The AI visibility metrics worth tracking are mention rate, citation rate, AI share of voice, answer position, sourced sentiment, and cited source coverage, measured per platform and per topic across a fixed prompt set. Skip raw mention counts, one-off screenshots, and domain authority as a stand-in. They fill a report and tell you very little.
Most teams would begin by typing their own brand search into the ChatGPT search bar, take a screenshot, and call it a day. Meanwhile, according to Pew Research Center, “when an AI-generated summary appeared on a search results page, Google users clicked on a traditional search result 8% of the time, compared with 15% when no summary appeared.” If the answer swallows your click, your analytics can no longer tell you where you are doing well and where you should pivot; you have to invest in metrics that focus on the answer itself.
📘 Definition
AI visibility metrics
AI visibility metrics measure how often, prominently, and favorably particular AI platforms (e.g., ChatGPT, Gemini, Perplexity, Google AI Overviews, Google AI Mode, etc.) are mentioned or cited in relation to a particular brand, for a certain set of prompts, relative to named competitors.
Which Metrics Measure AI Visibility?
Six indicators perform the heavy lifting. The mention rate and citation rate indicate your presence. The share of voice and position reflect your position against the competition. The sourced sentiment and cited source coverage represent the information from AI about you and the sources that affect you.
The six AI visibility metrics that earn a place on your report
| Metric | What it answers | How to calculate | Watch out for |
|---|---|---|---|
| Mention rate | Does the AI name us at all? | Answers naming your brand ÷ prompts run | Branded prompts inflate it |
| Citation rate | Does the AI point people to our pages? | Answers citing your domain ÷ prompts run | Split explicit citations from implicit ones |
| AI share of voice | How do we compare on the same prompts? | Your citations ÷ citations for all tracked brands | Only as good as your competitor list |
| Position | Are we the first pick or a footnote? | Average order your brand appears in the answer | Lists reorder between runs, so read the trend |
| Sourced sentiment | What story is the AI telling about us? | Positive, neutral and negative mentions, traced to the cited URL | A score with no URL behind it can’t be fixed |
| Cited source coverage | Are we on the pages AI trusts in our category? | Cited URLs that include you ÷ all cited URLs for your prompts | This is where your outreach list comes from |
Why citation rate is the one to watch
Mentions tell you the model has heard of you. Citations tell you it trusts your page enough to send a reader there, which carries authority and is the only clicks AI answers still hand out.
“If your boss asks for one number, give them citation rate.”
Why sentiment needs a source
Shadow reputation
Every brand has one in AI that’s assembled from reviews, forums, and comparison pages that you didn’t write. A sentiment score without the urls behind it can tell you there’s a leak, but not which pipe it’s coming from.
Which AI Visibility Metrics Should You Ignore?
Ignore anything you can’t trace back to a prompt, a platform, and a competitor. That rules out most quick audits.
❌ Single prompt screenshots
Run the same prompt twice, and you may get a different list.
❌ Raw mention counts
“We got 340 mentions” means nothing without the number of prompts behind it.
❌ “We rank #1 in ChatGPT”
There is no fixed ranking. Order shifts with wording, location, and session.
❌ Domain authority as a stand-in
In the Wellows study of 20.5 million AI citations, 47.9% went to domains under DA 40. Authority helps at the very top (301 domains at DA 90 or above took 17.0%), but most citations land on smaller sites.
❌ Daily swings on a tiny prompt set
With 20 prompts, one changed answer moves your rate by 5 points. Read weekly trends.
❌ AI referral traffic as the full picture
Pew found users clicked a link inside the AI summary in only 1% of visits. Referrals are real but badly undercount presence.
💡 Did you know?
In a Wellows analysis of 2.37 million AI citations, 77% of cited sources had a domain authority under 50, and AI engines pulled from 48,560 unique domains. (Source: Wellows)
How Do I Measure My Brand’s Visibility in ChatGPT and Gemini?
Build a prompt library of questions your buyers actually ask, and run it on a periodic basis across each channel, capturing four things for each answer: whether you were named, mentioned, cited, and where (what URLs). Then, compare the answers to your most direct competitors on each run.
Write the prompt set
Mix category, problem, comparison, and branded prompts, with branded kept small so they don’t flatter you. Aim for at least 50 so one answer can’t swing the numbers.
Tag every prompt by topic
Wellows found visibility can vary by more than 4.5x across topics for one brand.
Run each platform separately
ChatGPT, Gemini, Perplexity, AI Overviews, and AI Mode draw on different sources. Wellows found ChatGPT cited branded domains 11.1 points more than Google. The median DA of cited sites was 34 on Perplexity and 48 on AI Mode.
Log explicit and implicit citations
Explicit means your brand is named. Implicit means your domain sits inside a page the AI cited.
Freeze your competitor list
Change it mid-quarter and share of voice stops meaning anything.
Set a cadence
Collect daily, review weekly, report monthly.
Turn gaps into tasks
Cited URLs that leave you out become outreach targets. Prompts where no strong source is cited become content briefs.
What KPIs Matter for AI Search?
Create a report in three levels. The presence (mention rate, citation rate), competitive standing (share of voice, position, sentiment), and business impact (AI referral sessions, conversions from them, branded search trend). Leadership views the third level. Your team analyzes the first two.
Here’s how the old scorecard maps across:
Then vs Now: classic SEO KPIs and their AI search equivalents
| Then: classic SEO KPI | Now: AI search KPI | Why it changed |
|---|---|---|
| Keyword rankings | Position and AI share of voice | Answers have no fixed list of ten links |
| Impressions | Mention rate across prompts | Visibility happens inside the answer |
| Backlink count | Cited source coverage | The pages AI cites shape what it says about you |
| Click-through rate | Citation rate plus AI referral sessions | Fewer clicks, and citations are the path that remains |
For benchmarks and targets on each one, see our guide to generative engine optimization KPIs.
Brand Performance Metrics in AI Search: A Monday Checklist
Copy this into your workflow
Tick these off in order. The setup list takes one afternoon; the rest is routine. Your ticks are saved in this browser.
0 of 14 done
🎉 All 14 done. Your AI visibility tracking is set up and running.
How Wellows Tracks These Metrics
You can run all of this in a spreadsheet for a while. It gets painful around the third platform and second competitor. Wellows tracks each metric in this guide daily:
| Metric | Where it lives in Wellows |
|---|---|
| Mention and citation rate | Explicit and implicit citation tracking across ChatGPT, Gemini, Perplexity, AI Overviews and AI Mode |
| AI share of voice | AI Visibility Score, a 0 to 100 score measured against your tracked competitors |
| Sourced sentiment | Sentiment read from the cited pages themselves, down to the URL |
| Position and movement | Daily citation movement: earned, lost, held or reworded |
| Cited source coverage | An outreach list of cited URLs that leave you out, with a verified contact |
First, it differentiates between being named and being cited, so that you can see if the AI tools are actually driving traffic to you. Second, it reveals your shadow reputation, or the specific pages by which ChatGPT and Perplexity are learning about you, compared to your competitors.
It also closes the loop on the content you create – those gaps in knowledge can be filled by your existing content, or they can inspire new content creation, all while being compared to what you already publish. You can start with Wellows Monitoring, or perhaps examine how the AI Visibility Score is calculated.
Conclusion
Good AI visibility measurement is mostly discipline: a fixed prompt set, five platforms tracked separately, a frozen competitor list, and sentiment you can trace to a URL.
This week, write your prompt set and run a first baseline. Then pick three cited URLs that leave you out and start your outreach there. If you'd rather not build it by hand, set up tracking in Wellows and have your first baseline tomorrow.
FAQs
