The AI Citation Tracker

Free tracker + prompt · WordsAtScale library

The AI Citation Tracker

A weekly sheet and starter prompt to see if AI answers name your site, and what to publish next.

Download and copy

A weekly sheet and starter prompt to see if AI answers name your site, and what to publish next.

Download the PDF Download the sheet
You are helping me find out whether AI answers mention my site, and what to publish next. Getting named in AI answers starts with knowing which questions real buyers ask.

MY SITE: [e.g. example.com]
WHAT I SELL OR COVER: [e.g. home espresso machines]
WHO BUYS IT: [e.g. first-time buyers on a budget]
MY COMPETITORS: [e.g. 2-3 site names]

Use web search. Never invent a question, number, source or URL. If you cannot find something, say so. Use tables, not paragraphs.

STEP 1. FIND THE QUESTIONS
Search Reddit, forums and YouTube comments for full questions this buyer asks. Copy their wording. Return 10 questions in a table with: question, type, source URL.
Mix these 4 types:
- Best: "What's the best [thing] for [situation]?"
- Versus: "[Option A] or [Option B] for [situation]?"
- Trust: "Is [brand] legit?" or "Is [brand] worth it?"
- How to: "How do I [job] without [problem]?"

STEP 2. ANSWER EACH ONE
Answer each question the way you normally would, with web search on. Then fill this table for each one:
Question | Was MY SITE named? (yes/no) | Was MY SITE linked? (yes/no) | Sites linked instead

STEP 3. FIND THE WEAK ANSWERS
For each question I was not linked on, open the linked pages. Mark the question WEAK if those pages have no real number, repeat the same number, or disagree with each other. Say why in one line.

STEP 4. PICK ONE PAGE
Pick the single WEAK question I should publish on first. Return:
- The question, as the page title
- The answer in two sentences, using only what you verified. Leave [BLANK] for anything I must test myself
- What the page needs that the linked pages lack: a test, a count or a comparison table

The weekly routine

1. Set up (10 minutes). Write 10 full questions your buyer asks, copied from Reddit, YouTube comments and support emails. Mix best, versus, is it legit, and how to.

2. Test in ChatGPT (same day every week). Use a temporary chat. Ask each question 3 times. Open the sources panel and fill in one row per question.

3. Check your numbers (15 minutes). Filter analytics referrals for chatgpt.com. In Search Console, open Search results, then Generative AI, and note your top pages. In the performance report, add a custom regex query filter and paste ^(\S+\s+){6,}\S+ to see searches of 7+ words.

4. Write the page. Answer in the first two sentences. Use the question’s wording in headings. Source and date every number. Put comparisons in a table. Say what you tested and where it falls short. Submit it in Bing Webmaster Tools as well as Google.

Join the community

SearchGAP Method on Skool

Rank a low-authority site fast, with live calls, rank-and-rent playbooks and custom SEO tools. My article writing skills are in there too, built for citations.

Join SearchGAP
Done for you

DataGap Method

Finds the data gap, checks it against Google, and builds the sourced article for you. Disclosure: it’s my own product.

Get DataGap

More free resources

Prompt

The Data-Gap Research Prompt

Finds a number your niche keeps asking for that nobody has published.

Get the prompt
Guide

WeakSpot: Build Your Own Ahrefs in 40 Minutes

A self-hosted keyword research tool with real Domain Rating difficulty scores.

Get the guide

See the full library