AI Visibility Audit: Brand Mention and Citation Test Plan for AI Answer Engines

Design and analyze a repeatable prompt test set that checks how AI answer engines mention and cite your brand. It turns results into a citation table, evidence-based gap findings, and prioritized page and entity fixes.

You get: Decision summary, test prompt set and recording sheet template

More like this

The prompt

ai-visibility-audit-brand-mention-and-citation-test-plan-for-ai-answer-engines.md
Role: You are an SEO analyst specializing in AI visibility, covering how answer engines such as ChatGPT, Perplexity, Gemini, and Google AI features mention and cite brands.

Context: {brand} competes with {competitors} for {audience} searching about {topic}. The decision this answer supports: which pages, entities, and sources to improve first so {brand} is represented accurately and cited more often in AI answers.

Task: Phase 1: design a test prompt set and recording sheet. Phase 2: once I paste results, analyze them and recommend actions.

Constraints:
- Do not claim to have run live queries or browsed. You cannot know current AI answers. Work only from results I paste, and label everything else as a hypothesis.
- AI answers vary by run, location, and model version. Treat single runs as anecdotes and recommend at least {runs_per_prompt} runs per prompt per tool.
- Do not promise citations or ranking changes. No fabricated sources or statistics.
- Separate observed evidence, assumptions, and recommendations.

Inputs:
- Brand, key products or services, and canonical URLs: {brand}, {key_urls}
- Competitors: {competitors}
- Target topics and buyer questions: {topic}, {buyer_questions}
- Tools to test and their modes (e.g., with or without web search): {tools_tested}
- Pasted results from a previous test, if any (prompt, tool, date, answer text, cited URLs): {pasted_results}
- Known facts that must be stated correctly (pricing model, locations, features): {ground_truth_facts}
List missing inputs first, then assumptions.

Workflow:
1. Phase 1: write 20 test prompts across types: category discovery, comparison, problem-solution, brand-specific, and reputation. Add a recording sheet with columns for tool, mode, date, run number, brand mentioned (Y/N), position, sentiment, accuracy versus {ground_truth_facts}, cited URLs, and competitors named.
2. Phase 2: from {pasted_results}, build a citation table and compute mention rate per prompt type only from the data provided. Show the counts.
3. Review evidence: list inaccuracies, missing facts, and which domains are cited most often.
4. Diagnose likely causes as hypotheses (content gaps, unclear entity data, lack of third-party mentions, blocked crawlers, thin comparison content) and say how to verify each.
5. Prioritize actions by impact, confidence, and effort.

Tool-specific notes:
- Perplexity: record the listed source URLs for each answer and note whether the cited page is ours, a competitor's, or third-party.
- ChatGPT: record whether search was on, and note any named sources.
- Gemini and Google AI features: note whether answers appear with source links, and record the country and language.

Output format:
1. Decision summary (max 6 lines)
2. Missing information and assumptions
3. Test prompt set (table: ID | Prompt | Type | Intent | Why it matters)
4. Recording sheet template
5. If results were provided: citation table (Prompt ID | Tool | Brand mentioned | Accuracy | Cited URLs | Source type) and a findings table (Finding | Evidence | Confidence)
6. Prioritized action list: Action | Page or entity affected | Hypothesis tested | Impact | Effort | How to verify
7. Risks and next actions

Acceptance criteria:
- Test prompts cover all five prompt types and reflect real buyer questions.
- Every finding cites a row in the pasted data, or is labeled a hypothesis.
- Actions include a way to verify progress on a re-test.
- Run-to-run variability is acknowledged.

Quality checks:
- No invented mention rates, citations, or sources.
- Counts match the pasted data.
- Inaccuracies are compared against the supplied ground truth only.
- Recommendations are specific to named pages or entities, not generic advice.

Example output

AI-written example

1. Decision summary

  • Brand: Quillpay (payroll for UK businesses with 1-20 staff). Competitors: PayNest, Rotaly, LedgerWorks.
  • No results were pasted, so this is Phase 1 only. No mention rates or citations are claimed.
  • Run each prompt 3 times per tool and mode (ChatGPT with and without search, Perplexity, Gemini). Treat single runs as anecdotes.
  • Priority pages to check first: quillpay.example/pricing, /features/auto-enrolment, /compare.

2. Missing information and assumptions

  • Missing: pasted_results, and the country and language setting for Gemini.
  • Assumption: ground truth is "£6 per employee per month, no setup fee, HMRC-recognised, UK only."
  • Assumption: the tests run from a UK location in English.

3. Test prompt set

IDPromptTypeIntentWhy it matters
P1Best payroll software for a 5-person UK business?DiscoveryShortlistCore category query
P2Payroll tools for UK startupsDiscoveryShortlistSegment reach
P3Easiest payroll software for non-accountantsDiscoveryShortlistEase-of-use angle
P4Cheapest HMRC-recognised payroll softwareDiscoveryShortlistPrice framing
P5Quillpay vs PayNest for small teamsComparisonChooseHead-to-head
P6Quillpay vs Rotaly pricingComparisonChoosePrice accuracy
P7PayNest alternatives for UK small businessComparisonSwitchAlternative-seeker capture
P8LedgerWorks vs Quillpay featuresComparisonChooseFeature accuracy
P9How do I run payroll without an accountant?ProblemLearnTop-of-funnel
P10How do I set up auto-enrolment pensions for staff?ProblemLearnFeature page citation
P11How do I submit RTI to HMRC?ProblemLearnAuthority test
P12Payroll software that handles holiday payProblemSolveFeature match
P13What is Quillpay?BrandVerifyEntity clarity
P14How much does Quillpay cost?BrandVerifyGround-truth check
Generated from this prompt by AI. Your results will vary.

Tips and common edits

Run Phase 1 first, execute the prompts manually in each tool, log the results, then paste them back for Phase 2. Re-test on a fixed schedule with the same prompts to see change over time. Keep a record of tool, mode, location, and date, since answers differ between runs.

Questions about this prompt

More in SEO

GEO Page Audit: Citation Likelihood Score and Rewrites for AI Answers

Get a scored estimate of how citable your page is to ChatGPT, Gemini and Perplexity, plus a summary box, question headings, fact prompts and an FAQ.

AI Visibility Prompt Tracker List: 25 Buyer Prompts by Funnel Stage

Produces 25 realistic buyer prompts grouped by funnel stage, ready to run monthly to track whether AI assistants mention your brand.

SERP-Based Content Brief with Search Intent Analysis for a Single Keyword

Create a writer-ready content brief from SERP observations you supply. It covers intent, required sections, evidence gaps, and a go or no-go call on whether the page is worth producing.

Long-Form Article Brief With Source Discovery Plan and Citation Table

Creates an editorial brief and outline for a researched article. It includes a source discovery plan, an evidence review template, and a citation table, without pretending any live browsing took place.

Productivity Tool Selection: Source Discovery, Evidence Review, and Citation Table

A research workflow to choose between candidate productivity or automation tools. It plans online source discovery, structures evidence review in a citation table, and ends in a decision memo without claiming live browsing took place.

Decaying Content Refresh Plan: Outdated Facts, Gaps and a New Intro

Audits an old article for outdated facts and missing subtopics, proposes new sections and rewrites the intro for your target keyword.

FAQ Block for Rich Results: 8 Customer Questions and FAQPage JSON-LD

Extracts 8 genuine customer questions with standalone answers from your page and outputs matching FAQPage JSON-LD.

Keyword List to Content Plan: Intent Groups, Page Types and Titles

Turns a raw keyword list into intent-based page groups, each with a recommended page type, title and build order.

Meta Title and Description Variants: 5 Click-Angle Pairs Within Limits

Writes 5 meta title and description pairs within character limits, each using a different angle to earn the click.

Technical SEO Crawl Triage: Prioritized Fix Plan for the Next Sprint

Turn a crawl export and Search Console data into a scored, sprint-ready technical SEO fix list. It separates facts from assumptions and flags what needs verification before developers start work.

Topic Cluster Map: Pillar and Supporting Pages Ranked by Build Order

Build a topic cluster from a keyword list and existing content, then rank which cluster and pages to create first. It includes a keyword-to-page mapping table, internal link rules, and cannibalization risks.

Questions Buyers Ask ChatGPT and Perplexity About Your Category, With a Plan to Get Cited

Lists the questions buyers ask AI assistants about your product category, checks which sources those answers lean on, and plans the pages and facts you need to be cited.