AI Visibility Audit: How to Measure Your Brand's Presence in AI Search

An SEO audit tells you how search engines see your website. An AI visibility audit answers a broader question: when someone asks ChatGPT, Gemini, Google AI Mode, Perplexity or Claude about your category, does your brand show up, is it described correctly, and which sources shape that answer?

The answer depends on more than your website. It depends on your reputation across the web, on whether AI crawlers can read your pages, and on how clearly your facts are stated. This guide gives you a repeatable audit with five pillars, a scoring rubric and a report template.

The five pillars

Pillar	Core question	Where the evidence comes from
1. Access	Can AI crawlers reach and read your pages?	robots.txt, firewall/CDN rules, server logs, rendered vs raw HTML
2. Clarity	Is it obvious what you do, for whom, and at what price?	Your site's key pages, structured data, fact consistency
3. Presence	How often are you mentioned for buyer questions?	A fixed prompt set run across AI engines
4. Accuracy	When you're described, is it correct?	Brand-direct prompts checked against your real facts
5. Authority	Which sources do AI answers cite, and are you among them?	Cited URLs, third-party coverage, reviews

Pillars 1 and 2 are about your own site and are fully in your control. Pillars 3 to 5 measure the outcome, which you influence but don't control.

Step 1: Audit access

Start here, because nothing else matters if AI crawlers can't read you.

Check	How	Pass looks like
AI crawler rules	Read robots.txt for OAI-SearchBot, GPTBot, ClaudeBot, PerplexityBot, Google-Extended, Googlebot	Rules match a deliberate policy (for example, search crawlers allowed, training opt-out if you want it)
Firewall and CDN	Check bot-fighting and WAF settings for blocks on AI user agents or published IP ranges	Allowed crawlers aren't challenged or blocked
Actual crawl activity	Search server or CDN logs for AI user agents	Allowed crawlers visit key pages regularly
Rendering	Compare raw HTML with the rendered page on your top pages	Key text, prices and headings are in the raw HTML
Indexability	Check key pages for noindex, canonical errors and broken links	Key pages are indexable and return 200

Remember that crawler settings are independent. OpenAI, for example, lets you allow its search crawler while blocking its training crawler.

Step 2: Audit clarity

AI engines summarize. Make sure there's a clear, consistent story to summarize.

Core facts on dedicated pages. What you sell, who it's for, pricing, locations, integrations, policies. One canonical page per topic.
Answer-first writing. Important pages open with a direct, one- or two-sentence answer before the detail.
Structured data. Organization, Product or Service, FAQ and LocalBusiness markup where relevant, matching the visible content.
Consistency. Your name, description and prices match across your site, social profiles, review sites, directories and press pages.
Specifics. Numbers, comparisons and original data give AI answers something concrete to repeat.

Step 3: Measure presence

Build a prompt set that mirrors how buyers actually ask. Keep it fixed so results stay comparable over time.

Prompt group	Examples	Suggested share of set
Category discovery	"Best [category] for [customer type]"	40%
Use case / problem	"How do I [problem you solve]?"	20%
Comparison	"[You] vs [competitor]", "alternatives to [competitor]"	20%
Brand direct	"What is [brand]? How much does it cost?"	10%
Reputation	"Is [brand] good? Common complaints?"	10%

Run every prompt several times on each engine you care about, in fresh sessions with no personal memory. Answers vary between runs, so you're measuring rates, not one-off results. Then calculate:

Mention rate: share of answers that name you
Average position: where you appear in lists when named
Share of voice: your mentions as a share of all brand mentions in the set
Competitor overlap: which brands appear when you don't

Expect a gap between recognition and recommendation. A Q2 2026 study by Victorious found that AI platforms described 96% of brands accurately when asked about them directly, yet mentioned far fewer of them in category questions. Category prompts are where an audit earns its keep.

Step 4: Check accuracy

For every answer that describes you, compare it with the truth:

Fact type	What to check	Common problem
What you do	Category, core offering, audience	Old positioning from years ago
Pricing and plans	Prices, tiers, free trial	Outdated or invented prices
Features	Key capabilities, integrations	Missing new features, or a competitor's features attributed to you
Company facts	Location, founding, ownership	Confusion with similarly named companies
Reputation	Strengths and weaknesses mentioned	One old complaint repeated as current

Every inaccuracy should be traced to a source. Usually it's an outdated page of yours, an old directory listing or a third-party article.

Step 5: Map authority

Collect every URL cited in the answers from Step 3 and sort them:

Source type	Example	Your action
Your own pages	Your pricing or comparison page	Keep them updated and strengthen them
Encyclopedic	Wikipedia	Make sure facts about you are accurate and well-sourced
Community	Reddit, forums, Q&A sites	Join genuinely; answer questions; fix misinformation
Media and reviews	Industry press, review sites, comparison lists	Pitch, get listed, earn reviews
Competitors' pages	Their "vs" and "alternatives" pages	Publish your own balanced comparison

Concentration is the norm. Similarweb's early-2026 data showed Wikipedia and Reddit each taking around 12–13% of ChatGPT citations in the US, so the specific sources that dominate your category's answers are where outreach pays off most.

Score it

Rate each pillar from 0 to 4 and keep the scoring rules the same for every audit:

Score	Access	Clarity	Presence	Accuracy	Authority
0	Key crawlers blocked	Core facts missing	Never mentioned	Mostly wrong	Never cited; few third-party mentions
1	Partly blocked or not rendered	Facts scattered or inconsistent	Mentioned only in brand-direct prompts	Several major errors	Rarely cited; thin coverage
2	Reachable, some gaps	Facts present but buried	Mentioned in some category prompts	Minor errors	Cited occasionally; some coverage
3	Fully reachable	Clear, mostly consistent	Mentioned in most relevant prompts	Accurate, small gaps	Often cited; solid coverage
4	Reachable, monitored	Clear, consistent everywhere	Top positions across engines	Fully accurate and current	Frequently cited; strong coverage

A total out of 20 makes progress easy to communicate, but the pillar breakdown is what tells you where to act.

Turn findings into a plan

Lowest pillar	First moves
Access	Fix robots.txt and WAF rules; server-render key content
Clarity	Rewrite key pages answer-first; add structured data; align facts everywhere
Presence	Earn reviews and comparison-list inclusion; publish category and use-case pages
Accuracy	Update stale pages; correct third-party listings and old press
Authority	Targeted PR and outreach to the sources your category's answers cite

Re-run the audit every quarter with the same prompts and rubric, and keep monthly presence checks in between.

Audit report template

Section	Contents
Summary	Total score, pillar scores, change since last audit
Access	Crawler policy, issues found, fixes
Clarity	Pages reviewed, inconsistencies, schema status
Presence	Mention rate, position, share of voice by engine and prompt group
Accuracy	Errors found, likely source, correction owner
Authority	Top cited domains, owned citation rate, outreach targets
90-day plan	Prioritized actions with owners and dates

Key takeaways

Five pillars	Access, clarity, presence, accuracy, authority
Fix the foundation first	Access and clarity are fully in your control
Measure rates	Run a fixed prompt set several times per engine
Trace every error	Inaccuracies come from somewhere: find and fix the source
Repeat quarterly	Same prompts, same rubric, so progress is real