Semrush Keyword Gap: how to run it and turn the overlap sets into a plan

Semrush Keyword Gap compares your domain against up to four competitors in one chosen country database and sorts every keyword into sets: the ones you all rank for, the ones they rank for and you do not, the ones where you are weaker, the ones where you are stronger, and the ones nobody in the set has yet. You enter the domains, pick organic, paid or shopping, choose the country, and read the table. The value is not in the export. It is in the three filters you apply afterwards: remove their brand terms, remove queries your site has no business answering, and group what is left into pages. A clean run ends with eight to fifteen page ideas, not four thousand rows.

This page covers the Keyword Gap tool specifically. For the other Semrush keyword reports and the order to use them in, see the five Semrush keyword tools, and for the method that sits around any tool, see keyword research in five steps.

What the tool actually compares

You can compare root domains, subdomains, subfolders or single URLs, and you can mix them. Comparing a competitor's whole domain against your one product subfolder produces a gap list that is mostly noise, so match the level of the thing you are comparing. If you want to know why their blog outranks yours, compare blog subfolder to blog subfolder.

The sets the tool sorts keywords into are the part people misread, so here is what each one is for.

Set	What it means	What to do with it
Shared	Every domain in the comparison ranks for the keyword	Your competitive core. Check where you sit inside it and improve the pages that are on page two
Missing	All the competitors rank, you do not rank at all	The highest value set. Coverage you simply do not have yet
Weak	You rank, but below every competitor in the set	Fastest wins. The page exists, so the work is improvement rather than creation
Strong	You rank above all of them	Your defensible ground. Keep it updated and link to it internally
Untapped	At least one competitor ranks and at least one does not	Where a category is still being decided. Good for second mover pages
All	Every keyword in the comparison with each domain position	The raw table to export when you want to do your own grouping

Choosing the competitors, which decides everything after

The output is only as good as the four domains you type in. Three rules make the difference.

• Use search competitors, not business competitors. The company your sales team loses to may not rank for anything. Pull the domains that actually appear for your ten most important queries and use those.
• Keep them the same size and shape. Comparing a thirty page site to a national marketplace returns a Missing list of tens of thousands of product queries you will never serve. Pick competitors within reach.
• Run it more than once. One run against three direct rivals, one run against the publishers and comparison sites that keep appearing above you. The second run is usually where the content ideas are.

Set the country database to the market you sell in before you read a single row. The same domains produce entirely different gaps between one country and another.

Turning the export into a plan

Start with Missing and Weak, apply the filters in this order, and the list shrinks to something a team can actually deliver.

• Exclude branded terms. Filter out every competitor name and product name. These inflate the gap and you will not win them.
• Exclude what you cannot serve. If the query asks for a product, a service or a location you do not offer, it is not a gap. It is somebody else business.
• Filter by intent. The tool labels intent as informational, navigational, commercial or transactional. A page type follows from the label: guides for informational, comparison and category pages for commercial, product and service pages for transactional.
• Put a difficulty ceiling on it. Set a keyword difficulty range that matches your site current authority rather than filtering on volume alone. A low volume query you can win this quarter is worth more than a high volume query that is out of reach.
• Group before you decide. Sort alphabetically and by intent, then cluster. Fifteen phrasings of the same question are one page, and treating them as fifteen pages is the most common way a gap analysis produces thin content.

Then score the clusters so the order of work is a decision rather than an argument.

Column	What goes in it
Cluster	One row per topic, with every phrasing listed inside it
Combined volume	The sum of the pooled phrasings, not the single headline term
Hardest term	The highest difficulty score in the cluster, never the average
Page type	Guide, comparison, category or service page, taken from the intent label
Do we have a real answer	Evidence, a product capability, or a specialist who can be quoted
Order	The rank you will actually work in

The column that stops bad pages is the fifth one. If nobody at your company can say anything true and specific on the topic, the cluster goes to the bottom whatever the volume says.

Reading the numbers well

Volume figures in any tool are modelled estimates built from clickstream and search data, so use them to rank keywords against each other rather than to forecast visits. Positions are sampled, so a competitor shown at position four sits somewhere near there on the day you check. Both are sound enough for prioritisation, which is all you are asking of them. Re run the comparison monthly so you are working from current data, and keep the date on every export.

Two patterns are worth more than any single row. First, a competitor appearing in Missing across a whole topic rather than in scattered keywords means they built a page type you do not have. That is the finding. Second, a long Weak list with a short Missing list means your issue is quality on pages you already own, which is cheaper to fix than a content programme.

The paid side of the same report

Switching the report to paid keywords shows which terms competitors bid on and you do not. This is the quickest read on what converts for them, because nobody keeps paying for a query that does not pay back. A term that appears in both their paid list and their organic top ten is one they have decided is worth owning twice, which makes it a strong candidate for your own plan.

What gets asked about Semrush Keyword Gap

How many competitors should I compare at once?

Use three plus yourself for a clean read. Adding a fourth competitor widens the Missing set a lot, because a keyword only needs all of them to rank for it to land there, and tighter sets produce clearer decisions.

What is the difference between Keyword Gap and Backlink Gap?

Keyword Gap compares what you rank for. Backlink Gap compares which domains link to them and not to you. They answer different questions and are usually run in the same session, because a cluster you are missing often sits behind a link profile you are missing too.

Do I need a paid plan to use Keyword Gap?

The report is part of the Semrush subscription, and the number of comparisons and exported rows you get is set by your plan tier. Semrush publishes the current limits on its pricing page, which is the place to check before you build a workflow around daily exports.

Can I do a keyword gap analysis without Semrush?

Yes. Ahrefs has Content Gap, Moz has a keyword gap view, and you can assemble a version by hand from Search Console queries plus a competitor sitemap and page titles. The method on this page applies to all of them, because the filtering is what produces the plan.

How often is a gap analysis worth repeating?

Quarterly for the full run and monthly for the Weak list. The Weak list moves fastest because it reacts to your own publishing, which makes it the honest progress report on last quarter work.

The gap no keyword tool is looking at yet

A keyword gap tells you which queries a competitor wins in Google. There is now a second gap worth measuring. Buyers put the same comparison questions to ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews, and read back a short answer that names two or three companies and cites the pages it used. Those five answer differently from each other and from the ten blue links, and the sources they lean on are often directories, forums and trade publications rather than the domains in your keyword report. A company can hold position one on a term and still be absent from the answer a buyer actually reads. AI Knows Us measures which businesses those answers name on your category questions and which sources they cite, which makes the second gap as countable as the first. See how it works.