Identify the type of caption problem
First distinguish no caption option, text that appears at the wrong time and a word rendered incorrectly while the surrounding timing works. TVNZ’s help describes a CC indicator on show or episode pages and acknowledges AI caption limitations. A spelling or recognition error is a different observation from a player failing to display an available track.
Keep the exact episode link, its title and the position of the passage. Note the caption language and the wording displayed. If you cannot establish the intended word, record the uncertainty; do not replace an unfamiliar Māori name with a guess merely because another spelling looks more familiar.
Verify a proposed correction against the programme context
Replay the passage as needed and consult any name or term supplied in the programme’s own credits, on-screen label or associated official story. A spoken proper noun can be difficult to infer from sound alone. Preserve macrons and other meaningful spelling details when an authoritative programme source supplies them.
Keep displayed wording and proposed correction in separate fields. Include the basis for the correction, such as an on-screen name, rather than presenting an automated transcript as an authoritative spelling reference. The broadcaster’s use of mixed caption sources does not identify which production method created any particular error unless the item itself says so.
Submit feedback about the specific passage
TVNZ’s caption help directs accuracy feedback to its contact form. A useful report gives the episode, time position, displayed text, proposed correction and supporting context. Keep it brief enough for the relevant passage to be found without searching an entire series. No feedback was sent during this documentation review.
For personal viewing notes, retain the verified spelling separately from the caption wording until a correction is confirmed. Do not describe a captioned programme as uncaptioned because one name is wrong, or infer that every episode uses the same AI-generated text. An item-level accuracy report preserves the service that is available while making the actual language error concrete and reviewable.