What Is AI Visibility? How to Know Whether ChatGPT Recommends Your Brand (Free Methods Included)

Table of Contents

1. A new question: when AI answers, are you in it?
2. Layer 1: manual question panels (most direct, anyone can do it)
3. Layer 2: AI referral traffic in analytics (the hard evidence)
4. Layer 3: automated AVS scoring (find the specific gaps)
5. Has this score been validated?
6. What to fix first: ordered by return on effort
7. Want it done for you?

#A new question: when AI answers, are you in it?

For twenty years brands asked "where do I rank on Google". Now there is a harder question: when a user asks ChatGPT which vendor to use, does it mention you?

This is not the same thing as search ranking. Rankings have Search Console; AI recommendations ship no official report. But it is not mysticism either. It can be measured systematically. Here is the three-layer method we actually use, with the validation data behind it.

#Layer 1: manual question panels (most direct, anyone can do it)

1. List 10 questions you want to be recommended for (say, "what are good Threads automation tools in Taiwan")
2. Ask ChatGPT, Perplexity, and Claude the same set monthly
3. Log: were you mentioned, which page was cited, in what position

Two disciplines: keep the questions fixed (changing them destroys the trend), and judge trends, not single runs (AI answers are stochastic; missing one round means nothing, missing three consecutive months is a signal).

#Layer 2: AI referral traffic in analytics (the hard evidence)

Open your analytics tool (GA4 or Plausible both work) and look for these referral domains: chatgpt.com, perplexity.ai, claude.ai, copilot.microsoft.com.

If they are there, AI engines are citing your content and sending users over. This is the most direct proof of AI visibility, because it is no longer "might be cited"; someone actually clicked through from an AI answer.

Our first-hand data point: after completing the AEO setup on Ultra Lab's own site, our August 2026 GA4 began showing referral traffic from claude.ai. Still small, but it went from zero to existing with no ad spend.

#Layer 3: automated AVS scoring (find the specific gaps)

Manual panels tell you the outcome, referral traffic tells you the result, but neither tells you what to fix. That is why we built AVS (AI Visibility Score): an automated scan of any URL producing a 0-to-100 score with itemized gaps.

To be clear upfront: AVS is UltraProbe's scoring framework, not an industry-wide standard. The current v2.0 weighs three dimensions:

AVS = SEO × 0.35 + AEO × 0.35 + AAO × 0.30

The SEO dimension covers traditional search health, AEO covers answer-engine readability (structured data, llms.txt, AI crawler access), and AAO covers AI agent accessibility. Scanning is free; drop in a URL and you get the report.

#Has this score been validated?

Yes, and we published the method and data. In April 2026 we ran an experiment: 155 real queries sent to AI search engines, 816 citations collected, 721 cited websites scanned for AVS (then a two-dimension SEO plus AEO version). Full write-up in the validation study. Two key findings:

1. About 60 percent of AI-cited sites scored B-grade (75) or above: 75 is the reference threshold for the AI answer pool
2. Recommendation-type queries showed a higher bar around 80: AI's willingness to cite and willingness to recommend are two different standards, and the latter is stricter

So if your site scans below 75, close that gap first. If you sit between 75 and 80 and your goal is active recommendation (queries like "which vendor should I hire"), keep pushing past 80.

#What to fix first: ordered by return on effort

1. Technical readability (one-time, do first): allow AI retrieval crawlers in robots.txt, add FAQPage and Organization structured data, publish llms.txt. Instructions in the AEO guide.
2. Content citability (ongoing discipline): Q&A structure, at least three concrete numbers per piece, sourced quotations. Principles and evidence in the GEO guide (the Princeton paper measured 30 to 40 percent visibility gains from these tactics).
3. Bilingual coverage: AI retrieval skews English; one piece in two languages holds a seat in both answer pools.

#Want it done for you?

Measurement costs nothing, so scan first: UltraProbe free scan. If you would rather not fix the gaps yourself, UltraGrowth exists to work through that list and keep it maintained (setup from NT$19,800, from NT$6,800/month). As always, read the SEO pricing and red-flags guide and compare deliverables before hiring anyone.

FAQ

What is AI visibility?

AI visibility is the degree to which your brand and website are read, cited, and recommended by AI engines (ChatGPT, Perplexity, Claude, Google AI Overviews). The search era measured rankings; the AI era measures whether you appear inside the answer when AI responds to a question. It is measurable through three layers used together: manual fixed-question panels, AI referral traffic in analytics, and automated scoring tools.

How can I check my website's AI visibility for free?

Three free methods: one, pick 10 questions you want to be recommended for and ask ChatGPT, Perplexity, and Claude monthly, logging cited sources; two, check your analytics referral sources for chatgpt.com, perplexity.ai, and claude.ai; three, scan your URL with UltraProbe (ultralab.tw/probe) for a free AVS score with itemized gaps across SEO, AEO, and AAO dimensions.

What AVS score counts as passing?

We validated the score against 155 real queries and 816 AI citations: about 60 percent of sites cited by AI scored B-grade (75) or above, and recommendation-type queries (asking AI which tool to use or whom to hire) showed a higher bar around 80. So 75 is the reference threshold for entering the AI answer pool, and being actively recommended requires pushing past 80.

If my AI visibility is low, what do I fix first?

By return on effort: first technical readability (allow AI crawlers in robots.txt, add FAQPage structured data, publish llms.txt), which is one-time engineering; then content citability (Q&A structure, concrete numbers, cited sources), which is ongoing discipline. A scan report lists which specific items your site is missing.