An AI visibility audit is the starting point for any serious GEO/AEO strategy. Without a baseline, you can’t measure progress, prioritize correctly, or justify investment. This guide walks through a complete audit — from query set construction to gap prioritization — that any brand can run in 2–4 hours.
What You’re Auditing and Why
An AI visibility audit answers four core questions:
- Presence: Does your brand appear in AI responses for your target queries?
- Positioning: When you appear, how prominently — first mention, listed, or buried?
- Accuracy: When AI engines describe your brand, are the claims correct?
- Competitive standing: How do you compare to the 2–3 competitors you most often lose to?
These four questions map directly to the levers available to you: content creation (presence), content optimization (positioning), brand safety (accuracy), and competitive strategy (standing).
Step 1: Build Your Query Set
Your audit is only as useful as the queries you test. Build a set of 30–50 queries across three tiers:
Tier 1 — Category discovery queries (highest priority)
These are the queries your customers ask before they’ve decided which brand to use. They’re also where the biggest competitive wins live.
- “Best [your product category] for [primary use case]”
- “How do I [core problem you solve]”
- “What tools do [your target job title] use for [core workflow]”
- “Compare [Competitor A] vs [Competitor B]”
- “[Your category] software recommendations”
Tier 2 — Brand evaluation queries
Queries buyers ask once they’ve heard of you but haven’t committed.
- “[Your brand name]”
- “[Your brand] vs [top competitor]”
- “Is [your brand] good for [use case]”
- “[Your brand] pricing”
- “[Your brand] alternatives”
Tier 3 — Problem/use case queries
Longer-tail queries from buyers who describe their situation rather than naming a category.
- “[Specific pain point your product solves]”
- "[Industry] + [workflow] + “tools”
- “How to [specific task your product does]”
Document these in a spreadsheet before running any queries. Having a fixed set ensures your audit results are reproducible and comparable over time.
Step 2: Run Queries Across Five AI Engines
Test each query on all five major AI engines. For each query/engine combination, record:
| Field | What to capture |
|---|---|
| Brand present? | Yes / No |
| Mention position | First / Prominent / Listed / Absent |
| Sentiment | Positive / Neutral / Negative / Mixed |
| Accuracy | Correct / Partially correct / Incorrect |
| Citation | Link included? URL |
| Competitors mentioned | List all brands named in the same response |
Do this in a clean, logged-out browser session. AI engines personalize responses in some contexts — you want baseline results, not personalized ones.
Engines to test:
- ChatGPT (web, default GPT-4o model, with and without browse mode)
- Perplexity (default search mode)
- Google AI Overviews (search the query in Google; note whether AI Overview appears)
- Claude (claude.ai, default model)
- Microsoft Copilot
Run each query at least twice. Responses vary due to temperature; two runs surface variance.
Step 3: Assess Accuracy and Brand Safety
For every response where your brand is mentioned, fact-check the claims against your current product reality:
Common accuracy issues to watch for:
- Outdated pricing or plan names
- Deprecated features described as current
- Incorrect category positioning (“X is an analytics tool” when you’ve repositioned as a data platform)
- Wrong founding year, CEO name, or headquarters
- Capability overclaims or underclaims
- Conflation with a competitor or similarly-named company
Score each mention: Accurate, Partially accurate (some correct, some wrong), or Inaccurate (materially wrong or misleading).
Any inaccurate mention is a brand safety issue. Flag these for a remediation plan — they’re addressable but require a different playbook than content creation. See How to Fix AI Brand Safety Issues.
Step 4: Map Competitor Presence
For every query where you’re absent, note which competitors appear. This gives you a content gap map:
Query: "Best project management tools for remote teams"
Your brand: Absent
Competitors present: Asana (first), Monday.com (prominent), Notion (listed)
After running your full query set, count how many queries each competitor appears in that you don’t. Sort by frequency. The competitors appearing most often in your absence are your highest-priority competitive targets — their content strategy is working somewhere yours isn’t. To turn this into an ongoing discipline, see competitor benchmarking and content gap analysis.
Step 5: Calculate Your Baseline Scores
Once your spreadsheet is complete, calculate these baseline metrics:
Impression rate = (Queries where you appear) ÷ (Total queries tested) × 100
- Across all engines
- Per engine (expect significant variance)
Average mention position = Weighted score where First=3, Prominent=2, Listed=1, Absent=0; divide by total queries
Accuracy rate = (Accurate + Partially accurate mentions) ÷ (Total mentions) × 100
Competitive share of voice = Your impressions ÷ (Your impressions + Competitor impressions for same queries) × 100
Document these as your Day 0 baseline. You’ll compare every subsequent audit against these numbers — a composite visibility score is just these four numbers compressed into one line on a chart, which is useful for reporting but no substitute for keeping the components.
Step 6: Prioritize Gaps by Impact
Not all gaps are equal. Prioritize which content gaps to close first using this framework:
High priority: Queries in Tier 1 where a top competitor appears and you don’t, and where the query has clear commercial intent.
Medium priority: Queries where you appear but only in a listed position, and competitors appear first or prominently.
Lower priority: Tier 3 long-tail queries where you’re absent but no competitor dominates either.
Brand safety first: Any query producing inaccurate claims about your brand overrides all content priorities. Misinformation spreading across all users of an AI engine does compounding damage.
Step 7: Identify Technical Barriers
Some absence isn’t a content problem — it’s a crawlability or indexability problem. Check:
For Perplexity / AI Overviews / Copilot (RAG-powered):
- Is your site crawlable? Check robots.txt for AI crawler blocks (GPTBot, PerplexityBot, Googlebot). Google’s robots.txt guide is also a useful corrective on what the file does and does not do: it manages crawler traffic, it is not a way to keep a page out of results, and a disallowed URL can still be listed if other sites link to it
- Is your sitemap current and submitted?
- Do your key pages load fast? (slow pages get crawled less frequently)
- Is your content behind a login or paywall?
For ChatGPT / Claude (training-data engines):
- Is there substantial third-party coverage of your brand?
- Have you existed long enough for training data to include you? (brands <12 months old are often underrepresented)
- Is your category niche enough that LLMs simply have less data on it overall?
Technical barriers require different fixes than content gaps. Identify them before building a content plan.
Building Your Audit Into a Recurring Practice
A one-time audit gives you a baseline. A recurring audit gives you a trend line. Schedule:
- Monthly full audits against the same 30–50 query set (pair these with ongoing multi-engine monitoring)
- Biweekly spot checks on your 10 highest-priority queries
- Event-triggered checks after major product launches, pricing changes, or significant press coverage
Store results in a versioned spreadsheet. Trends matter more than snapshots — a brand with 30% impression rate increasing to 45% over 90 days is doing something right, even if the absolute number is still modest.
The audit is the foundation. Everything else — content creation, authority building, schema markup, earned media — flows from knowing specifically where your gaps are and which ones matter most.
Frequently Asked Questions
How do I audit my brand’s AI visibility?
Build a fixed set of 30–50 queries across category, brand, and use-case tiers; run them across the major AI engines in a clean session; and record for each whether you appear, your position, sentiment, accuracy, citations, and which competitors show up. Then calculate baseline scores and prioritize the gaps.
How long does an AI visibility audit take?
A thorough first audit typically takes 2–4 hours for a 30–50 query set across five engines. Recurring spot checks on your top 10 queries take far less, and the value compounds because you’re tracking a trend against your Day 0 baseline.
Which AI engines should I include in an audit?
At minimum ChatGPT (with and without browsing), Perplexity, Google AI Overviews, Claude, and Microsoft Copilot. Each sources and cites differently, so per-engine results reveal whether a gap is a retrieval problem or a training-data problem.
What metrics should an audit produce?
A baseline impression rate (overall and per engine), average mention position, accuracy rate, and competitive share of voice. Document these as Day 0 so every later audit measures progress against them.
How often should I re-run the audit?
Run a full audit monthly against the same query set, do biweekly spot checks on your highest-priority queries, and run event-triggered checks after launches, pricing changes, or major press. Trends over 30–90 days matter more than any single snapshot.