Verified against ChatGPT · 2026-08-09
Map internal links across a topic cluster so every page points to the right neighbor, not just back to the pillar
Builds a full internal linking map across a set of cluster pages and their pillar, specifying which pages should link sideways to each other and with what anchor text — not just the standard pillar-to-child pattern.
The prompt
Ready to copy — highlighted parts are example details you can swap.
You are building an internal linking map for a topic cluster — the pillar page plus its supporting cluster pages — with a specific focus on sideways links between cluster pages, since most guidance stops at "link every page back to the pillar" and misses that pages within a cluster should also link to each other where genuinely relevant. PILLAR PAGE 'Complete Guide to Employee Onboarding' — covers the full onboarding process at a survey level. CLUSTER PAGES 'New Hire Paperwork Checklist', 'Onboarding Timeline: First 90 Days', 'Remote Onboarding Best Practices', 'Onboarding Software Comparison'. CURRENT LINKING STATE All four cluster pages link up to the pillar. No sideways links exist between any cluster pages currently. For every pair of cluster pages, decide whether a sideways link is genuinely warranted — a reader on page A would actually benefit from being pointed to page B at a specific point in the content, not just "these are topically related so let's link them." Reject a link between two pages if the only justification you can produce is topical adjacency; require a specific reason like "page A mentions X in passing, page B is the deep answer to X." For every link you do recommend, specify: the source page, the target page, roughly where in the source page's existing structure the link should sit (which section or near which existing point), and suggested anchor text that describes what the reader will get, not a bare keyword match repeated verbatim across multiple links. Separately, check whether every cluster page currently links up to the pillar, and flag any that don't. Finally, flag anchor text that's identical or near-identical across more than two links pointing to the same target page — repeated exact-match anchor text across many internal links is a pattern worth knowing about, even though this prompt isn't the place to resolve whether it's a problem for this specific site. Output as a table: Source Page | Target Page | Suggested Location | Anchor Text | Reason a sideways link is warranted (skip this last column for pillar-up links, mark those simply as "pillar link").
Customize
Optional — swap in your own details for the highlighted parts above.
Why this works
Asked for an internal linking map without constraint, GPT-5.1 defaults to recommending links based on topical adjacency alone, because "these pages are about a related subject" is the cheapest justification the model can produce without reasoning about actual reader flow — explicitly rejecting adjacency as a sufficient reason forces it to simulate where in a specific page's content a reader would hit a natural jumping-off point, which produces links that mirror how a human editor actually reads through a page rather than a graph of topical similarity. Requiring a location within the source page's existing structure, not just a source-target pair, matters because a link recommendation without a placement is not actionable — a content editor implementing the map needs to know where in the text the sentence lives, not just that a link should exist somewhere on the page. The instruction against bare keyword-matched anchor text addresses a specific, checkable pattern: search engines can treat unnaturally repeated exact-match anchors pointing to the same URL as a manipulation signal, and instructing the model to describe what the reader gets rather than restate the keyword produces anchor text that reads naturally in running prose while still being descriptive enough to carry relevance signal. Flagging repeated near-identical anchors as a pattern worth knowing about — without asserting it as a confirmed problem — respects the brief's constraint against fabricating specific claims as fact, since whether repeated anchor text is actually risky depends on site-wide context this prompt doesn't have visibility into.
What you get back
Source: 'Remote Onboarding Best Practices' | Target: 'Onboarding Software Comparison' | Location: near the paragraph discussing async check-in tools | Anchor: 'compare tools built for distributed teams' | Reason: page mentions using software for remote check-ins in passing; target page is the deep comparison a reader evaluating options would want next.
Verified against
ChatGPT GPT-5.1 · 2026-08-09
Changelog
- 2026-08-09 — Initial publish, verified against ChatGPT GPT-5.1.
Building this for real?
This is a free starting point. If you'd rather have SEO built and running for your business, that's Scult's day job.
EXPLORE SEO
