Content Gap Analysis With Claude: Find Your Missing Pages

Say you run a small invoicing app and your closest rival ranks for these three searches. You rank for none of them.

"late payment fee wording"        880 searches/mo   their position 3
"late fee on invoice"             590 searches/mo   their position 4
"interest on overdue invoices"    320 searches/mo   their position 6

A keyword tool calls that three gaps. It is one. All three searches want the same page, a guide to charging late fees on an invoice, and your rival has it while you do not. That page is the content gap. The three rows are just the evidence.

That distinction is the whole reason to run content gap analysis with Claude rather than in a spreadsheet. The subtraction is easy and any tool can do it. Turning 400 leftover keywords into the 15 pages you should write is the hard part. That part is what a language model is genuinely good at, provided it is looking at real rankings and not guessing them. This guide is the workflow we use with the tool we build, ContextBolt SEO, and every step names what it costs.

What a content gap is, and what it is not

Content gap analysis is the job of finding the topics your competitors have a page for and you do not. It is usually confused with its raw material, the keyword gap, which is the list of individual search terms a rival ranks for and you do not. Search Engine Land’s gap analysis guide draws a similar line, with keyword gaps as the terms and content gaps as the depth and coverage of what you have written. This post uses the plain version. A content gap is a page you do not have.

The difference matters because the two lists are wildly different sizes. Pull the keyword gap between any two sites of similar size and you get hundreds of rows. Cluster those rows by the page that would rank for them and you get a couple of dozen. The clustering move itself, including how to settle the pairs that could go either way, is its own guide, keyword clustering with Claude. Most of the rows were phrasings of the same question. The invoicing example above is three rows and one page, and that ratio of three or four to one holds across most niches.

So the unit of a content gap is a page. If you cannot name the page you would build to close it, it is not a gap yet. It is a row in a spreadsheet, and rows in spreadsheets do not get written.

One more scope note. This post is about finding the topics you have no page for. Sizing up a single competitor, who they are and whether you can beat them, is the competitor analysis with Claude guide, and that is where the keyword gap itself is explained. The link side of the gap, the sites that link to your rivals and not to you, is backlink research with Claude. This post takes the keyword gap as an input and turns it into a content plan.

Why Claude cannot find the gap on its own

Ask Claude “what do my competitors write about that I don’t” with no data connected and you get a fluent list. It will be wrong in the way that is hardest to catch, because it is plausible. The model has a training cutoff and no ranking data at all. It has read a lot of SEO advice about invoicing apps, so it can guess what such a site probably ranks for. It cannot tell you what yours actually ranks for this month, and it cannot tell you what your rival ranks for either.

The fix is MCP, the open standard Anthropic introduced in late 2024 for connecting an agent to live data. Connect Claude to a server that holds ranking data and it stops guessing. It calls a tool, gets the real ranked keywords for a domain, and reasons over those.

That split is worth holding onto for the rest of this guide. The data tools do the looking up. Claude does the grouping, the judgment about what is one page versus two, and the arguing about which gaps are worth your time. Neither half is much use without the other, which is why the walkthrough alternates between them.

Step 1: find the rivals who actually share your keywords

The first mistake in most gap analyses is picking the competitors by reputation. The two companies you watch on X are your business rivals. Your search rivals are whoever ranks for your queries, and that list usually includes a forum, an aggregator and a niche blog you have never heard of.

In ContextBolt SEO the tool for this is competitor_keywords. Give it your domain and it returns the sites competing for the same keywords, each with the number of shared keywords, their average position, and estimated traffic. One credit.

Copy this prompt

Find the organic search competitors for ledgerly.example.
For each one, show shared keywords, average position and traffic.
Then recommend the two or three I should run a content gap against:
similar size to me, most shared keywords, and not a giant I cannot catch.

Swap in your own domain. Read the recommendation rather than taking the top of the list, because the domain with the most shared keywords is often a site ten times your size. A rival with 240 shared keywords and traffic close to yours is a better gap partner than one with 900 shared keywords and a hundred times the authority, since the second one’s gap will be everything.

Two or three rivals is the right number. One tells you what that site chose to write about. Five or more produce a union so wide that every topic in your industry lands on it.

Step 2: pull what each rival ranks for, highest traffic first

Now the data. Run ranked_keywords on your own domain, then on each rival. It returns up to 50 keywords per domain, sorted by estimated traffic, with position, monthly volume, difficulty and whether the ranking is new, rising or falling. The default is 25, so ask for 50 by name. One credit per domain either way, so three or four credits for this step.

That cap of 50 sounds like a limit and it is doing you a favor. The 50 highest-traffic keywords a site ranks for are, almost by definition, its best pages. A page that ranks well for a real query pulls several keywords onto that list at once. The long tail below the top 50 is mostly phrasings of the same pages plus terms nobody visits. For finding pages you are missing, the top of the list is where the pages are.

Copy this prompt

Pull the top 50 ranked keywords for ledgerly.example, then for
paperclip.example and for billtidy.example.
Keep all three lists in the conversation. Do not summarize yet.

The last line matters. You want the raw rows held in context for the next step, not a paragraph about them.

Step 3: subtract, then cluster into pages

This is the step that makes the whole exercise worth doing inside an agent. Ask Claude to remove every keyword you already rank for, then group what is left by the page that would rank for it. Not by topic in the loose sense. By page, meaning “one URL could rank for all of these”.

Copy this prompt

Take every keyword paperclip.example or billtidy.example ranks for
that ledgerly.example does not.
Group them into pages: one group per URL that could rank for all
of the keywords in it. Name each page as a working title.
For each page, list its keywords with volume and the rival's position,
and add up the volume. Sort pages by total volume.
Mark any page that BOTH rivals rank for.

What comes back looks like this for the invoicing example.

Four pages out of what was a dozen keyword rows, and the volume column now means something, because it is the demand a single page could capture rather than the demand behind one phrasing. The “both rivals” column is the strongest signal on the table. A topic that two different competitors independently chose to build a page for is a topic the market rewards, not one site’s whim.

The fourth row is there on purpose. It is a gap, and it is not yours. Step 5 deals with it.

Step 4: missing page or weak page?

Here is the check that most gap tools skip, and it changes what you do next. A keyword absent from your ranked list could mean one of two things. Either you have never written about the topic, or you have a page and it ranks so badly that it never reaches the traffic-sorted list. Those need opposite fixes. One is a new page. The other is improving, retitling, or consolidating a page you already own, and writing a second page on the same topic would make that one worse.

Your own Google Search Console answers this, and it is the one source that knows about your position-60 pages. ContextBolt SEO reads it once you connect the property, and those calls cost no credits.

Copy this prompt

For each page in the gap list, check my Search Console over 90 days:
for queries containing the cluster's head term, break results down by
page and show impressions, clicks and average position.
Tag each gap MISSING if no page of mine has any impressions,
or WEAK if one does, and name that page.

Backlinko’s content gap guide puts auditing your own content beside the competitor half for the same reason. Two things show up here that no rival-side tool can see. First, a page you forgot about. On a site with 200 posts, a three-year-old post already covers “late fees on invoices” at position 38, and the fix is a rewrite that costs a morning rather than a new page that competes with it. Second, a cluster that is not one page. If two different URLs of yours each earn a few impressions for terms in the same cluster, you have split the topic, and consolidating is the fix. Google’s own guidance on helpful content keeps returning to the same idea, that one page should comprehensively serve the need, and a topic split across two thin pages serves it twice, badly.

Step 5: throw out the gaps that were never yours

Now the opinionated part. Most content gap lists are a third padding, and the padding is why the exercise gets a bad name. Three kinds of row need to go.

Their brand and product terms. “Paperclip pricing” is a gap in the same sense that “Coca-Cola” is a gap for Pepsi. Some agencies build “rival pricing” pages to catch that traffic, and it works for a while and reads as desperate. Cut them unless comparison content is genuinely your strategy.

Wrong intent for your site. A rival that sells to accountants ranks for “bookkeeping certification cost”. You sell to freelancers. The keyword is real, the page would rank, and every visitor would leave. This is the cut people skip, because volume makes the row look valuable.

Formats you will not build. If the cluster only ranks for a free tool, a calculator or a template download and you are not going to ship one, an article will not close it. Mark it and move on.

Copy this prompt

Go through the gap list and remove three kinds of row:
1. The rival's own brand or product names.
2. Topics where the searcher is not someone who would use an
   invoicing app for freelancers.
3. Clusters that only rank for a tool, calculator or download.
Show me what you removed and why, then the list that is left.

Always ask to see what was removed. The model will occasionally cut a good cluster on a bad guess about your audience, and a two-line review catches it.

Step 6: gate the survivors before you write

A gap is a topic a rival ranks for. It is not proof that you can. The last step is the same gate any new topic gets, run on the head term of each surviving cluster.

keyword_difficulty returns the difficulty score, volume and intent for one keyword, and, more usefully, the average number of referring domains pointing at the pages in the current top 10. That average is the real bar. If your site has 15 referring domains and the top 10 average 140, the gap stays a gap until you have earned links, however clean the cluster looks. One credit per keyword, so gate the five biggest clusters first and come back for the rest if those five do not fill your quarter.

Then serp_overview on the three you like best, to read who is actually on page one and what kind of page they are. If it is all forums and one thin vendor page, write it this week. If it is nine SEO publishers and an AI Overview, put it at the bottom. The full method for reading a results page is in SERP analysis with Claude, and it is the last check before a gap becomes a brief.

Copy this prompt

For the five remaining pages with the most total volume, run keyword
difficulty on the head term and show volume, difficulty and the average
referring domains of the top 10.
For the three with the best volume-to-difficulty ratio, show the SERP.
Then rank those five by how likely I am to reach page one, given that
my site has about 15 referring domains.

The whole thing as one prompt

Once you have run it by hand a couple of times, the six steps collapse into one instruction. This is the version worth saving.

Copy this prompt

Run a content gap analysis for ledgerly.example.
1. Find my organic competitors and pick the two closest to my size
   with the most shared keywords. Tell me who you picked and why.
2. Pull the top 50 ranked keywords for me and for each of the two.
3. List every keyword they rank for that I do not, then group those
   into pages, one group per URL that could rank for all of them.
   Total the volume per page and flag pages both rivals cover.
4. For each page, check my Search Console by page for queries
   containing the head term. Tag MISSING or WEAK and name my page.
5. Remove their brand terms, wrong-audience topics, and tool-only
   clusters. Show me what you cut.
6. For the five remaining pages with the most volume, run keyword
   difficulty on the head term, and show the SERP for the best three.
   Rank those five by likelihood of reaching page one for a site with
   15 referring domains, and give me each as a one-line brief.
Tell me the credits this will cost before you start.

The last line is a habit worth keeping on any multi-step ask. The agent quotes the cost, you say go, and there is never a surprise.

What it costs, in credits

Every tool named above is a ContextBolt SEO tool and each has a fixed price, so the bill is easy to add up before you start.

Twelve credits out of 1,000 a month on the $35 plan. The 7-day free trial comes with 100, so the first analysis and the next seven are covered before any money moves. Content gap analysis is bursty work, hard when you plan a quarter and untouched for weeks after, and a flat monthly price with the trial in front fits that rhythm better than a $129-a-month suite you keep paying between quarters. Add a third rival and the total goes to 13. Add the link half with backlink_gap and it goes to 15, because the backlink tools cost three each.

Running it with ContextBolt SEO

ContextBolt SEO is a hosted SEO MCP server. You start a 7-day free trial, paste one URL into Claude, Cursor, Codex or ChatGPT, and ask in plain language. There is nothing to install and no account with a data vendor to fund. The tools above are the research half of the toolkit, alongside site audits, backlinks, AI visibility checks and your connected Search Console, and you do not pick a tool, you ask.

Every lookup saves automatically to your SEO Board, a live dashboard where each rival’s ranked keywords sit with the date they were pulled, so the raw material is there next quarter without re-running the conversation. The clustered gap list itself is the conversation’s output, so paste it wherever you plan work. A content gap is a snapshot, and the useful habit is to pull the same three domains again in three months and ask which of the pages you built now rank and which new ones the rivals opened.

On the numbers, the honest framing is the one we hold everywhere. The ranked keywords and volumes are estimates from a large keyword index, decision-useful and directionally right, not a copy of Google’s own data. No third-party tool has that. For deciding which fifteen pages to write next, an estimate that is right about the order is all you need, and where a single figure matters, read it as an estimate the way you would read any tool’s number.

Where content gap analysis stops

Three limits, stated plainly, because a gap list that is treated as a plan is how sites end up with 40 mediocre posts.

The gap is your rival’s judgment, not the market’s. Two rivals agreeing on a topic is good evidence. One rival ranking for it might mean they were wrong to write it. That is why Step 3 flags the overlap and why the volume total is on the table, since a page both rivals built for 1,790 searches a month is a different bet from a page one rival built for 300.

The reverse gap is worth one look. Ask the same question the other way, what you rank for that neither rival does, and you get the pages that make your site distinct. Those are the ones to protect and deepen, and they are also a warning about the gap list, because your rivals are running this analysis on you and will find them.

And the tool sees pages through the top 50 keywords by traffic. A rival’s small pages, the ones ranking for a handful of long-tail terms nobody clicks, will not appear. That is the right trade for finding the pages that matter, and it is a trade. If you want a rival’s entire footprint, thousands of rows deep, that is what the big suites sell for the price of a big suite, and for most sites doing their own SEO it is a spreadsheet they will never finish reading.

The list you are left with after Step 6 is short, usually five to fifteen pages, each with the keywords it will carry, the rival page you are up against, and a read on whether you can reach page one. That is a content plan, and you got it from one conversation. ContextBolt SEO starts with a 7-day free trial, so the first one costs nothing.