CITED: The AI Visibility Operating System

The 5-part operating system that shows where your brand appears in ChatGPT, Claude, Perplexity, Gemini, and AI Overviews, why competitors get named instead, and which sources move the number.

By Matt Shealy, CEO of SwayyEm Built from the audit, citation, Reddit, attribution, and reporting workflows we run across AI visibility accounts.

---

Guide length: ~4,800 words Estimated reading time: 18 minutes Number of parts: 7 Last updated: 2026-06-29

---

Table of contents

Part 1 - Why we built this The reason rankings reports stopped answering the question buyers and boards now ask.

Part 2 - What you will have at the end The 5 modules inside CITED and the weekly report they produce.

Part 3 - The CITED score The simple scoring model for measuring whether your brand exists inside AI answers.

Part 4 - The 6 skills The copy-paste skill files that turn Claude into an AI visibility analyst.

Part 5 - The source map How to find the publications, Reddit threads, review pages, and comparison pages that engines actually cite.

Part 6 - The weekly workflow The 45-minute cadence that turns raw checks into a CMO-ready visibility report.

Part 7 - When to bring in a placement partner The honest line between what you can fix with the system and what still requires editorial relationships.

---

Part 1: Why we built this

Most marketing teams still report on rankings.

That made sense when buyers opened Google, clicked the first few blue links, and compared five vendor pages for themselves.

That is not how the buying journey works now.

Buyers ask:

What are the best tools for this problem?

Which company should I shortlist?

Why is this brand better than the incumbent?

What do real users say about them?

The answer comes back from ChatGPT, Claude, Perplexity, Gemini, or Google AI Overviews. The buyer does not see ten tabs. They see a short answer with a few named brands and a few cited sources.

Your brand is either named there or it is invisible.

That broke the old reporting stack.

An SEO ranking report can tell you where your page sits on Google. It cannot tell you whether ChatGPT names you when a buyer asks who leads your category.

A PR coverage report can tell you how many placements landed. It cannot tell you whether that coverage is changing the answer engines.

A backlink report can tell you who linked to you. It cannot tell you whether the engines trust those sources enough to cite them.

So we built CITED.

CITED is the operating system we wanted every CMO, growth lead, and comms lead to have before they spent another month reading a deck full of rankings, audience estimates, and placement counts.

It answers five questions:

1. Are we named inside the AI answers our buyers see?
2. Which competitors get named instead?
3. Which sources are causing the engines to trust them?
4. Which gaps can we fix ourselves?
5. Which gaps require editorial coverage, expert commentary, or a placement partner?

This is not a theory guide.

It is a working system: 6 Claude skills, a source map, a Reddit and forum workflow, an attribution layer, and a weekly CMO report.

Set it up once. Run it every week. Stop reporting on rankings nobody clicks. Start proving where your brand gets cited.

---

Part 2: What you will have at the end

At the end of this guide, you will have a reusable AI visibility system with 5 modules.

Module 1: The citation audit stack

This checks your brand across ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews using a frozen buyer-question set.

It tells you:

• Which prompts name your brand
• Which prompts name competitors
• Which engines ignore you completely
• Which answers cite sources you can influence
• Which answers cite sources you cannot touch

Module 2: The source authority map

This maps the sources behind the answers.

For every buyer question, the system records:

• Cited URLs
• Publication type
• Whether the source is editorial, forum, vendor-owned, review site, marketplace, or government and education
• Whether your brand is present on that source
• Whether a competitor is present instead
• How hard the source is to influence

Module 3: The Reddit and forum citation workflow

The engines keep pulling from forums because forum answers sound like real human experience.

This module helps you find:

• The Reddit threads engines already cite
• The comparison questions buyers ask
• The comment formats that survive moderation
• The answer structures engines lift
• The rules for mentioning your brand without getting removed

Module 4: The coverage-to-citation tracker

This is the attribution layer.

It connects PR, editorial mentions, podcasts, guest posts, review pages, and forum activity to changes in your AI visibility score.

The goal is simple: stop asking whether a placement got audience. Ask whether it moved the answer.

Module 5: The weekly CITED report

This is the output your CMO can read in 3 minutes.

It includes:

• Overall CITED score
• Engine-by-engine visibility
• Top 3 competitor wins
• Top 5 source gaps
• Best fix this week
• Editorial targets to pursue
• What changed since last week
• What to bring to a placement partner

What this system will not do: manufacture real authority from nothing.

If the engines cite your competitor because they have 40 credible third-party mentions and you have 3, Claude cannot fix that. The system will show you the gap. Closing it still takes editorial strategy, expert commentary, and source-building work.

That is the point.

You should know which problem you have before you spend money trying to solve it.

SUBSCRIBER GATE

---

Part 3: The CITED score

CITED uses a simple scoring model.

You are not trying to measure every possible mention on the internet. You are trying to measure the buyer moments that matter.

Start with 15 prompts.

Use 5 informational prompts, 5 consideration prompts, and 5 conversion prompts.

Informational prompts

These are category-learning questions.

Examples:

what is [category] and who are the leaders
best [category] tools in 2026
how do [buyer type] companies solve [problem]
[category] for [vertical or company size]
what should I look for when choosing a [category] platform

Consideration prompts

These are shortlist-building questions.

Examples:

[competitor A] alternatives
[competitor A] vs [competitor B]
best [category] for [specific company type]
[category] tools with [feature]
which [category] platform is best for [use case]

Conversion prompts

These are brand-specific questions.

Examples:

is [your brand] worth it
[your brand] reviews and results
[your brand] vs [closest competitor]
[your brand] pricing and is it worth the cost
does [your brand] do [key differentiator]

Run the same 15 prompts across 5 engines:

• ChatGPT
• Claude
• Perplexity
• Gemini
• Google AI Overviews

That gives you 75 answer checks.

For each answer, record 5 things:

1. Did it name your brand?
2. Did it name a competitor?
3. Which source did it cite?
4. Did that source mention you?
5. Did that source mention a competitor?

Then score the result.

CITED score = brand-named answers / total answer checks x 100

If your brand appears in 18 of 75 checks, your CITED score is 24.

That number is useful, but the layer split matters more.

If you score 70 on conversion prompts but 8 on informational prompts, buyers can find you only after they already know your name.

If you score 40 on informational prompts but 5 on consideration prompts, the market recognizes your category but does not shortlist you.

If you score 0 on competitor-alternative prompts, your competitor owns the comparison set.

That is where the work begins.

The weekly report should show both:

• The overall score
• The three layer scores

Your goal is not a perfect score. Your goal is movement.

Invisible means the engines do not know you.

Building means the engines know you but do not trust you enough to name you consistently.

Authority means the engines name you in category, comparison, and brand-specific answers.

---

Part 4: The 6 skills

Create a Claude project called:

CITED - AI Visibility OS

Paste this as the project instruction:

You are an AI visibility analyst for a B2B brand.

Your job is to measure where the brand appears inside ChatGPT, Claude,
Perplexity, Gemini, and Google AI Overviews, identify which sources cause
competitors to get cited instead, and produce a weekly CMO-ready report.

You do not invent citations. You only analyze answer outputs, source URLs,
brand mentions, competitor mentions, and user-provided exports.

Every recommendation must name:
1. The buyer question
2. The engine
3. The source creating the visibility gap
4. The competitor currently benefiting
5. The action that can move the number

Never tell the user to "create more content" without naming the exact page,
source, publication, forum thread, or comparison surface that needs to change.

Always end with:
1. CITED score
2. Top 3 gaps
3. Best fix this week
4. Editorial targets
5. What requires a placement partner

Then create these 6 skill files.

Skill 1: citation-gap-finder

File name:

citation-gap-finder.md

Prompt:

You find where the user's brand is missing from AI answers.

Input:
- Brand name
- Competitors
- 15 buyer prompts
- Answer outputs from ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews

Output:
- Prompt
- Engine
- Brand named? yes or no
- Competitor named? yes or no
- Source cited
- Gap type: absent, competitor-owned, source-missing, weak proof, stale proof
- Priority: high, medium, low

Rank gaps by buyer intent first, then competitor risk, then source influence.

Skill 2: answer-engine-scorer

File name:

answer-engine-scorer.md

Prompt:

You calculate the CITED score.

Input:
- A table of prompts, engines, brand mentions, competitor mentions, and cited sources

Output:
- Overall CITED score
- Informational score
- Consideration score
- Conversion score
- Engine-by-engine score
- Competitor share of answer
- 3-line interpretation for a CMO

Keep the math visible. Do not round until the final score.

Skill 3: source-authority-mapper

File name:

source-authority-mapper.md

Prompt:

You map the sources that control AI answer visibility.

Input:
- Cited URLs from answer engines
- Brand mentions on each URL
- Competitor mentions on each URL
- Source type

Output:
- Source URL
- Source type: editorial, forum, review, marketplace, owned, government, education, social
- Influence level: high, medium, low
- Can we influence it? yes, no, maybe
- Best action: pitch, update, comment, contribute, ignore, monitor
- Why this source matters

Prioritize third-party sources the brand can influence within 30 days.

Skill 4: reddit-source-mapper

File name:

reddit-source-mapper.md

Prompt:

You identify Reddit and forum threads that answer engines already cite.

Input:
- Buyer prompts
- Answer outputs
- Cited forum URLs
- Subreddit or forum rules

Output:
- Thread URL
- Query that triggered it
- Engine that cited it
- Is brand mentioned?
- Is competitor mentioned?
- Contribution opportunity
- Comment format to use
- Risk of removal

Never recommend spam. The contribution must be useful enough to survive moderation.

Skill 5: coverage-to-citation-tracker

File name:

coverage-to-citation-tracker.md

Prompt:

You connect editorial coverage and third-party mentions to AI answer movement.

Input:
- New placements or mentions
- Prior CITED score
- Current CITED score
- Source map
- Buyer prompts

Output:
- Placement or mention
- Which prompts it could influence
- Which engine cited it
- Score movement since baseline
- Confidence: high, medium, low
- What to do next

Do not claim causation when the data only supports correlation.

Skill 6: cmo-report-builder

File name:

cmo-report-builder.md

Prompt:

You turn the weekly CITED audit into a one-page CMO report.

Input:
- CITED score
- Layer scores
- Engine scores
- Top gaps
- Source map
- Recommended actions

Output:
1. Executive summary
2. Scorecard
3. What changed this week
4. Top 3 gaps
5. Best fix this week
6. Editorial targets
7. What requires a placement partner

Write in plain English. No jargon. No hedging. No vanity metrics.

---

Part 5: The source map

The score tells you what happened.

The source map tells you why.

For every answer you check, capture the cited sources.

Put them into a table:

Buyer question	Engine	Source URL	Source type	Brand present	Competitor present	Action
best [category] tools	Perplexity	example.com/list	editorial	no	yes	pitch update
[competitor] alternatives	ChatGPT	reddit.com/thread	forum	no	yes	contribute
[your brand] reviews	Gemini	g2.com/page	review	yes	yes	monitor

The important column is action.

There are only 6 actions that matter:

Pitch

Use this when the source is editorial and the page is fresh enough to update, or when the publication covers your category repeatedly.

Update

Use this when the source is your own page, your own docs, a partner page, a marketplace page, or a profile you control.

Comment

Use this when the source is Reddit, Quora, a forum, or a community thread where a real practitioner contribution can survive.

Contribute

Use this when the source accepts outside expert commentary, quotes, contributed articles, research, or data.

Monitor

Use this when the source already names you and the job is to make sure it does not decay.

Ignore

Use this when the source is impossible to influence or too far from your buyer.

This prevents the classic mistake: treating every source gap like a content problem.

Some gaps are content gaps.

Some are credibility gaps.

Some are third-party source gaps.

Some are Reddit and forum gaps.

Some are editorial relationship gaps.

If you diagnose the wrong gap, you waste the month.

---

Part 6: The weekly workflow

Run CITED once per week.

Do not rebuild the prompt set every week. Freeze it for 90 days.

The weekly workflow takes about 45 minutes once the project is set up.

Step 1: Run the answer checks

Run the same 15 prompts across the 5 engines.

Save each answer in a simple doc or sheet.

Record:

• Engine
• Prompt
• Brand named
• Competitor named
• Cited sources
• Notes

Step 2: Run citation-gap-finder

Paste the table into Claude and invoke:

Run citation-gap-finder on this week's answer outputs.

Claude returns the missing-answer map.

Step 3: Run source-authority-mapper

Paste the cited source list and invoke:

Run source-authority-mapper and rank the sources we can influence in 30 days.

Claude returns the source action map.

Step 4: Run reddit-source-mapper

If Reddit, forums, Quora, or community URLs appear in the source list, invoke:

Run reddit-source-mapper on the forum sources and identify contribution opportunities.

Claude returns safe contribution options.

Step 5: Run coverage-to-citation-tracker

Paste new coverage or third-party mentions from the last week.

Invoke:

Run coverage-to-citation-tracker against the latest score movement.

Claude separates correlation from proof.

Step 6: Build the CMO report

Invoke:

Run cmo-report-builder and create the weekly CITED report.

The output should fit on one page.

Here is the report shape:

CITED Weekly Report

Overall score: [x]/100
Informational: [x]
Consideration: [x]
Conversion: [x]

What changed:
- [change 1]
- [change 2]
- [change 3]

Top 3 gaps:
1. [prompt + engine + competitor + source]
2. [prompt + engine + competitor + source]
3. [prompt + engine + competitor + source]

Best fix this week:
[single action]

Editorial targets:
- [publication/source 1]
- [publication/source 2]
- [publication/source 3]

What requires a placement partner:
[honest boundary]

That is the report.

Not rankings. Not audience estimates. Not "visibility was up."

Named answers. Named competitors. Named sources. Named actions.

---

Part 7: When to bring in a placement partner

CITED will show you the gaps.

It will not magically close all of them.

If the answer engines are citing sources you control, you can move fast. Update pages, improve profiles, add proof, fix schema, contribute to active threads, and rerun the check.

If the engines are citing sources you do not control, the work changes.

That usually means editorial.

Examples:

• The source naming your competitor is a category roundup on a publication you have never pitched.
• The engine keeps citing an old comparison article that excludes your brand.
• A competitor is present in analyst commentary and you are absent.
• Reddit threads mention competitors because real users have heard of them, not because the thread needs a better comment.
• AI Overviews cite third-party explainers, not vendor pages.

Those are not prompt problems.

They are authority problems.

Authority comes from third-party sources saying credible things about you repeatedly enough that the engines learn the pattern.

That is where SwayyEm comes in.

We use the CITED report to decide which sources matter, then build the editorial and source strategy around those gaps.

The system tells us where to aim.

The placement work changes what the engines remember.

If you want to run it yourself, use this guide every week for 90 days. If the same source gaps keep showing up, that is your signal. The problem is not the report. The problem is the authority layer.

That is the layer we build.

Comment CITED and I will send you the full operating system.

(Make sure we are connected so the DM lands.)

Repost helps another team stop reporting on rankings and start proving where they are cited.