The keywords they rank for. The ones you don’t.

One run compares your site with up to 5 of the rivals you have named — you may name more than that, and the cap is on what one run measures rather than on what you may record. For your site and for each rival it measures, a run collects the top 1,000 keywords that domain ranks for, ordered by estimated traffic — the head of each profile rather than the whole tail, because the tail is where a comparison stops being informative and starts being expensive.

If you do not know who your rivals are, leave the list empty and the run finds candidates for you: domains that rank for the same keywords you do, with how many keywords they share. You pick which to keep.

• Gaps: keywords a rival ranks for and you do not appear for at all.
• Outranked: keywords you both rank for, where they are ahead of you.
• A keyword matrix, with every measured domain’s position side by side, so you can see whether one rival owns a theme or everyone is fighting over it.
• Share of voice across the set, as an estimate from a published click-through curve, labelled as the estimate it is.

The matrix draws a column for each domain the run actually measured, which is not always every rival you track: a run whose lookup for one rival failed leaves that rival out, and Keyword gap names it underneath instead of drawing a column of dashes. A column of dashes would read as "this rival ranks for none of your keywords", which is the opposite of "we did not get an answer about them".

A row carries a value and a difficulty band wherever both can be measured, and a dash where they cannot — a keyword the provider reported no search volume for cannot be priced, and one it scored no difficulty for is in no band. The value is a click estimate multiplied by what an advertiser pays for that click — a way of putting one keyword ahead of another, never a forecast of your traffic. The band comes from the measured difficulty of ranking, so an easy row worth a little and a very hard row worth a lot are told apart before you commit a week to either.

You do not need a project to compare. Name your own domain and up to 4 others in Keyword gap, and the comparison is built from the answers Domain overview fetches for each domain. Every domain you have not looked up yet costs one search from the day’s allowance to fetch, and one already looked up is read back at no cost. Domain overview itself shows any one of them on its own: its keywords, its pages, its competitors and its backlinks.

The worklist is yours to work through: mark a keyword done, dismiss one you will never target, or add a selection to the rank tracker in one go. Re-running the analysis refreshes the figures underneath without resetting any of that — which is the point of keeping the worklist on the project rather than on the run.