Long-tail Amazon keywords are more concentrated than head terms, not less. We measured 2,004 of them

The standard keyword playbook has a shape almost everyone recognises. Head terms are expensive and locked up by incumbents, so you go long. You add modifiers, you accept smaller volume, and you buy your way into demand nobody is defending yet.

We tested the second half of that claim against Amazon’s own click data. It does not hold.

What we measured

Amazon’s Brand Analytics Top Search Terms report publishes, for each search term, the three most-clicked products and the percentage of that term’s clicks and conversions each one received. Combined, those three numbers describe how concentrated a search result is: whether a handful of listings absorb the demand, or whether it spreads across the page.

We sampled three bands of the US search-frequency-rank curve for the July 2026 monthly period — head terms at rank 1–1,000, a middle band at 10,000–12,000, and a long-tail band at 100,000–102,000 — for 2,175 rows in total.

Before analysing anything we removed 171 rows, 7.9% of the sample, where the top three products showed a 0% conversion share. These are almost entirely media and streaming catalog queries: film and book titles like Obsession or Hail Mary, where clicks land on a Prime Video or catalog page rather than a purchasable unit. Left in, they would have manufactured a false concentration spike, since a single title can take 97% of clicks on its own name. That leaves 2,004 physical-commerce search terms.

Finding 1: about 72% of clicks go to products outside the top three

The median combined click share of the top three clicked products is 28.2%. The mean is 32.0%, pulled up by a thin right tail.

Read the other way: on a typical Amazon search term, roughly 71.8% of clicks go to products that are not in the top three. The single most-clicked product holds a median click share of just 13.0% — about 46% of whatever the top-three block controls.

That is a much flatter distribution than the ranking-is-everything framing suggests. Position one on a typical term is worth about an eighth of the clicks, not a majority of them.

The full distribution across the 2,004 terms:

Combined top-3 click share	Terms	Share of sample
Under 15%	303	15.1%
15–25%	534	26.6%
25–40%	635	31.7%
40–60%	355	17.7%
60% or more	177	8.8%

Fewer than one term in eleven is genuinely dominated by three products.

Finding 2: concentration rises as demand falls

This is the result that contradicts the playbook. Splitting the clean sample by demand band:

Band (search frequency rank)	Terms	Median top-3 click share	Terms where top-3 hold 50%+	Terms where top-3 hold under 25%
Head (1–1,000)	400	23.9%	9.5%	52.0%
Middle (10,000–12,000)	821	28.5%	15.1%	42.5%
Long tail (100,000–102,000)	783	30.4%	19.2%	35.8%

Concentration increases monotonically as search volume falls. The proportion of terms where three products own at least half the clicks doubles from head to tail. The proportion of comfortably open terms — top three under 25% — falls from 52.0% to 35.8%.

The gap is visible at the extremes too. Among head terms, 36.5% have a top-three share under 20% and only 4.5% exceed 60%. Among long-tail terms, only 20.9% fall under 20% while 11.7% exceed 60%.

The biggest head terms are startlingly open. Summer dresses for women, at 1,113,887 monthly searches, gives its top three a combined 4.4% of clicks. Dresses gives them 2.7%. Meanwhile a tail query like acure brightening facial scrub, at 5,787 searches, hands its top three 90.7%.

Finding 3: there are two kinds of tail terms, and length cannot tell them apart

The mechanism becomes clear when you split terms by how settled they are and then look at conversion.

Term type	Median top-3 conversion share (head band)	Median top-3 conversion share (tail band)
Owned — top-3 hold 50%+ of clicks	48.0%	55.0%
Open — top-3 hold under 25%	9.8%	8.2%

Terms the top three own are terms where shoppers already know which product they want. They click it and they buy it, so click share and conversion share are both high. Open terms behave in the opposite way: even the leading products convert less than a tenth of the term’s purchases, because the query describes a category rather than a product.

The long tail contains far more of the first kind. That is why it is more concentrated — not because generic tail phrases are harder to win, but because a larger fraction of low-volume queries are named-product searches that were settled before you arrived.

And here is the part that breaks the standard heuristic. In the long-tail band, the terms the top three own have a median phrase length of three words. The terms they do not own have a median phrase length of three words. Word count carries no signal about winnability at all.

What this changes

Volume plus phrase length is not a difficulty model. It is two numbers that feel like a difficulty model, and on this sample they point the wrong way roughly as often as the right way.

The screen that does work is cheap: pull combined click share and conversion share alongside volume for every candidate term. High click share with high conversion share means the query is settled and you are buying traffic that has already chosen someone else. Low click share means demand is genuinely distributed and share is accumulable. High click share with low conversion share is the most interesting case — incumbents are absorbing clicks they cannot close, which is exactly where a better-matched listing takes ground.

Two honest limits on all of this. The sample is three narrow rank bands in one marketplace for one month, so treat the direction of the trend as the finding and the exact percentages as period-specific. And Brand Analytics only publishes three products per term; we can measure how much the top three hold, but not how the remaining ~72% is distributed behind them.

The Operator Read

Long-tail keywords are not a discount rack. On this sample they are more winner-take-all than head terms, because low-volume queries are disproportionately named-product searches that are already resolved. The head, where a million-search term can leave 95% of its clicks unclaimed, is frequently the more open ground.

Stop letting word count stand in for difficulty. Screen on click share and conversion share together, spend against the ~42% of terms where the top three hold under a quarter of clicks, and re-pull the numbers every cycle — share metrics move whenever anyone else does.