How the ranking works
This extracts recurring words and two-word phrases from your actual caption text, then sorts them by how often they appear - a word or phrase mentioned more than once, or matching your niche keyword, ranks higher. Tags are then grouped into broad, medium, and niche tiers by word length and commonness as a rough proxy for likely competition level.
Why using your actual caption produces better tags than a bare topic
A longer, more detailed caption gives the frequency-ranking approach genuinely more signal to work with - specific phrases and recurring themes in your own writing surface naturally, rather than relying on a short topic phrase alone to guess at relevant tags. This is why pasting your real caption, not just a one-word topic, produces noticeably more tailored results.
Why the niche keyword field gets a strong boost
Terms from your explicitly entered niche keyword receive a deliberate scoring boost over whatever the caption text naturally surfaces - since you're telling the tool directly what your content is actually about, that explicit signal is weighted more heavily than an inferred pattern from word frequency alone.
Why the broad/medium/niche tiers are a heuristic, not live data
This is a mechanical heuristic based on your own text, not a real popularity or trend analysis - tag length and general commonness are used as a rough proxy for likely competition level, which correlates reasonably well with actual tag popularity but isn't the same as checking real current search or usage volume on the platform itself.