AI citation audit

Summary#

The Free AI Citation Audit scores whether a page gives an answer engine something worth quoting and a reason to attribute it to you. It fetches the page, its robots.txt and its llms.txt, scores six weighted signals, and returns a 0-100 score with ranked fixes. It runs at /free-tools/ai-citation-audit with no account.

Purpose#

AI answers name their sources. Being named is different from being readable: a page can be perfectly crawlable and well structured and still be passed over because it contains nothing specific enough to quote, or because nothing on it says who wrote it or when.

This audit weights those two things heavily. Concrete facts are worth the most points in the set, and author plus date signals are close behind, because they are what separates a page an engine will attribute from one it will paraphrase anonymously.

Overview#

Three fetches run in parallel: the URL, /robots.txt and /llms.txt. Six checks are scored with fixed weights totalling 100.

Two checks are unique to this tool. Citable facts and data counts concrete numbers in the visible text: thousands separators, percentages, currency amounts, four-digit years and magnitude words such as million, billion, thousand and percent. Five or more matches passes. Author and trust signals requires both a named author signal and a machine-readable date.

Benefits#

• Measures quotability rather than readability, which is the gap most content audits miss.
• Counts your concrete facts and tells you how many it found.
• Checks author and date together, because a citation needs both.
• Runs on any public URL with no account and no credits.

Use Cases#

• Turning a good article into a citable one. The fix list usually points at adding specifics rather than rewriting.
• Auditing thought-leadership content. Opinion pieces score badly on citable facts, which is useful to know before you commission more of them.
• Checking a byline template. If the author markup is missing site-wide, one run tells you.
• Comparing with a rival that keeps getting cited. Run it on their page and count the difference in facts and trust signals.

Requirements#

A public URL returning HTML. Private, loopback and internal hosts are refused.

Permissions#

Anyone. No account, plan or role. A shared hourly fair-use limit applies per IP address across the free tools.

Cost#

Free. Zero credits.

Navigation Path#

metricvaultai.com → Free tools → AI Citation Audit

The page is at /free-tools/ai-citation-audit.

Inputs#

Helper line: No signup · No credit card · Tests the live page, robots.txt and llms.txt.

Step-by-Step Guide#

1. Open /free-tools/ai-citation-audit.
2. Enter the page in the field marked example.com.
3. Select Run citation audit. The button reads Checking… and the result area shows Reading the page, robots.txt and llms.txt, and scoring AI readiness….
4. Read the score, then work Priority fixes from the top.

Reading the Results#

The score ring. 0-100 with a tier badge:

The heading reads <N> of 6 AI-readiness checks passed, with the bands 80-100 strong, AI can cite you, 55-79 partly ready and 0-54 at risk.

Citation signals we score. The six checks:

Citable facts and data reports the count either way. On a pass it reads <N> concrete stats/dates an answer can quote and attribute; on a failure, Few concrete facts (<N> found). Add specific numbers, dates and stats engines can cite.

Author and trust signals names the missing half: No visible author. Add a named author and bio. or No publish/updated date. Add one so engines trust the page.

Priority fixes. Failed checks, numbered and weight-sorted.

Examples#

Screenshots#

Tips#

• The fact counter reads visible text only, so numbers hidden in images or charts do not count. Put the key figures in prose as well.
• A date in datePublished or dateModified satisfies half of the trust check even when the design shows no date.
• Question/FAQ format passes on a heading that ends in a question mark, so it is often the cheapest remaining points.
• Re-run after each edit. Nothing is cached.

Best Practices#

• Lead with specifics. One sentence carrying a number, a date and a source is more citable than a paragraph of adjectives.
• Give every substantive page a named author and a real date, in markup as well as on screen.
• Keep the facts accurate and dated. An engine that quotes a stale number quotes it in your name.
• Run GEO audit on the same URL. That tool checks whether the page can be read as an answer; this one checks whether it deserves the attribution.

Common Mistakes#

• Inflating the fact count with filler numbers. The check counts numeric patterns and cannot judge relevance, so padding raises the score without improving the page. That helps nobody.
• Assuming the score predicts citations. It measures signals on your page, not what any engine does.
• Adding an author name in an image or a graphic. It has to be in the HTML.
• Expecting the result to persist. Anonymous runs are not stored.

Limitations#

• One URL per run. No crawl.
• The fact counter is a pattern match. It cannot tell a meaningful statistic from a phone number or a price list, and it can be gamed. Read the count as a rough proxy.
• The author check looks for author markup, a rel="author" link, an author meta tag or a byline class. An unconventional byline pattern can fail it even when a human sees a name.
• Server-rendered HTML only.
• No model is queried, so the tool cannot report actual citations. That is AI Citation Tracker in the app.
• No export, history, scheduling or recommendations panel. See Free tools, no account needed.

Troubleshooting#

FAQs#

What is an AI citation? A named source in an AI-generated answer. The assistant summarizes the answer and credits the pages it drew from, usually with a link.

How is that different from a backlink? A backlink is another site linking to you. A citation is an assistant naming you inside an answer. They can move independently, and this tool measures only the on-page conditions for the second.

Does this cover ChatGPT, Gemini, Perplexity and Claude? The signals it checks, crawler access, structured data, concrete facts, author and date, are the ones answer engines generally rely on. It does not query any of them. Measuring which engines actually cite you over time is AI Citation Tracker.

Is it really free? Yes. No account, no card, no credits.

How often should I check? Once per substantive page, and again after any template change that touches author, date or schema markup.

What does the paid product add? Ongoing measurement of which sources the engines cite for your topics, whether you are among them, and how that changes. See What you get on Free.