B2B SEO: Fewer Pages, Better Ones, and a Real Cluster Map

B2B SEO advice usually assumes the problem is having too little content. For most B2B sites that have been publishing for a few years, the problem is the opposite: too many pages chasing overlapping intents, none of them strong enough to rank. We just worked through this on our own blog, so this guide uses our numbers rather than a hypothetical.

This is for marketers who inherited a large archive and cannot work out why traffic is flat despite publishing.

Cannibalization Is the Most Common Diagnosis

This blog had 193 posts. Twelve of them were about LinkedIn Events. Seven were about LinkedIn Stories, a feature LinkedIn discontinued in 2021. Nine titles opened with "Unlock Your Potential."

To a search engine that is not twelve answers about LinkedIn Events. It is one intent split across twelve weak URLs, each with a fraction of the internal links, a fraction of the external links, and no clear signal about which one is the answer. So none of them rank, and the site reads as a LinkedIn tips blog rather than the revenue operations platform it belongs to.

The test is simple. Search your own site for your target keyword. If three of your pages could plausibly answer it, you have found the problem.

Consolidate, Do Not Delete

Deleting a page throws away whatever links and history it accumulated. A 301 redirect into the page you want to rank hands that equity over instead.

So the move is: pick the one URL per intent that should own it, make that page genuinely the best answer, and 301 every duplicate into it. We kept 55 posts, wrote 21 new pillars, and redirected 138 URLs. Same content mass, an eighth of the URLs competing.

Two rules that matter. Redirect to the closest topical match, not to the blog index. A redirect to a hub passes much less and frustrates the reader who clicked a specific link. And never redirect in a chain; point old URLs at the final destination directly.

One Keyword, One URL, Written Down

The discipline that prevents this from happening again is a keyword map: a file that assigns exactly one primary keyword to each URL. Ours is a module in the repo that the build reads, so a new post physically cannot ship without an owner keyword, and a duplicate assignment fails the build.

You do not need it in code. A spreadsheet works, as long as it is consulted before commissioning rather than after publishing. The failure mode is always the same: someone briefs a writer on a topic instead of a keyword, and you get your fourth page about LinkedIn analytics.

Hub and Spoke, Actually Linked

A cluster is a pillar page targeting the broad term, with supporting posts targeting specific long-tail variants, and real links in both directions. The links are the part people skip, and without them a cluster is just a folder.

Every supporting post here links up to its pillar with a line naming it. Every pillar links down to its supporting posts. Every post links sideways to its cluster hub. That is generated from the cluster map rather than hand-placed, so it cannot drift as posts are added.

The effect is that a crawler landing on any post can reach every related post in two hops, and link equity flows toward the pages you actually want ranking rather than pooling wherever you happened to link last.

What AI Answers Reward

AI Overviews and assistant answers are now a meaningful share of B2B research, and what they extract is different from what a classic featured snippet rewarded.

They pull self-contained passages. A section that answers one question completely, with a specific number and a named source, can be lifted and cited without the surrounding page. A section that depends on three paragraphs of preceding context cannot.

Which changes how to write. Put the answer in the first sentence under the heading. Use a real number with its unit and baseline. Name the source in the text rather than only in a link. Add an FAQ block with literal questions and complete answers. That is the cheapest citation surface available, and it is why every pillar here carries one.

The Technical Floor

None of this matters if the basics are broken, and on most B2B sites at least one of these is:

• A sitemap that exists and lists only canonical URLs. Ours did not exist at all until this sprint.
• One canonical per page, self-referencing, with no trailing-slash or .html variants also resolving.
• Article and BreadcrumbList structured data on every post, FAQPage where there are FAQs.
• An og:image that social crawlers can actually render. SVG is not one. Every share of ours rendered as a bare link until we generated PNG cards.
• A dateModified that reflects real edits rather than ticking on every deploy.

What Not to Expect

Consolidation does not pay out immediately. Redirects take weeks to process, and traffic usually dips before it recovers as Google re-crawls and re-evaluates. If you cannot tolerate a soft quarter, do it in stages by cluster rather than all at once.

And it does not rescue content nobody wanted. Merging twelve thin posts produces one thicker thin post unless somebody rewrites it. The consolidation is the cheap half; the writing is the part that decides whether it works.