Content Cannibalisation Audit and Fix Plan
Find the pages competing for one intent and decide which survives.
Groups a set of URLs by the search intent they actually serve, identifies where several pages compete for the same one, and returns a fix per group — consolidate, differentiate, canonicalise or retire — with the surviving URL and the redirect and link changes each fix requires.
Ready-to-use prompt
The prompt
Copy it as-is, then swap the bracketed placeholders for your own details before running it.
Role: You are an SEO auditor resolving keyword cannibalisation.
Context:
- URLs with titles, target keywords and any performance data: {{url_inventory}}
- Site structure notes: {{site_structure}}
Task:
1. Group URLs by the search intent each one actually serves, not by its stated target keyword.
2. Flag every group containing more than one page as a cannibalisation candidate, and rate the severity — direct overlap, partial overlap, or false positive.
3. For each real conflict, recommend one fix: CONSOLIDATE, DIFFERENTIATE, CANONICALISE or RETIRE.
4. Name the surviving URL and what happens to each other page, including redirects.
5. List the internal links and navigation entries that must be updated for the fix.
6. State the risk of each fix, including what traffic could be lost.
Rules:
- Two pages sharing keywords are not cannibalising if they serve different intents — say so explicitly and move on.
- Every consolidation names one survivor and the fate of the rest.
- Never recommend deleting a page without a redirect target or an explicit reason not to redirect.
Output: intent groups, conflict table with severity, and a fix plan ordered by impact.
- Separate observed evidence from inferred intent, and mark decisions blocked by missing data.
- Make impact ordering provisional when traffic or conversion evidence is unavailable.
- Recommend canonicalisation only for duplicate or near-duplicate pages that must remain accessible; explain why redirection is unsuitable.
Estimated results
Editor's note
Why this prompt matters
Cannibalisation audits go wrong in both directions: teams merge pages that were serving genuinely different intents, or leave four near-identical posts fighting because nobody wants to delete anything. This prompt groups pages by the intent they actually serve, rates how badly they overlap, and forces one named survivor per fix — plus the redirect and internal link work each decision creates.
Anatomy
Prompt engineering breakdown
Role
Context
Goal
Constraints
Output format
What you'll get
Expected output
Scenario and evidence limits
A hypothetical appointment-software site supplies six URLs, page summaries and navigation notes, but no query-level performance data. The summaries support provisional intent grouping; they do not establish that rankings are being suppressed. Impact ordering below reflects the clarity of overlap, not a traffic forecast.
Intent groups and conflict table
| Intent | URLs | Severity | Decision | |---|---|---|---| | Learn how to reduce missed appointments | /blog/reduce-no-shows, /guides/prevent-missed-appointments | Direct overlap | CONSOLIDATE | | Find wording for appointment reminders | /blog/reminder-examples, /templates/appointment-reminders | Partial overlap | DIFFERENTIATE | | Evaluate reminder software | /features/appointment-reminders | False positive against the wording group | Keep separate | | Troubleshoot reminder delivery | /help/reminders-not-sending | False positive against the software group | Keep separate |
The feature page serves purchase evaluation; the help page serves existing users diagnosing failures. Shared reminder terminology is not sufficient evidence of cannibalisation. Neither needs a fix.
Fix plan, ordered by provisional impact
1. Consolidate the missed-appointment advice. Choose /guides/prevent-missed-appointments as the provisional survivor because the supplied summaries show broader coverage and the site structure places it in an evergreen resource hub. Before implementation, compare query coverage, backlinks and conversions; stronger evidence for the blog URL could reverse this choice.
Move useful examples and any distinct advice from /blog/reduce-no-shows into the guide, then permanently redirect the blog URL directly to the survivor. Update the resource-hub card, related-reading links and any in-body references found by a crawl. Remove the redirected URL from the sitemap. These are link locations to inspect, not a verified list of source pages.
Risk: Narrow queries answered only by the blog post could lose relevance if its details disappear during merging. Preserve those answers and record their queries before release.
2. Differentiate the reminder pages. Keep both URLs: /templates/appointment-reminders owns copy-ready wording, while /blog/reminder-examples becomes an explanation of when and why different messages work. Remove duplicated template blocks from the blog and link to the relevant template sections. Rename navigation anchors to distinguish “Reminder templates” from “Choosing reminder wording.” No redirects are required.
Risk: The blog may lose template-seeking visits; evaluate both URLs together rather than treating that shift alone as failure.
Why not canonicalise or retire?
Canonicalisation is not appropriate for these distinct editorial pages. Retirement lacks justification while useful material remains. Check indexation, redirect behaviour and group-level query performance after recrawling before addressing weaker candidates.
Under the hood
Why this prompt works
Grouping by served intent rather than by target keyword is what separates real conflicts from pages that merely share words. The explicit false-positive category gives the model permission to find nothing, which is how you avoid destructive merges, and requiring a named survivor plus link updates makes each recommendation something an engineer can actually execute.
Model fit
Best AI models for this prompt
Claude
Best at declining to merge pages with genuinely different intents.
ChatGPT
Fastest on large inventories and clear on redirect mechanics.
Gemini
Strongest when you paste query-level performance data per URL.
When to use
- When query reports show two URLs repeatedly exchanging visibility for the same user task.
- Before pruning, to establish survivors and redirect destinations.
- After a site merge creates overlapping resource libraries.
- Before briefing rewrites for a cluster with unclear page boundaries.
When not to use
- On a small site whose pages clearly serve distinct tasks.
- To choose new topics without a separate demand and opportunity assessment.
- For parameter duplicates or pagination; use a technical crawl and indexation checks.
- To approve high-value redirects without analytics, backlink evidence and review by the page owner.
Get more from it
Pro tips
- 1
Supply clicks and impressions by query and URL for the same date range; page totals can hide different intents.
- 2
Fix direct overlaps first. Log the release date and wait for recrawling before changing partial overlaps.
- 3
Keep false positives documented with the distinct user task each page completes.
- 4
Update internal links alongside redirects; crawl afterward for links still pointing through redirected URLs.
- 5
Include headings, summaries and conversion purpose. Titles alone cannot reliably establish the intent actually served.
- 6
Add backlinks, conversions and indexability before selecting survivors; the most comprehensive page is not automatically the strongest destination.
- 7
For a second pass, provide disputed groups with page excerpts and query data. Request revised decisions and the evidence that changed them.
Don't ship this
Common mistakes
✗ Merging pages that serve different intents because they share keywords.
Fix — Use the false-positive category and check served intent, not target keyword.
✗ Retiring pages without redirects.
Fix — Require a redirect target or an explicit reason not to redirect.
✗ Leaving internal links pointing at consolidated URLs.
Fix — Treat the link update list as part of the fix, not follow-up work.
People also ask
Frequently asked questions
Q.Do two pages ranking for one keyword always mean cannibalisation?
No. If they serve different intents, both can rank legitimately. The prompt marks those as false positives rather than forcing a merge.
Q.What data makes the audit most accurate?
Query-level impressions and clicks per URL from Search Console. Without them severity ratings are directional only.
Q.Should I consolidate everything at once?
No. Fix direct overlaps first, measure for a month, then decide whether the partial overlaps are worth touching.