AI-assisted practical guide. Examples are hypothetical; proposed workflows are editorial suggestions.
You can record a flower visit without knowing the visitor's species. The useful move is to preserve the observation at the level the evidence supports. An honest provisional label is more valuable than a precise guess that later becomes a false record.
Use temporary visual groups
Describe visible characteristics in ordinary terms and attach photographs. Give a temporary group a neutral code, such as V1. Explain that the code groups similar-looking observations; it does not establish a species, sex, or number of unique individuals.
Record the date, patch, observation effort, and plant identity if known. Keep an unknown plant separate from an unknown visitor so either identification can later improve independently.
Follow a hypothetical review
Suppose several V1 photographs turn out to show different insects. Split the interpretation in the index while preserving the original observation IDs. Do not rewrite the field notebook as though the distinctions were known at the time.
If a photograph cannot support further identification, retain it with that limitation. It may still document a visit even when it cannot support a species record.
Report the right result
Describe provisional groups and observation counts explicitly. Do not present the number of visual codes as species richness. A short methods note should tell readers how groups were assigned and which records were reviewed. That makes the collection usable for learning without pretending it is a professionally verified inventory.