How to Audit AI Search Citations at the URL Level
When you audit AI search citations at the URL level, you stop asking the vague question “Is our brand visible in AI?” and start asking a precise one: “Which of our pages do AI platforms actually use, and why?” That shift matters. Most websites have a small number of pages that earn the majority of AI citations, a large group that is never cited, and a few pages that are cited for the wrong topics. This guide explains how to find each group and decide where content improvements are genuinely justified.
Why URL-Level Auditing Matters
Brand-level visibility tells you whether you are winning overall. URL-level data tells you what to do on Monday morning. If your pricing page is cited constantly while your detailed service pages are ignored, that is a specific content and architecture problem with a specific fix. It also stops teams from rewriting pages that are already performing well. Understanding how AI chooses sources gives useful context before you begin.
Step 1: Collect Citation Data
From Prompt Testing
Run your tracked prompt set across ChatGPT with search, Perplexity, Gemini, Google AI Mode and Copilot. For every answer, record each cited URL, not just the domain. Normalise URLs by removing tracking parameters and trailing slashes so the same page is not counted twice.
From Analytics
Export landing pages for AI assistant sessions from GA4. A page that receives AI referral clicks is being cited somewhere, even if your prompt set missed it. Add these URLs to the audit.
From Server Logs
Filter logs for user-triggered AI fetchers such as ChatGPT-User and Perplexity-User. These requests often happen when an assistant retrieves a page to answer a live question, which makes them a useful proxy for citation demand.
Step 2: Build the URL Inventory
List every indexable URL on the site, then join the citation data to it. A spreadsheet works well with these columns:
| Column | Purpose |
|---|---|
| URL and page type | Group service, blog, product and location pages |
| Target topic | What the page is meant to answer |
| Citations by platform | Counts from prompt testing |
| AI referral sessions | Clicks from analytics |
| AI fetcher requests | Retrieval demand from logs |
| Organic clicks | Traditional search performance |
| Last updated | Freshness |
Step 3: Audit AI Search Citations by Segment
Sort pages into four groups:
- Cited and relevant: cited for the topics they target. Protect these pages and avoid unnecessary changes.
- Cited for the wrong topic: for example, a blog post cited for a service query where you have a better service page. This signals weak internal linking or a service page that does not answer the question clearly.
- Ignored but important: high-value pages with zero citations. These are your priority candidates.
- Ignored and low value: thin or outdated pages that may be better merged or removed.
Check Whether Ignored Pages Are Even Reachable
Before rewriting an ignored page, confirm AI crawlers can fetch it, that the main content appears in raw HTML, and that robots.txt does not block relevant user agents. Many “ignored” pages are simply invisible. Our guide to fixing AI crawler indexing covers these checks, and why content is not showing in AI search explains other common causes.
Step 4: Compare Cited and Ignored Pages
Look for patterns that separate the two groups:
- Does the cited page answer the question in its first paragraph?
- Does it include specific facts, numbers, definitions or comparison tables?
- Does it cite sources or show first-hand experience?
- Is it well linked internally from related pages?
- Does it have clear headings that match how people phrase questions?
- Is the author and organisation clearly identified?
These comparisons, drawn from your own site, are more reliable than generic advice because they reflect what platforms reward in your niche.
Look at Structure as Well as Words
Cited pages often share structural traits: short paragraphs, descriptive headings, lists for steps, tables for comparisons and a visible date. Ignored pages are often long blocks of marketing copy with few facts. When structure is the gap, FAQ optimization services and clearer formatting can close it without a full rewrite. Record these structural traits in your spreadsheet as simple yes or no columns so patterns become obvious when you filter the data.
Step 5: Decide Where Improvements Are Justified
Not every ignored page deserves work. Prioritise using three factors:
- Business value: does the page support revenue, leads or bookings?
- Prompt demand: do tracked prompts on this topic exist, and do competitors get cited for them?
- Gap size: how far is the page from the pattern of your cited pages?
High value, high demand and a large gap means rewrite. Low value and low demand means leave it, merge it or retire it. Document the reason for each decision so future audits can check whether the work paid off. When rewriting, apply proven methods from AI content optimization and strengthen the page with structured data optimization where relevant.
Step 6: Re-Audit and Measure Impact
Repeat the audit every month or quarter using the same prompt set. Keep your blog writing services or in-house writers informed of the results, so new content follows the patterns that earn citations from the start. Track whether improved pages begin to earn citations and whether pages cited for the wrong topic are replaced by the right page. Keep unchanged pages as a comparison so you can tell genuine improvement from platform-wide shifts.
Common Audit Mistakes
- Counting domain citations instead of specific URLs
- Ignoring citations of PDFs, images or old URLs that redirect
- Rewriting top-cited pages and accidentally removing the content that earned citations
- Relying on one platform when others behave differently
- Forgetting that citations from Google AI Overviews require separate prompt checks
Frequently Asked Questions
How Many Pages Should the Audit Cover?
Start with all commercial pages plus your top 100 informational pages. Expand to the full site once the process works.
What If Only a Few Pages Are Ever Cited?
That is common. Study why those pages win, then apply the same structure to similar topics rather than spreading effort across the whole site.
Should Old Cited URLs Be Redirected?
Keep cited URLs live and stable where possible. If you must change them, use permanent redirects to the closest equivalent page.
A URL-level audit turns AI visibility into a clear, prioritised work plan. For hands-on help, see our AI indexing case study, learn how to get featured in AI answers, explore on-page SEO optimization, or review our AI visibility optimization service.