Long-tail keyword
A query from the thin end of the demand curve: hardly anyone searches it, and it usually spells out what they want.
See it
What it is
Long tail is a claim about demand, not about word count. 'Running shoes' is the head of the curve: enormous volume, murky intent, every retailer on earth fighting for it. 'Best trail running shoes for flat feet under $150' sits far down the tail, where any single phrase gets typed a handful of times a month. The name comes from Chris Anderson's long tail, the idea that the thin end of a distribution adds up to more than the fat head.
So the definition is relatively low individual search frequency, full stop. Length, weak competition, and sharp buying intent are things that often travel with tail queries, not things that make a query one. A three-word phrase almost nobody searches is tail; a long, well-phrased question can still be brutally contested.
Reach for tail phrases when you are new, small, or outgunned: they are the only queries you can realistically win in year one, and they often convert better because the phrasing spells out what the person wants. They are also the natural shape of voice queries and of prompts typed into AI assistants.
Gotcha: do not build one thin page per phrase. Modern engines match by meaning, not exact string, so a single strong page can rank for hundreds of variants, while fifty near-identical pages just compete with each other (keyword cannibalization) and read as filler. Second gotcha: keyword tools report '0 volume' for plenty of long-tail phrases that still get typed. Zero reported volume means below the tool's rounding floor, not zero people.
Ask AI for it
Find long-tail keyword candidates for [product/topic], aimed at [audience]. Work from the Search Console export, keyword data, or live query evidence I give you, and select phrases with relatively low individual search frequency regardless of how many words they contain. Return a table with: phrase, the demand evidence behind it, search intent (informational, commercial, transactional), why this audience would type it, and the single page on my site that should own it. Label any phrase you generated without supporting data as a hypothesis rather than a finding, and never present an invented phrase as measured demand. Group phrases that share the same underlying need onto one page instead of inventing a page per variant, and flag any two rows that would end up cannibalizing each other.