Peec AI's tracked-query dataset shows AI Overviews appeared on 86.7% of business-intent searches in April 2026, up from 56.9% a year earlier. An AthenaHQ report covered by Business Insider found the average brand's own domain went uncited in 84% of AI answers in its sample. Put those two numbers together and a problem crystallizes: the content AI answers do cite might be yours, but the recommendation inside it might belong to your competitor.
Cannibalization Has Two Layers Now
The old audit was straightforward. Two pages rank for the same keyword, Google rotates between them, neither wins. Ahrefs still frames this correctly: keyword overlap alone doesn't prove cannibalization. The diagnostic is whether pages serve the same search intent and whether Google switches between them over time. That part hasn't changed.
What changed is the second layer. A legacy comparison page can sit in an AI Overview and actively recommend a competitor. The ranking report looks healthy. The page gets impressions. But the cited passage says "Rival A leads for audit-log compliance" based on a feature checklist you retired two product releases ago. That's a brand-risk problem dressed in SEO clothes.
Single Grain's audit framework captures this by adding a "beneficiary" field: for each URL and query cluster, who does this content actually help? Classic cannibalization calls for consolidation or redirect. Competitor-favoring content calls for a rebuild around current buyer criteria, or removal.
The Evidence Standard That Separates Audit From Cleanup
A content refresh queue starts with age and declining clicks. A cannibalization audit starts with query ownership and beneficiary. You need evidence that two pages serve the same intent, that the wrong destination appears, or that an outdated recommendation creates a commercial conflict.
An AI citation to your domain doesn't automatically help you. A cited roundup can accurately report a rival's advantage, and that might be credible editorial work worth keeping. Flag a problem when the recommendation is obsolete or inconsistent with the buying job the page now serves. Don't delete competitor references just because competitors benefit from them. Delete them when they're wrong.
Single Grain's URL-level decision sheet captures this with five fields: URL, query cluster, who benefits, action (consolidate / redirect / rebuild / remove), owner, and due date. Evidence gets linked to each row. No evidence, no action.
Run It This Week: The Four-Decision Framework
Setup: Pull your top 20 buying-intent queries from Search Console. For each, identify every URL that ranks or has ranked in the last 90 days. Screenshot any AI Overview that cites your domain (query, cited passage, source URL).
Step 1 — Check intent overlap. Do the competing URLs serve the same buying job, or distinct stages? If distinct, they aren't cannibalizing. Move on.
Step 2 — Check the beneficiary. Read the cited passage. Does it recommend your current product with accurate claims, or a competitor using outdated criteria? Record the answer per URL.
Step 3 — Pick the action.
- Consolidate when useful evaluation material exists on both pages.
- Redirect only after confirming the destination satisfies the old page's intent. Redirecting to a mismatched page is a visibility loss, not a fix.
- Rebuild when buyers still need a standalone comparison but the current version uses stale criteria.
- Remove only when the page serves no useful purpose and has no appropriate replacement.
Step 4 — Assign and ship. Each row gets an owner and a due date. The SEO lead approves redirects and removals. The content lead approves revised claims. No one autopublishes.
The hypothesis: if we rebuild competitor-favoring legacy URLs around current buyer criteria, then AI-cited recommendations will shift to our product within 60 days because answer engines prefer fresher, more specific content. (An Ahrefs analysis summarized by Security Boulevard found AI-cited content was 25.7% fresher than content in traditional organic results.)
Success = the intended URL serves the buying query and cited passages reflect current positioning. Guardrails = organic traffic to affected URLs doesn't drop more than 15% in the first 30 days. Stop-loss = if a redirected URL loses more than 25% of qualified traffic within two weeks, revert and investigate intent mismatch.
The Trade-Off You're Accepting
This audit will surface pages you'd rather not touch. Some rank well. Some drive decent impression volume. Rebuilding or removing them carries short-term risk: traffic dips, broken internal links, temporary ranking instability. But leaving a legacy page that recommends your competitor in 86.7% of business-intent AI Overviews is a slower, quieter form of damage. It won't show up in last-click attribution. It just keeps making someone else's case to your buyers, one AI answer at a time.
AI visibility statistics vary by dataset and query mix, so treat them as directional. Validate impact on your own tracked queries before scaling beyond those first 20. Most cannibalization audits ask which page should rank. This one asks which page is making your competitor's argument for them. That's the URL to fix first.
