Some B2B buyers now use AI tools to compare vendors, understand a category, or check claims before speaking with sales. An AI visibility audit helps you see whether tools such as ChatGPT, Perplexity, Gemini, and Google AI Overviews can find and accurately describe your brand and whether they cite your pages.
This checklist gives you a repeatable way to establish a baseline. It is a practical review, not a guarantee of rankings or citations: AI answers can change by prompt, product, location, and time.
An SEO audit reviews how your pages can be crawled, indexed, and found in traditional search. An AI visibility audit tests how answer tools describe your company, which sources they cite, and whether your content is eligible to appear in their answers. The two reviews overlap, but they answer different questions.
A few useful terms:
For a broader overview, read PMG360’s complete guide to Answer Engine Optimization for B2B.
Some buyers use AI to summarize a topic, compare providers, or prepare questions for a sales conversation. Others start with search, reviews, or recommendations from colleagues. AI is one part of a wider research process, and its role varies by buyer and purchase.
For more on the measurement challenge, see PMG360’s article on why much of the buying journey is invisible to your CRM. The complete AEO guide also discusses how AI can appear in the B2B buyer journey.
AEO builds on SEO, so an AI visibility audit adds a layer of testing rather than replacing your SEO work.
|
What you compare |
Traditional SEO audit |
AI visibility audit |
|
What it measures |
Crawlability, indexation, rankings, links, and page health |
Brand mentions, description accuracy, citations, and observable referrals from AI tools |
|
Typical inputs |
Search Console, analytics, SEO platforms, and site crawlers |
A fixed prompt set, selected AI tools, analytics, and search performance data |
|
Useful indicators |
Indexed pages, search visibility, organic traffic, conversions |
Repeatable prompt-level visibility, accurate descriptions, citations, and qualified outcomes where trackable |
|
Review cadence |
Set a regular cadence based on site activity, often quarterly or twice a year |
Recheck a fixed prompt set monthly; repeat the full audit after major site changes |
You can establish a useful baseline with a spreadsheet, access to the AI tools your buyers use, and a few hours. Keep your test conditions consistent so you can compare results over time.
Write 15–25 prompts and keep the wording stable between audits. Include three types:
Problem-stage prompts can show whether buyers might encounter your brand before they know its name. They do not represent every buyer’s research path, so include all three types in your baseline.
Test the tools that are relevant to your audience. Examples include ChatGPT, Perplexity, Gemini, Claude, Microsoft Copilot, Google AI Overviews, and Google AI Mode. Chatbots and Google Search features are different experiences, so log them separately.
For each run, record the date, location or market, tool, prompt, whether your brand appeared, how it was described, cited sources, and any link to your site. Repeat prompts when practical. A single answer is only a snapshot, not a dependable ranking.
For each question, score Yes = 2, Partial = 1, or No = 0. With 25 questions, the maximum score is 50. These score bands are working guidelines for prioritizing your own audit; they are not industry benchmarks.
|
Score |
Working interpretation |
Suggested next step |
|
40–50 |
Stronger visibility signals in this audit |
Maintain the prompt log and look for gaps by topic, platform, and competitor |
|
25–39 |
Mixed visibility signals |
Find the lowest-scoring section and address its largest gaps |
|
Under 25 |
Foundational gaps may be limiting visibility |
Check technical access and brand accuracy before expanding content |
This section checks whether your brand appears for relevant prompts. For possible causes when it does not, read PMG360’s guide to why a brand may not appear in ChatGPT.
Ask each tool which providers fit your category and audience. Score Yes if your brand appears consistently across the tools and repeat runs you tested; score Partial if it appears only in some.
Test a set of commercial queries in Google Search. Record whether an AI feature appears and whether it names or links to your site. Search Console reports AI feature activity within the overall Web search performance data; it does not provide a separate AI Overview visibility score. See Google’s guidance on AI features and your website.
Record the sources shown with each answer. A link gives the buyer a path to verify details and may provide a measurable referral; an unlinked mention can still contribute to awareness.
Track competitor mentions using the same prompts, tools, and repeat-run method. This gives you a prompt-set share-of-voice indicator, not a universal measure of AI market share.
Compare questions about the problem with questions that include your company name. If you appear only in branded prompts, investigate whether your category and problem-solving content clearly explains your fit.
Being mentioned is only part of visibility. Check whether the information in the answer is correct and current.
Ask, “What does [company] do, and who is it for?” Compare the response with your approved positioning. Score Partial if the general description is right but important details are incomplete or incorrect.
Review your website, LinkedIn, review profiles, directories, and other authoritative sources. Note outdated names, taglines, service descriptions, or leadership information. Consistency helps people verify your information, though no single profile controls an AI answer.
Search your company name by itself and with your location or category. If the answer blends your company with another, make your site and public profiles more specific about your location, services, and identity.
Ask about current offerings and note any stale details. Trace each error to a page or profile that can be corrected. Some responses may reflect old sources or generated errors, so verify the source before deciding what to change.
Check that structured data matches visible page content. A validator can help identify syntax issues; use the Schema.org Markup Validator for general markup checks. Accurate structured data may help systems interpret information, but it does not guarantee inclusion or citations. Google says there are no special structured-data requirements for AI Overviews or AI Mode.
A page that cannot be accessed or indexed may be harder to surface. Review crawler settings carefully because different bots can have different purposes.
Review your robots.txt and any applicable shared policy signals. OpenAI distinguishes OAI-SearchBot for ChatGPT search from GPTBot for potential model training; ChatGPT-User handles certain user-initiated visits and is not an automatic web crawler. Perplexity documents PerplexityBot and separate user-directed fetching. Anthropic distinguishes ClaudeBot, Claude-SearchBot, and Claude-User. Google’s Googlebot controls Search crawling, including eligibility for AI Overviews; Google-Extended is a robots.txt token for some Gemini training and grounding uses, not a separate crawler or a control for AI Overviews.
Read the current OpenAI crawler documentation, Perplexity crawler documentation, Anthropic bot guidance, and Google’s pages on Google-Extended and AI features in Search. Policies change, so verify current documentation before editing directives. Google has also announced plans to support a shared AI training control; check its adoption and guidance before relying on it.
A robots.txt rule is only one part of access. Ask your technical team to review relevant logs and security rules for blocked requests. Avoid relying on user-agent names alone when validating important access; follow each provider’s current verification guidance.
Check important pages in a browser and review their rendered output. Google can render JavaScript, but crawler behavior differs by provider. If essential answers appear only after client-side scripts or interaction, ask your technical team to verify they are available to the crawlers you care about.
Review Google Search Console and Bing Webmaster Tools for indexing issues. Bing is relevant to Microsoft search experiences, including Copilot features that use web search. Do not assume that every AI assistant uses the same index.
Validate the schema types you actually use, such as Organization, Article, or BreadcrumbList. Fix errors and keep markup consistent with page content. Structured data can clarify facts for systems, but Google does not require special schema to appear in AI Overviews or AI Mode, and markup alone will not secure an AI citation.
Clear structure can help readers and automated systems locate key information. There is no guaranteed word count, heading formula, or format that ensures an AI citation.
Review the first section of each key page. Make sure it answers the page’s main question directly, then provide the explanation and evidence. There is no required 40–60-word answer length.
Compare headings with sales-call notes, support questions, and search queries. Use question phrasing where it improves clarity; do not force every heading into a question.
Check whether your pages use structure to clarify comparisons or instructions. Tables and lists should serve the reader, not be added solely in hopes of influencing AI results.
Look for original research, customer examples, practical frameworks, or expert perspective that you can support. Original material can make a page more useful and distinctive, but it does not guarantee citations.
Provide author names and relevant credentials when appropriate, and update pages when facts change. These practices help readers evaluate the content; a date or author bio by itself is not a ranking or citation guarantee.
AI responses may cite sources beyond your website, and selection varies by tool and prompt. For a deeper explanation, see the source-selection section in PMG360’s AEO guide.
Run your category prompts and record the sources shown. Check whether your company is represented accurately on relevant review sites, industry publications, or directories. Treat the list as a research and communications priority—not as a promise that outreach will earn a citation.
Compare review presence and recency on platforms buyers in your market use. Address factual issues and invite honest customer feedback through appropriate processes. Do not assume review volume or recency directly determines AI visibility.
Review the GA4 Traffic acquisition report for sources such as chatgpt.com, perplexity.ai, or other AI-related domains. The exact source can vary, and some visits may not preserve referral information. A custom channel group can make observed AI traffic easier to review; it will not capture every AI-assisted journey. See Google’s GA4 Traffic acquisition guidance.
Keep a dated log of prompts, tools, and relevant test conditions, then compare trends over time. Without a consistent method, results from different prompts or sessions may not be comparable.
Where source information is available and your consent practices allow it, preserve it in analytics and CRM fields. Review lead quality and downstream outcomes alongside referral counts. Attribution may be incomplete, so distinguish observed source data from self-reported or inferred influence. PMG360’s CRM services cover campaign tracking, source attribution, and reporting setup.
Add your 25 scores for a total of up to 50. Use the bands as a prioritization aid, not as a grade or industry benchmark. Also compare the five section totals so you can identify whether your biggest gaps involve mentions, accuracy, access, content, or measurement.
Keep the prompt set running and watch for changes after major site updates or shifts in the tools you test. Expand the audit to additional buyer questions and relevant competitors.
Start with the lowest-scoring section. For example, accuracy issues may call for profile cleanup, while content gaps may call for clearer pages that answer buyer questions.
Review whether priority pages can be accessed and indexed, then check whether your brand information is accurate and consistent. Add or revise content after those foundational checks.
An audit is useful when it leads to a clear next step. PMG360’s AEO and AI visibility services include prompt testing, technical and content reviews, and reporting on citations, AI referrals, and leads. Scope and results depend on your audience, site, and campaign goals.
For CRM measurement, PMG360’s CRM services cover source attribution and reporting setup. Connecting the systems can help teams evaluate tracked referrals and lead outcomes, while recognizing that some AI-assisted research will remain unobserved.
Run the full checklist once to establish a baseline. Recheck a stable prompt set monthly and repeat the full review after a major site change or when your buyer research changes. Adjust the cadence if your own results show that answers shift more or less often.
Yes. A spreadsheet, access to the AI tools you want to test, Google Search Console, Bing Webmaster Tools, and GA4 can support a basic audit. Paid tools may help automate testing or reporting at scale.
Answers can vary by prompt, tool, session, location, and time. Record your test conditions and repeat the same prompts. Treat the results as patterns to investigate, not a definitive ranking from one response.
Google says Google-Extended does not control inclusion in Google Search or AI Overviews. Googlebot and Search controls govern Search crawling; Google-Extended applies to certain Gemini training and grounding uses. Check Google’s current crawler documentation before changing rules.
An SEO audit reviews crawlability, indexing, rankings, and site health. An AEO or AI visibility audit tests how selected AI tools mention and describe your brand, what they cite, and what referral activity can be observed. The work overlaps, and neither audit replaces the other.
There is no universal 50-point industry standard. In this checklist, 40–50, 25–39, and under 25 are working bands for prioritizing your own findings. Compare results using the same prompts and conditions over time.
Start with tools your buyers are likely to use and features relevant to your market. ChatGPT, Perplexity, Gemini, and Google AI Overviews are common starting points. Add other tools as needed, and report search features separately from chatbot answers.
It can be useful to track, but interpret it alongside lead quality, conversions, and other channels. Small or incomplete referral counts alone do not establish the value of AI visibility.
That is a business decision and may differ by crawler. Search, training, and user-directed bots can have separate controls. Review each provider’s current documentation and consult your technical team before changing access rules.
Timing varies by the type of change, the platform, and when its systems revisit the source. Monitor the same prompt set over time; do not promise a fixed timeline or treat one changed response as proof of impact.
Use this checklist to establish a repeatable baseline, identify the sections that need attention, and track observable changes over time. If you want help testing prompts and prioritizing fixes, request a free AI visibility audit from PMG360.