AnswerRank Quick Start: From Prompts to Your First Sources Strategy
8 min read
AI tools like ChatGPT and Gemini read websites before they answer. If those websites talk about your brand, the AI is more likely to recommend you. AnswerRank shows you which websites the AI reads, and this guide gets you from zero to a first plan in three steps: add your prompts, read the Sources report, and turn it into a to-do list.
A full worked example of one possible brand and its path is at the end of this guide, separate from the steps.
New to AnswerRank? Start with How to Create Your First Prompt to Track in AnswerRank and How to Start Tracking Your First Prompt in AnswerRank.
Words you need to know
| Word | Meaning |
|---|---|
| Prompt | A question people ask the AI, such as "best moisturizer for dry skin" |
| Source | A website the AI used to write its answer |
| Citation | One time the AI used that website |
| Prompt Coverage | How many of your prompts that website appeared in |
Step 1: Add your prompts
Your prompts decide what AnswerRank measures, so write the questions your real customers would ask an AI. Aim for 20 to 50 to start. For a detailed walkthrough, see How to Create Your First Prompt to Track in AnswerRank.
1. Think like a buyer, not a brand. Most prompts should not mention your brand name. People ask "best moisturizer for dry skin", not "is [your brand] good". 2. Mix four kinds of questions: - "Best" questions: *best moisturizer for sensitive skin* - Problem questions: *how to stop dry skin in winter* - Comparison questions: *[your brand] vs [competitor] moisturizer* - Buying questions: *where to buy vegan moisturizer online* 3. Add them in AnswerRank and let the tool run them across the AI models. See How to Start Tracking Your First Prompt in AnswerRank. 4. Wait for the first results before changing anything.
Step 2: Read the Sources report
Open the Sources report once your prompts have run. Everything below is done inside the tool, and you can export the list when you need it.
- Sort by Citations. Click the arrows next to "Citations" and look at the top 10 to 15 websites.
- Check the Category column. It tells you the type of each website: Retailer, Editorial, Review, Corporate or Community. Note which type appears most.
- Compare Citations with Prompt Coverage. Divide Citations by Prompt Coverage. A low number means the site appears in many answers (wide). A high number means the AI leans on it heavily for fewer questions (deep). Both matter.
- Check the Model icons. More icons means more AI tools trust that website.
- Open a website to see the exact pages the AI used. AnswerRank shows whether you and your competitors are mentioned on each page. Most important: check the context of each mention. Are you recommended, ranked low, compared unfavorably, or only listed in passing?
- Find yourself. Type your own domain into "Filter Domain" to see if the AI uses your site.
- Export the main sources so your team can work from one shared list.
Step 3: Build your first strategy
Give each top website a label, then pick one action per gap. Fix Missing first, then Weak.
- Present: you are there and the information is complete and correct.
- Weak: you are there, but the information is short, old or wrong.
- Missing: you are not there at all.
Use the website type to choose the action:
| Website type | Why the AI uses it | What to do |
|---|---|---|
| Retailer | To check products, prices and reviews | Get listed, fill in every detail, collect real reviews |
| Editorial | For expert advice and "best of" lists | Contact the writers of the exact articles the AI uses |
| Review | To compare products and ratings | Ask to be included in comparisons |
| Corporate | For facts from the brand itself | Publish clear guides and FAQs on your own site |
| Community | To hear what real people think | Join conversations honestly, encourage happy customers to share |
One person can start this in a week. Check the Sources report again in 4 to 6 weeks.
Common mistakes
- Using only brand-name prompts. You will only see sources for people who already know you.
- Looking at one number. Always read Citations and Prompt Coverage together.
- Trusting the famous name. A small expert site can matter more than a big one.
- Vague plans. "Get into that magazine" is hard. "Get into this exact article" is clear.
- Expecting instant results. AI tools update over time. Give it a few weeks.
Check your progress every month
- Filter for your own domain: are Citations and Prompt Coverage going up?
- How many websites moved from Missing to Weak, or Weak to Present?
- Do more Model icons show next to your website?
- Are there new websites in the top list? Add them to the plan.
- Do your prompts still match what customers ask? Add new ones if needed.
For a wider view of which numbers to report, see How to Measure AI Visibility: The Metrics That Actually Matter.
Example: Velunia, step by step
Here is an example of one possible brand and its path. Velunia is a small skincare brand that sells moisturizers on its own website and through a few online stores. Customers like the products, but when people ask ChatGPT "what's the best moisturizer for dry skin?", Velunia rarely comes up. Here is how they can use AnswerRank to find out why and what to fix first.
Step 1: Adding prompts
Velunia adds 30 prompts: 12 "best" questions, 10 problem questions, 5 comparisons and 3 buying questions. Only the 5 comparisons use a brand name.
Step 2: Reading the Sources report
Velunia's 30 prompts have run. They have never used the tool before, so they go one question at a time. For each one, they write down what they see and what it tells them to do.
1. Which websites does the AI read most? They sort by Citations. The top 10 are 6 online stores, 3 beauty magazines and 1 competitor's brand website. - *Direction:* the AI mostly checks stores in this category, so stores come first. 2. Is each website wide or deep? One big store appears in 25 of 30 prompts with 50 citations (2 per prompt): wide. A skincare magazine appears in only 6 prompts but has 40 citations (almost 7 per prompt): deep. - *Direction:* the store gets them seen in many answers. The magazine is the AI's expert on dry skin, so being in it gets them recommended for those exact questions. 3. Are we there? They open each store in the tool. Velunia is on 4 of the 6, but 2 of those pages have only one sentence. They are missing from the other 2. - *Direction:* fix the 2 thin product pages first. It's the fastest win, then work on getting listed in the missing 2. 4. How are we mentioned? The 3 magazine pages the AI uses are "best moisturizers" lists. Two competitors are named "best for dry skin". Velunia is not mentioned at all. On one store page, Velunia appears, but reviews complain about the pump bottle. - *Direction:* contact the writers of those 3 lists. Reply to the reviews and fix the packaging message, because the AI repeats what it reads. 5. Is our own website a source? Filtering for velunia.example shows zero citations. The competitor's brand site is cited often, mainly for its "How to choose a moisturizer for your skin type" guide. - *Direction:* the AI trusts brand websites here, so Velunia should publish its own clear guide. 6. Save the work. They export the top 15 sources and add a Present, Weak or Missing label to each. This list becomes their plan in Step 3.
Step 3: Velunia's first plan
| Website | Label | Action |
|---|---|---|
| 2 stores | Missing | Apply to sell there, or go through a distributor |
| 2 stores | Weak | Add full descriptions, ingredients, skin types and how to use |
| 3 "best moisturizers" lists | Missing | Email the writers with samples and product facts |
| Own website | Missing | Write a "How to choose a moisturizer for your skin type" guide |
| Skincare magazine (deep source) | Missing | PR: send a short story to the editor, such as a new product, a key ingredient or founder tips for dry skin |
| Beauty blogs cited for dry-skin questions | Missing | Offer a guest post or expert quote, for example "5 winter mistakes that dry out your skin", with a link to Velunia's guide |
| Own website (after the guide is live) | Missing | Earn backlinks from the cited sites to the new guide, so the AI is more likely to find and trust it |
Summary: insights and actions from the sources data
From one sources export, Velunia gets a clear picture of where the AI learns about moisturizers and a short list of what to do about it.
| What the data shows | Insight | Action |
|---|---|---|
| 6 of the top 10 sources are online stores | The AI mostly checks stores in this category | Make stores the first priority |
| One store is wide (25 of 30 prompts), one magazine is deep (7 citations per prompt) | The store brings visibility, the magazine brings recommendations | Keep both on the list: fix the store page, pitch the magazine |
| Listed on 4 stores, 2 with one-sentence pages, missing from 2 | The AI finds little or nothing to say about Velunia there | Rewrite the 2 thin pages, then apply to the 2 missing stores |
| Competitors named "best for dry skin" in 3 lists, Velunia absent | Competitors win the "best of" answers | Email the writers of those exact articles with samples and facts |
| Reviews mention a pump bottle problem | The AI may repeat this complaint | Reply to the reviews and fix the product message |
| Own website has zero citations, a competitor's guide is cited often | The AI trusts brand websites in this category | Publish a "how to choose a moisturizer" guide |
| Editorial sites and blogs are a big share of citations, and the AI opens specific articles | The AI learns from earned media, not ads | Run PR: send product news and expert comments to the writers of the cited articles |
| Some cited blogs publish outside experts and guest authors | A guest article on a trusted site can become an AI source itself | Pitch guest posts on dry-skin topics the prompts cover |
| Cited sites link to competitors' guides, but not to Velunia | Links from trusted sites help the AI find a brand's own pages | Ask for backlinks to Velunia's guide in guest posts, PR coverage and "best of" updates |
In 4 to 6 weeks, Velunia exports the sources again and compares: more citations, more prompts covered, and fewer Missing or Weak labels mean the plan is working.