A four-partner accountancy firm in Leeds runs its first gap analysis and gets back 4,217 rows. Somewhere in there are the three articles they should write next quarter. The rest is noise: competitor brand names, terms about Australian tax law, a thousand long-tail phrases with no measurable UK volume, and a stack of keywords owned outright by a national comparison site that will never lose them. The export took ninety seconds. Working out what to do with it is the actual job, and almost nobody writes about that part.

Competitor keyword tracking is the repeated comparison of your ranking footprint against a named set of rivals, run on a schedule, so you can see what they rank for that you do not. The emphasis is on repeated. A one-off export is a snapshot of one Tuesday. Tracking is a diagnostic you keep running, because the gaps that are winnable this month were not winnable in March, and some of the ones you dismissed in March have opened up since.

What a content gap actually is, and what it is not

A gap is not "a keyword a competitor ranks for and you do not". That definition produces thousands of rows and no decisions. A gap is a query your buyers plausibly search, that you could realistically compete for, that your site does not currently answer well. Three conditions, all of them, not one.

It also helps to stop treating gaps as a single category. There are four, and they need four different responses:

• Missing topic. You have never covered it. Two competitors have, and they both rank. This is the one everybody means when they say "content gap".

• Weak page. You have a page on the topic. It sits at position 18 and has done for a year. The gap is quality and depth, not existence.

• Wrong intent. You have a page, it is decent, and it answers a different question from the one being asked. A service page where the top 10 is all guides will not move, however good the service page is.

• Cannibalised cluster. You have two pages, sometimes three, splitting the same query between them. Neither wins. The export shows this as a gap because your best URL never holds a stable position.

Most exported gaps are not gaps at all. Strip out competitor brand terms first, because you are not going to rank for someone else's company name and you would not want the traffic. Strip out anything a national publisher or marketplace owns through sheer domain authority. Strip out terms attached to service lines you do not sell: an accountancy firm that does not do R&D tax credits should not be writing about R&D tax credits because a rival happens to rank for it.

And set your expectations on timing. A gap you find in October is a brief for November and a page that tells you something useful in January. New pages need an indexing and reassessment cycle before position data means anything, which is why we review movement at 8 to 12 weeks, not 8 to 12 days. Anyone promising faster is selling something.

Choosing the competitors you track, because the whole report depends on it

A gap report is only as good as the competitor set behind it. This is the step teams rush and then wonder why every opportunity looks impossible.

Start by separating search competitors from business competitors. The firm you lose pitches to may not be the site you lose SERPs to. Your business rival might be a two-person consultancy with a brochure site and no blog, in which case they tell you nothing. Meanwhile a sector publication you have never pitched against is taking every informational query your buyers type before they ever reach a shortlist.

A workable set is usually 5 to 10 crawled sites: two direct rivals you actually compete with commercially, two search-only rivals that keep appearing in your target SERPs, and one aspirational site a size or two above you to show where the topic eventually goes. Hand-picked beats auto-generated every time. Tools that guess your competitors from keyword overlap will hand you Wikipedia and a national newspaper.

Back to the Leeds accountants. Their first set included four local rivals and two national comparison sites. Every term in the report came back at difficulty 70 or above, because the comparison sites rank for everything and drag the apparent competition up with them. Drop those two from the set and a cluster of practical questions appears underneath: the sort of thing the local rivals already rank for, at difficulty levels an SME can genuinely take. The keywords were always there. The national sites were burying them.

Refresh the set when the market moves. A competitor rebrands, gets acquired, merges their blog into a parent domain, or simply stops publishing. Tracking a site that has not posted since 2023 gives you a frozen baseline and a false sense of progress. Every quarter is a reasonable review cadence for the list itself.

From raw crawl to a shortlist you would actually commission

Crawl your own site first. Not second. The comparison is meaningless until the system knows what you already rank for, what exists but sits buried on page four, and which of your URLs are competing with each other. We read your own site the way a new hire would before we look at anybody else's.

Then the sequence is straightforward:

1. Pull ranking keywords for each tracked competitor.

2. Subtract your own ranking keywords. What remains is the raw gap list.

3. Filter on UK volume and difficulty, refreshed weekly rather than carried over from a static annual export.

4. Cluster what survives by topic, so one strong article replaces six thin ones.

5. Score each cluster for commercial relevance to what you actually sell, before anything reaches a brief.

Step three is the one people skip, and it matters more than it sounds. Difficulty and volume are not fixed properties of a keyword. They are descriptions of a SERP at a moment in time. A term that looked closed because the top 10 was full of long-established pages can open up within weeks after a reshuffle, a site migration, or an update that demotes a few thin pages. Weekly refreshed UK data changes which gaps are live in a way a yearly snapshot never will. If you are already running rank tracking and reading the numbers properly, you have half the picture; gap analysis supplies the other half.

Done properly, those 4,217 rows become roughly a dozen topic clusters once branded terms, irrelevant service lines and zero-volume long tail come out. The relevance and SERP check takes it down again, to three or four briefs. That is not a disappointing result. That is the result.

Read the SERP before you commission anything

Gap analysis without a SERP read produces the wrong page type. This is the expensive mistake, because you find out three months after publication.

A term can look like a service-page opportunity in a spreadsheet: commercial wording, decent volume, a competitor's service page ranking fifth. Open the live results and the other nine are how-to guides. Google has already decided what that query means, and the answer is an article. Build a landing page for it and you will sit at position 30 wondering why.

So for every shortlisted gap, open the top 10 and note three things: the dominant format, the depth of the leading results, and which SERP features are present. Featured snippets reward a direct, tightly structured answer near the top of the page. People Also Ask tells you which sub-questions belong in your H2s. A video pack above the fold cuts the click-through on everything below it, which changes the value of a win. A local pack on a service term means proximity and your Google Business Profile matter more than the article does.

Match the intent to the output. Informational queries want guides and explainers. Commercial comparison queries want comparison pages with real detail. Transactional queries want a service or product page. Navigational queries you ignore unless the brand is yours.

And sometimes the honest answer is to skip it. If the top 10 is nine national brands and a government page, that keyword is not a gap for a ten-person firm. Write it off and spend the hours on the mid-difficulty cluster underneath. We would rather tell you a term is not worth chasing than take the brief and let you find out in April.

Check your own site before you write a word

The most common failure is not missing a gap. It is filling one you already filled.

Inventory first: existing service pages, old blog posts, anything published two years ago and forgotten. Duplicate titles happen constantly on sites where content was commissioned in batches by different people, and the result is two of your own URLs splitting the same query between them.

Cannibalisation has clear tells. Two URLs swapping positions week to week for one query. A thin page outranking your properly researched one. Impression share spread across several URLs with no single page consolidating. Take a services site with two older posts on near-identical topics, trading positions 11 and 14 every few days. Adding a third article makes it worse. The fix is consolidation: merge the useful parts into one expanded page, redirect the weaker URL, update the internal links pointing at it.

Deciding between a refresh and a new page comes down to whether the existing URL is close. If it ranks anywhere in the top 30 and covers the right intent, expand it. If it answers a different question, or the topic genuinely is new to your site, write the new piece and link it into the cluster properly. A gap closes as a group of pages, not as a single orphan post nobody links to. Our editorial process for every brief includes that check before a draft exists, because a duplicate is harder to undo than to prevent.

Turning tracking into a rhythm you can keep

Weekly works for most lean teams. Scan the refreshed data, confirm one or two briefs, write, publish, and review movement at 8 to 12 weeks. That is the whole loop. Twenty minutes of scanning beats a four-hour audit every six months, because the four-hour audit produces a shortlist that is stale by the time anyone writes from it.

Decide which device you track against. Mobile and desktop results differ in ordering and in how much space features take up, and the split that matters is your own analytics split, not an industry average. A B2B firm whose traffic is overwhelmingly desktop should not be optimising around a mobile SERP it rarely sees.

Ramp volume gradually. Closing a batch of gaps at once is self-defeating: publishing twenty pages in a fortnight on a site that normally posts twice a month creates an unnatural spike, spreads the research thin across all twenty, and leaves you unable to tell which page moved what. Google's own guidance on creating helpful content is blunt about producing large amounts of content primarily for rankings. Four good pages a month, each properly researched, beats twenty rushed ones and gives you clean signal.

Watch the traffic going the other way too. If a tracked competitor starts appearing in SERPs where you were comfortable, that is them closing gaps on you. Early signs: their new URLs entering positions 15 to 30 on your terms, a sudden increase in their publishing frequency, a redesign of a page you outrank. Respond to those while they are at 20, not after they are at 4.

What makes gap reports useless

Sorting by volume. Volume is the laziest sort order there is, and it puts the hardest, least relevant terms at the top of your list. Sort by relevance, then filter by difficulty, then look at volume last.

Writing for a gap your audience does not have. One competitor ranking for something is not evidence of demand. Two or three, consistently, with volume behind it, is a signal. One is an accident of their history.

Chasing everything. A dozen clusters tackled simultaneously produces a dozen mediocre pages. Pick three.

Never re-running the analysis. Competitor keyword tracking only pays off as a habit. Run it once and your shortlist is stale within a quarter, and you will not notice until the pages you wrote stop converting. If you want a read on where your site stands before building the habit, our free content marketing snapshot is a reasonable starting point, and the content marketing glossary covers the terminology if any of this is new.

Frequently Asked Questions

What is competitor keyword tracking and how is it different from rank tracking?

Rank tracking monitors your own positions for a list of keywords you have already chosen. Competitor keyword tracking compares your whole ranking footprint against a named set of rivals, repeatedly, so it surfaces terms you never thought to add to that list. One tells you how you are doing. The other tells you what you are missing.

How many competitors should a small business track?

Five to ten crawled sites is usually enough: two direct commercial rivals, two sites that keep appearing in your target SERPs, and one aspirational site above your weight class. Fewer than five and the data is thin. More than ten and you drown in rows without adding insight. Hand-pick them rather than accepting an auto-generated list.

How often should keyword and competitor data be refreshed?

Volume and difficulty are worth refreshing weekly, because both describe a SERP at a point in time and SERPs move. The competitor set itself needs reviewing roughly quarterly, or whenever a tracked site rebrands, gets acquired or stops publishing. An annual export is a historical document, not a planning tool.

How do you tell a genuine content gap from a keyword you should ignore?

Three tests. Does it map to something you actually sell or advise on? Could a site your size realistically reach the top 10, judged by the live results rather than a difficulty score alone? And does your site genuinely fail to answer it already? Anything failing one of those is noise.

Should you write a new page or update an existing one to close a gap?

If you already have a page on the topic that sits anywhere in the top 30 and answers the same intent, expand and refresh it. If your existing page answers a different question, or there is nothing on the site at all, write a new piece and link it into the surrounding cluster. When two of your own URLs are splitting one query, neither option applies: consolidate them into one page and redirect the weaker URL.