fake caption or that every data table necessarily lacks a name without caption. Other naming mechanisms exist, and real grouped headers can span columns. The job is to encode the actual information structure, not blindly remove colspan.

Table naming and header associations solve different needs. A correctly named table can still have wrong column or row relationships, while a complex valid header arrangement should not be simplified into inaccurate flat data merely to satisfy a heuristic.

How WebAuditOS evaluates this finding

The supplied table-fake-caption heuristic targets certain detected data tables whose first row contains one cell spanning the width. The source labels the rule experimental and states it is excluded from the current tag-selected run.

It therefore raises no finding in the documented report. Its nominal impact is not a reported severity, and other table rules can still produce findings. Even if enabled, the candidate shape would not prove that the cell is a title rather than a legitimate grouped header.

How to fix it

1. Identify the real table title, introductory explanation, and header hierarchy. Ask whether the merged cell names the whole table or actually categorises data below it.
2. For an actual table title, prefer an appropriate native caption. A basic illustration is:

<table>
  <caption>Illustrative delivery rates</caption>
  <thead><tr><th scope="col">Destination</th><th scope="col">Rate</th></tr></thead>
  <tbody><tr><th scope="row">Czechia</th><td>5 EUR</td></tr></tbody>
</table>

The values are examples. Replace them with actual content and preserve whatever header structure the real data needs.

3. Move the title text out of an unnecessary data row without deleting genuine data or grouped headers. Check rowspan and colspan effects on the remaining grid before changing cells.
4. If an existing visible heading is intentionally used as the table's accessible name, verify the appropriate explicit relationship. A nearby heading is not automatically the table name, and adding another overriding label can mask a useful caption.
5. Keep additional instructions separate from the concise identity where practical. Explain units, abbreviations, and navigation needs in accessible surrounding content or appropriate descriptions.
6. Repair the generating template or CMS table component so later imports do not recreate fake titles. Spreadsheet paste workflows may need editorial review rather than assuming every first merged row has the same meaning.
7. Test reading order, header associations, keyboard and assistive navigation, zoom, and responsive presentation. A card-style mobile transformation must still convey relevant relationships.
8. Record manual evaluation and rescan other table checks. The absent experimental finding cannot certify captions, proper header IDs, or the accessibility of the data itself.

What the fix can break

Removing a row can change cell indices, formulas, exports, CSS selectors, or scripts that assume fixed positions. Verify the actual application uses after the semantic repair.

Replacing a true grouped header with a caption can destroy useful relationships. Review meaning rather than applying the same change to every spanning cell.

Do not add a meaningless caption to a layout-only table to improve a score. Prefer proper CSS layout where appropriate and reserve data-table semantics for data relationships.

Questions

Is every full-width header a fake caption?

No. It can express a legitimate grouped header. Evaluate the content and intended relationships.

Is a heading above the table automatically its name?

No. Use caption or an appropriate explicit naming relationship where table identity needs that association.

Does no finding mean the table was checked?

Not for the supplied experimental rule, which is excluded. Other table checks have separate scopes.

Which related checks help?

Review headers references, cell headers, and the automation limits guide.

What severity does this get in the report?

None, today - the rule is experimental in axe-core and our scan does not run experimental rules, so it never produces a finding. axe declares the rule serious; if it ran and failed, the mapping of critical and serious to high, moderate to medium and minor to low would make it high.