AI changed 04/05/2026 into a date. Was it April 5 or May 4?

Confirm the source’s day/month order before converting. Both readings can be valid, so a conversion without an error can still be wrong.

Keep the original text and ask the source owner which format it uses. Microsoft documents that Excel date conversion depends on system Date and Time settings. Changing how a date is displayed does not establish what the original string meant.

One string, two valid results

We parsed the invented string 04/05/2026 locally with Python. The explicit format %d/%m/%Y gave 2026-05-04; %m/%d/%Y gave 2026-04-05. Both succeeded. This test demonstrates ambiguity, not which date the writer intended.

Include cases that must fail

Our test parsed 13/05/2026 as 2026-05-13 with day first and rejected it with month first. It also rejected 31/02/2026 with day first. Do not repair an invalid date by guessing another date unless the source confirms it.

Give AI a declared rule

Suggested input: “The source owner confirms DD/MM/YYYY. Preserve the original column, add a YYYY-MM-DD result and flag impossible dates. If a row uses another format or the source rule is missing, ask instead of guessing.” Keep unknown cases in a review list.

Check before replacing originals

Review a few known dates against the source, including a day above 12 and an ambiguous day below 13. A successful parser is not an independent source. Preserve the original text until the intended values are confirmed.

Published October 14, 2026. We tested explicit parsing of fictional strings in Python 3.11.3. We did not run Excel, change regional settings, access real date records or test an AI service.

Sources: Microsoft · DATEVALUE · Python · datetime.