AI Citation Audit Checklist: Test 4 AI Engines
RedHub AI Editorialupdated August 17, 20266 min read
Jump to a section9
TL;DR
- What it is: A step-by-step AI citation audit checklist — how to test whether ChatGPT, Claude, Perplexity, and Gemini name your business, in about 90 minutes.
- Who it's for: Operators and agencies who want a repeatable baseline, not a one-off spot check — see the AEO Citation Audit & Optimizer Kit.
- How it works: Build 30–50 buyer-intent queries across four funnel stages, run them through the four engines, save the responses as evidence, score on five axes.
- Bottom line: Save your day-1 responses. Engines shift daily — evidence beats memory when you re-audit.
What is an AI citation audit checklist?
An AI citation audit checklist is the ordered set of steps for testing your visibility in AI answers: define your niche and ideal customer, generate a buyer-intent query set, run every query through ChatGPT, Claude, Perplexity, and Gemini, save each response, and score the results on a fixed rubric. Following the same checklist every time is what makes your day-30 re-audit comparable to your day-1 baseline.
Best for: a first baseline audit — the AEO Citation Audit & Optimizer Kit ships the query generator and scoring rubric this checklist runs on.
Most people "audit" their AI visibility by asking ChatGPT one question, wincing at the answer, and closing the tab. That is a mood, not a measurement. A real audit uses a fixed query set, all four major engines, saved evidence, and a rubric — so the number you get on day 1 means something when you re-check on day 30.
This is the full checklist. It is the baseline step of the loop described in our AEO audit pillar — read that first if you want the why; stay here for the how.
Before you start: what you need
- Access to the four engines — ChatGPT, Claude, Perplexity, and Gemini. Free tiers are usually enough for a baseline.
- A place to save evidence — a spreadsheet or doc where every response gets pasted with its date, engine, and query.
- A clear picture of your buyer — niche, location if you're local, and the offer you most want to be recommended for.
- ~90 minutes — one focused session. Don't split the baseline across days; answers drift.
Step 1: Build the buyer-intent query set
The audit is only as good as its questions. You want 30–50 queries that mirror what real buyers ask an assistant — not what you'd type into Google. Buyers ask engines in full sentences, with context, at different stages of the decision. Cover four funnel stages:
The buyer doesn't know who exists yet. Example shapes: "Who are the best [service] providers in [city]?" · "What should I look for in a [category] company?" · "I need help with [problem] — where do I start?" These queries test raw Presence: does any engine know you exist?
The buyer is weighing options. Example shapes: "Compare [category] options for a small business" · "[Your niche] vs [adjacent option] — which is better for [situation]?" · "What are alternatives to [big-name competitor]?" These test Prominence: when you're named, where do you land in the list?
The buyer is close to choosing. Example shapes: "Is [your brand] legit?" · "What does [your brand] actually do and what does it cost?" · "Reviews of [category] providers in [city]." These test Accuracy and Sentiment: does the engine describe you correctly, and in what tone?
The buyer already chose — maybe you, maybe not. Example shapes: "How do I get the most out of [category] service?" · "Common problems with [category] and how to fix them." These test Freshness and whether engines associate you with ongoing expertise, not just the sale.
Write the queries in your buyers' words, not your industry's jargon. If you sell "revenue operations consulting," your buyer may be asking "why is my sales team missing forecast." The kit's query generator builds this set from your niche, location, ICP, and offer across 8 industries plus a custom mode — but the principle stands even by hand: four stages, 30–50 queries, buyer language.
Step 2: Run the queries — cleanly
- Use fresh sessions. Start a new chat for each engine so earlier questions don't contaminate later answers. Don't tell the engine who you are — you want the answer a stranger gets.
- Ask verbatim. Run each query exactly as written. Resist the urge to rephrase when an answer disappoints — the rephrase belongs in your next audit cycle, not this one.
- Paste every response into your evidence sheet with the date, engine, and query. This is the step everyone skips and everyone regrets. Answers change daily; your saved day-1 responses are the only proof of your starting point.
- Note the sources. Perplexity and Gemini usually show citations; ChatGPT and Claude name brands inline. Record which pages and domains the engine leaned on — these tell you who currently owns your answer space.
Honesty check: one run is a sample, not a census. Engines are non-deterministic — the same query can name different brands tomorrow. That's exactly why the checklist fixes the query set and saves evidence: you're building a fair before/after comparison, not chasing a single perfect answer.
Step 3: Score what you found
Read your evidence sheet and score the whole set on five axes — Presence, Prominence, Accuracy, Freshness, Sentiment — at 20 points each, for a 100-point Citation Index. The full scoring method, with a working calculator, is in your AI citation score.
The single number tells you how bad it is. The five sub-scores tell you what to fix first. A Presence problem calls for entity and schema work — start with schema markup for AI search. An Accuracy or Freshness problem calls for content and consistency fixes, which the 30-day AEO plan sequences day by day.
The one-page checklist
| # | Task | Output |
|---|---|---|
| 1 | Define niche, location, ICP, offer | One-paragraph audit brief |
| 2 | Generate 30–50 buyer-intent queries, 4 funnel stages | The fixed query set |
| 3 | Run every query in fresh sessions on all 4 engines | Raw responses |
| 4 | Save each response with date, engine, query | The evidence sheet |
| 5 | Record the sources each engine cited | Competitor / source map |
| 6 | Score on the 5-axis, 100-point rubric | Day-1 Citation Index |
| 7 | Name the weakest axis | Your fix priority |
Don't build the audit from scratch
The AEO Citation Audit & Optimizer Kit ($79) ships the query generator (8 industries + custom, 4 funnel stages), the 100-point rubric, the 12-template schema library, and the 30-day playbook that turns your weakest axis into a fix plan. One-time purchase, 30-day refund, unlimited projects.
Get the AEO Audit Kit — $79 →Decision Guide
Use this checklist if: you've never measured your AI visibility, or your last "audit" was one question in one engine.
Skip it if: you already run a fixed query set on a schedule with saved evidence — go straight to the fix work.
Best first step: block 90 minutes this week and run the baseline. Every later decision gets easier with a number in hand.
FAQ
How many queries do I need for a real audit?
30–50, spread across discovery, comparison, decision, and post-purchase stages. Fewer than that and one odd answer skews your score; more and the session drags without adding much signal.
Why test four engines instead of just ChatGPT?
Because your buyers are spread across them, and the engines don't agree. A brand can show up well in Perplexity and be invisible in Gemini. Testing all four — ChatGPT, Claude, Perplexity, Gemini — shows you where the gap actually is.
Do I need to save every response?
Yes. Answers shift day to day, so your saved day-1 responses are the only honest baseline for the day-30 comparison. A screenshot or pasted text with date, engine, and query is enough.
Should I sign in or stay anonymous when running queries?
Use fresh sessions and don't tell the engine who you are. You want the answer a stranger gets, not one shaped by your own chat history.
What if no engine names my business at all?
That's a Presence score near zero — common, and fixable. It usually means your entity signals (schema, canonical naming, citation sources) are too weak for engines to know what your business is. Start with the foundation phase of the 30-day AEO plan.
How is this different from AI visibility strategy?
This checklist is the hands-on measurement layer: run, record, score. The strategy layer — how AI visibility fits your whole marketing motion — is covered by the GEO / AI Visibility Playbook. Audit first; strategize with real numbers.
Run your baseline in one sitting
Query generator, 100-point rubric, schema library, and the 30-day playbook — everything this checklist needs, in one $79 kit.
Get the AEO Citation Audit & Optimizer Kit →