An AI visibility audit checks whether ChatGPT, Claude, Perplexity and Google AI Overviews can reach, understand and recommend your brand. This template gives you 24 checks in six areas, a 0 to 2 score for each, and a total out of 48, so you can record where you stand and rerun it each quarter.
Key takeaways🔗
- The template covers six areas: access, indexing, entity, content, third-party proof and measurement.
-
Score every check 0 (fail), 1 (partial) or 2 (pass). The total out of 48 is a progress number for your own site, not an industry benchmark.
-
Access and indexing come first. A low score there caps everything else, because assistants cannot cite pages they cannot read.
-
Our GEO audit checklist explains the 10 core checks and why each one matters. This post is the worksheet you fill in.
-
Rerun the same template every 90 days and compare section scores, not just the total.
How is this template different from a GEO audit checklist?🔗
This template is a fill-in worksheet with a score for every line, while our GEO audit checklist is an explainer of 10 pass or fail checks. Use the checklist to understand what to look for. Use this template to record findings, assign owners and track a score over time.
The differences in short:
| GEO audit checklist | This audit template | |
|---|---|---|
| Format | 10 checks explained in prose | 24 checks in fill-in tables |
| Scoring | Pass or fail, 10 points | 0, 1 or 2 per check, 48 points |
| Best for | Learning what matters | Running and repeating the audit |
| Output | A list of gaps | A scored worksheet with owners and next steps |
| Time | 2 to 3 hours by hand | Longer the first time, faster on reruns |
How do you use this template?🔗
You use this template by copying the six tables below into a spreadsheet or doc, working through one section at a time, and writing a score and a short note on every line. Do not skip lines you think you pass. The note is what makes the next audit faster.
Set up before you start:
-
Pick your pages. List 5 to 10 buyer pages: home, pricing, top product or service pages, and any comparison pages.
-
Pick your prompts. Write 10 to 25 questions a buyer would ask an assistant, such as "best payroll software for a 20-person restaurant." Mix category, comparison and problem questions.
-
Pick 2 to 5 competitors. Use the brands buyers actually compare you with.
-
Use this scale everywhere: 0 = fail or unknown, 1 = partly done, 2 = done and verified.
Section 1: Can AI crawlers access your site?🔗
AI crawler access is the first section because every later check depends on it. If a search crawler is blocked by robots.txt or a firewall, that assistant cannot read your pages, no matter how good they are.
| # | Check | How to test | Score (0-2) | Notes |
|---|---|---|---|---|
| 1.1 | Search crawlers allowed in robots.txt | Confirm OAI-SearchBot, Claude-SearchBot, PerplexityBot, Googlebot and Bingbot are not disallowed on key pages | ||
| 1.2 | Training crawler policy is a deliberate choice | Check GPTBot, ClaudeBot, Google-Extended and Applebot-Extended rules match what you decided | ||
| 1.3 | CDN and firewall do not block AI bots | Review AI bot settings in Cloudflare or your host | ||
| 1.4 | Server logs show AI crawler visits | Filter the last 30 days by user agent |
What the docs say: OpenAI states that sites opted out of OAI-SearchBot "will not be shown in ChatGPT search answers," while GPTBot is a separate training setting (OpenAI). Anthropic says blocking Claude-SearchBot "may reduce your site's visibility and accuracy in user search results" (Anthropic). Perplexity asks sites to allow PerplexityBot to appear in its search results (Perplexity).
Section 2: Are your key pages indexed and readable?🔗
Indexing is the second section because AI answers often draw on search indexes. Google says AI Overviews and AI Mode have "no additional requirements" beyond normal Search eligibility, which means being indexed and eligible for a snippet (Google).
| # | Check | How to test | Score (0-2) | Notes |
|---|---|---|---|---|
| 2.1 | Key pages indexed in Google | URL Inspection in Google Search Console | ||
| 2.2 | Key pages indexed in Bing | URL Inspection in Bing Webmaster Tools | ||
| 2.3 | Key text is in the raw HTML | View source (Ctrl+U) and search for a price or product name | ||
| 2.4 | No stray noindex, nosnippet or canonical errors | View source on each key page |
Check 2.3 matters more than most teams expect. A December 2024 Vercel and MERJ study found that OpenAI's and Anthropic's crawlers download JavaScript files but do not run them (Vercel), so text that loads client-side may be invisible to them. Bing setup is covered step by step in Bing SEO for ChatGPT.
Section 3: Do AI systems understand who you are?🔗
The entity section checks whether your brand is described the same way everywhere, so an assistant can connect your name to what you sell, who you serve and where. Mixed descriptions make it harder for a model to place you in the right category.
| # | Check | How to test | Score (0-2) | Notes |
|---|---|---|---|---|
| 3.1 | One clear sentence on what you do | Read your homepage hero and About page | ||
| 3.2 | Same name, category and description across profiles | Compare your site, LinkedIn, Google Business Profile, Crunchbase or G2 | ||
| 3.3 | Organization schema with sameAs links | View source or a schema validator | ||
| 3.4 | Assistants describe you correctly | Ask ChatGPT, Claude and Perplexity "What is [brand]?" |
For 3.4, write down each answer word for word. Wrong pricing, an old product name or a mix-up with a similarly named company all count as a 0 or 1. Our schema markup guide covers Organization and the other types worth adding.
Section 4: Does your content answer buyer questions?🔗
The content section checks whether your pages answer the questions in your prompt list in plain, quotable passages. Assistants lift single sections, so each one needs to make sense on its own.
| # | Check | How to test | Score (0-2) | Notes |
|---|---|---|---|---|
| 4.1 | A page exists for each core buyer question | Map your prompt list to URLs | ||
| 4.2 | Answers come first in each section | Read the first sentence under each H2 | ||
| 4.3 | Comparison pages for top competitors | One page per main rival, fair and specific | ||
| 4.4 | Prices, plans and specifics stated in text | Search key pages for numbers | ||
| 4.5 | Pages updated in the last 12 months | Check visible "last updated" dates |
Microsoft's guidance for AI answers makes the same point: clear headings, direct question and answer pairs, and key facts kept in HTML text rather than only in images, plus a warning not to hide answers in tabs because "AI systems may not render hidden content" (Microsoft Advertising).
Section 5: What do other sites say about you?🔗
The third-party section checks whether independent sources mention your brand, because assistants often cite review sites, lists and forums rather than the brand's own pages.
| # | Check | How to test | Score (0-2) | Notes |
|---|---|---|---|---|
| 5.1 | Listed on the main review sites for your category | Search G2, Capterra, Yelp, Clutch or your trade directories | ||
| 5.2 | Included in "best of" lists that assistants cite | Note which list pages appear as sources in your prompt answers | ||
| 5.3 | Real mentions in Reddit or forum threads | Search Reddit for your category and brand | ||
| 5.4 | Recent press or trade coverage | Search news for your brand in the last 12 months |
For 5.2, the source links in Perplexity and ChatGPT answers to your prompts are the best guide to which third-party pages matter in your category. Our post on Reddit citations in AI answers explains how to take part in threads without spamming.
Section 6: How are you measuring AI visibility?🔗
The measurement section checks whether you can see change over time. Without a baseline, you cannot tell if the fixes from sections 1 to 5 worked.
| # | Check | How to test | Score (0-2) | Notes |
|---|---|---|---|---|
| 6.1 | Baseline of brand mentions per assistant | Run your prompt list and record who is named | ||
| 6.2 | Competitor share recorded for the same prompts | Count mentions per competitor | ||
| 6.3 | AI referral traffic tracked in analytics | A GA4 segment for AI referrers | ||
| 6.4 | Bing AI Performance report checked | Bing Webmaster Tools, AI Performance | ||
| 6.5 | Repeat schedule set | Calendar the next run |
Microsoft's AI Performance report, in public preview since February 10, 2026, shows citations of your pages in Microsoft Copilot, AI summaries in Bing and "select partner integrations" (Microsoft Bing). Microsoft does not name those partners, so do not treat it as a ChatGPT, Claude or Perplexity report. Check 6.1 still needs its own run. For 6.3, see how to measure AI search traffic in GA4.
How do you read your total score?🔗
You read your score by section first and total second. A site with 40 out of 48 but a 2 out of 8 on access has a bigger problem than a site with 30 that passes access cleanly.
These bands are our own rule of thumb for prioritizing, not a published benchmark:
| Total (out of 48) | What it suggests | Where to start |
|---|---|---|
| 40 to 48 | Strong foundation | Expand prompts and third-party coverage |
| 28 to 39 | Solid base with clear gaps | Fix the lowest-scoring section |
| 15 to 27 | Several areas missing | Access, indexing and entity first |
| 0 to 14 | Assistants likely cannot find or place you | Sections 1 and 2 this week |
Then turn each 0 or 1 into a task with an owner and a date. Sort by section order, since a fix in section 1 can lift results in every later section.
How often should you rerun the audit?🔗
You should rerun the full audit every 90 days and rerun sections 1 and 2 after any site migration, redesign or CDN change. AI answers shift as models and sources change, so a single audit goes stale quickly.
Section 6.1 is the part worth automating. TrendlyAI runs your buyer prompts weekly across ChatGPT, Perplexity, Claude and Google AI Overviews, records which brands get named, and drafts fixes for the gaps. It does not track Gemini, Grok, Copilot or AI Mode, and it takes 1 sample per prompt per engine, so treat it as a trend line. I track TrendlyAI's own domain this way, and our score is still low, which is exactly why I keep rerunning audits like this one.
For a quick first data point for section 6, run the free TrendlyAI check: 5 prompts on ChatGPT and Perplexity, about 30 seconds, no signup. For everything else in AI search, start at our GEO hub.
Frequently asked questions🔗
What is an AI visibility audit?🔗
An AI visibility audit is a structured review of whether AI assistants such as ChatGPT, Claude, Perplexity and Google AI Overviews can reach your site, understand your brand and recommend you for buyer questions. It covers technical access, content, third-party mentions and measurement.
How long does an AI visibility audit take?🔗
It depends on how many pages and prompts you include. Expect the first run to take longer than a quick checklist, since you test and note every line. Reruns are faster because your notes show what was already fixed.
What is a good AI visibility audit score?🔗
There is no industry benchmark for this score. It is a way to compare your site with itself over time. Look for each section to move toward 2s, starting with access and indexing.
Should agencies use this template for clients?🔗
Yes. The scored sections make it easy to show a client where they started and what changed after each quarter. Keep the same prompt list between runs so the comparison is fair.
Can I run this audit without paid tools?🔗
Yes. Every check can be done with free tools: robots.txt, Google Search Console, Bing Webmaster Tools, GA4 and the assistants themselves. Paid tools mainly save time on running prompts and tracking results each week.