For a long time, measuring SEO came down to two numbers: where you ranked, and how many clicks that ranking earned. Both still matter, and I am not going to pretend they stopped. But a growing share of what people decide about a brand now happens inside an AI answer, before any click, on a surface you do not control and mostly cannot see in your analytics.
That is the gap these metrics try to close. AI search visibility metrics measure whether ChatGPT, Google’s AI Overviews, Perplexity, and Gemini mention you, cite you, and how they describe you when they do. Where you rank on a page of links is a different question, and the two overlap less than most people assume once you go looking at the numbers.
Most teams cannot read that scoreboard yet. Semrush surveyed marketing leaders for its 2026 AI Visibility Index and found 45% could not accurately measure their brand inside AI-generated answers, and only 9% had the tools to track every relevant metric across platforms.1 What follows is a plain walk through the metrics that count, what each one actually tells you, and which ones I would spend time on versus which are mostly vendor noise. It supports my wider SEO Trends 2026 report, so start there if you want the full year in context.
Why it is a separate scoreboard from SEO
The tempting assumption is that AI visibility is just your rankings under a new name. Rank well, get cited, carry on with SEO as usual. It holds up on Google’s own surface and falls apart nearly everywhere else, and the numbers below show the gap.
Ahrefs ran the check both ways. When they looked at what standalone assistants cite, only about 8% of the pages ChatGPT, Gemini, and Copilot named also ranked in Google’s top 10 for the same query, and roughly 80% did not rank in the top 100 at all.2 Perplexity landed higher at 28.6%, which fits, since it leans harder on live search. Then they looked at Google’s own AI Overviews and got a very different answer: 76.1% of cited pages ranked in the top 10, with a median cited position of three.3
The split makes sense once you know how these systems fetch. Google’s AI Overviews are grounded in Google’s own index, so they cite what already ranks. ChatGPT and the others run their own retrieval, rewrite your question into several searches, and pull passages from wherever they find them. I went through the mechanics in what RAG is, because it explains a lot of what looks random here.
So you can sit at position one and never appear in ChatGPT, and you can get quoted by ChatGPT while ranking nowhere near the top. On Google’s own surface the two move together. Off it they mostly don’t, and a rank tracker will tell you nothing about the second outcome.
The metrics worth tracking
There are five that keep coming up. I have ordered them by how much I trust the data behind each, most solid first, and I will be straight about the two at the bottom where the numbers thin out.
AI referral traffic and conversion
This is the one you can already see in your analytics, which makes it the natural place to start. Filter your GA4 sessions by referrer and you can watch visits arrive from chatgpt.com, perplexity.ai, gemini.google.com, and the rest. That traffic is attributable, and it is climbing quickly. Similarweb estimated AI platforms sent 1.13 billion referral visits in June 2025, up 357% from a year earlier, and total AI referrals grew more than threefold between September 2024 and September 2025.45
That growth comes with a heavy concentration problem, though: almost all of the traffic arrives from one product.
- ChatGPT 84 (84%)
- Perplexity 8.9 (9%)
- Gemini 4.5 (4%)
- Copilot 2.1 (2%)
- Claude 0.6 (1%)
Previsible tracked 6.77 million assistant-driven sessions across 166 properties and found ChatGPT accounted for 92.4% of them on average.6 When someone shows you a tidy pie of “AI traffic sources,” remember most of it is one product, and your ChatGPT presence is doing nearly all of the work.
The louder claim about this traffic is that it converts better than organic. The studies that measured it disagree enough that I would rather show you three of them side by side than pick the flattering one.
| Study | What it found | The catch |
|---|---|---|
| Visibility Labs, 94 ecommerce brands, 2025 | ChatGPT visitors converted at 1.81% vs 1.39% for non-branded organic, about 31% higher | ChatGPT drove $474K in revenue against $32.1M from organic, roughly 1.5% of the total7 |
| Ahrefs, its own site | AI visitors converted 23x higher than organic; 0.5% of traffic drove 12.1% of signups | One site, signups as the goal, so read it as a case study, not a benchmark8 |
| Amsive, 54 sites | LLM traffic converted at 4.87% vs 4.60% for organic | The gap was not statistically significant (p = 0.794), so probably noise9 |
The visitors AI sends are often further along and worth more per session, though the volume stays small for most sites. Anyone quoting you a clean 23x with a straight face is selling something. Measure your own AI referral conversion rate; the multiples from other people’s sites will not transfer to yours.
Citation frequency
Referral traffic only counts the people who clicked through. Plenty of AI visibility never produces a click at all, and citation frequency is how you count that part: how often you show up as a named source inside the answer, visit or no visit. For a lot of queries this is closer to the real point, since the mention itself does the work.
One quirk to build into how you measure it. The platform changes the math. Semrush found ChatGPT cites an average of 15 sources per response while Gemini cites about 3.1 Your odds of being one of the named sources are wildly different across engines, so a single blended “citation rate” averages away the thing you actually need, which is where you stand on each one.
Citation position
Frequency tells you how often you get named. Position tells you whether it counts for anything. Where you land in the answer, near the top or buried at the end, decides whether the mention gets read at all, the same way rank three and rank thirty were never worth the same on a page of links. I dug into the retrieval data on this in the RAG piece, where the short version is that the front of a page and the front of an answer get read far more closely than the rest.
AI Share of Voice
Share of Voice is the percentage of a tracked set of prompts where your brand shows up, measured against your competitors. Pick 50 or 200 questions your buyers actually ask, run them across the assistants on a schedule, and count how often you appear versus the other names in your category. As a directional, competitive gauge it is genuinely useful. You learn where you are invisible and who owns the answer instead of you.
The tooling oversells it, though. There is no credible industry benchmark for what a “good” Share of Voice looks like. Vendors will tell you to aim for 30%, or for category parity, and those numbers are invented. What you can trust is your own trend line as it moves while you publish and earn mentions. Any absolute target that shows up without a source behind it, I would ignore.
Brand sentiment in LLMs
Sentiment covers how the mention frames you. When ChatGPT summarizes your category, you might come out as the trusted default or as the example of what to steer clear of, and that framing is worth spot-checking by hand on your key prompts.
I would keep expectations low on the automated version, because this is the thinnest area for real data. There is no benchmark and no agreed scoring method, and the vendor tools that hand you a sentiment number are usually running a generic model over the answer and reporting back whatever it says. Reading the actual responses the assistants give about your brand, on the prompts you care about, will tell you more than any of those gauges do right now.
What actually moves it
Measuring only helps if you know which lever to pull. The one with the clearest evidence behind it is getting mentioned across the web. Ahrefs looked at 75,000 brands and asked what correlates with showing up in AI Overviews, and branded web mentions came out on top, correlating about three times more strongly than raw backlinks.10
Correlation is not a switch you flip, so treat this as a pointer at the mechanism rather than a checklist item. Getting talked about across the web, in the places these models read, tends to travel with getting cited by them. It lines up with the on-page retrieval work too, the clear and quotable answers near the top of pages a crawler can actually load. If you want a starting point, I built a free AI content visibility checker to see whether the assistants are pulling you in on the queries you care about, and AI Overviews vs classic SEO covers the formatting side.
Keep the whole thing in proportion
One more number, because it keeps the rest honest. In June 2025, AI platforms sent 1.13 billion referral visits while Google search sent 191 billion in the same window.4 AI referrals are growing fast and, for most sites, still tiny beside classic search.
So do not torch your SEO program to chase a channel that is 0.6% the size of the one you already have. Track AI visibility now for a different reason: the influence is running ahead of the traffic. People form opinions inside these answers and either arrive already decided or never arrive, because the answer was enough on its own. On Google, the top organic result loses about a third of its clicks, 34.5%, once an AI Overview sits above it.11 None of that registers as a session, so you only catch it if you go looking. It is cheaper to build that habit now, while the numbers are small, than to backfill it after the channel matters.
Frequently asked questions
What are AI search visibility metrics?
They are the KPIs that measure how your brand shows up inside AI answers rather than on a page of links. The main ones are AI Share of Voice, citation frequency, citation position, brand sentiment in LLMs, and AI referral traffic and its conversion rate. Together they tell you whether ChatGPT, Google's AI Overviews, Perplexity, and Gemini mention, cite, and recommend you.
How is AI visibility different from SEO rankings?
Ranking is where your page sits in a list of links. AI visibility is whether an AI answer names you as a source. On Google's AI Overviews the two overlap, since about 76% of cited pages rank in the top 10. But for standalone assistants like ChatGPT, only around 8% of cited pages rank in Google's top 10, so you can rank well and still be invisible in the chat, or get cited without ranking at all.
Does traffic from AI search convert better than organic?
Sometimes, but the evidence is mixed and you should measure your own. Visibility Labs found ChatGPT ecommerce traffic converted about 31% higher than non-branded organic, and Ahrefs saw 23x on its own site, but that is a single site. Amsive analyzed 54 sites and found the difference was not statistically significant. AI visitors are often further along, but the volume is small for most brands.
What is AI Share of Voice?
It is the percentage of a tracked set of prompts where your brand appears, measured against competitors. You choose the questions your buyers ask, run them across the assistants on a schedule, and count how often you show up versus rival brands. It is useful as a competitive trend line. There is no credible industry benchmark for a good score, so watch your own movement rather than a vendor's target.
How do I measure AI referral traffic?
Filter your analytics by referral source and look for sessions from domains like chatgpt.com, perplexity.ai, and gemini.google.com. It is real and attributable, though most of it comes from ChatGPT, which accounted for about 92% of standalone AI-assistant referrals in one large study. Track the conversion rate of that traffic on your own site rather than trusting a headline multiple from elsewhere.
Is it worth tracking AI visibility if the traffic is still small?
Yes, because the influence runs ahead of the traffic. AI referrals are still a fraction of classic search, roughly 0.6% of Google's referral volume in mid-2025, but people form opinions inside AI answers before they ever click, and often without clicking. That pre-click influence does not register as a session. Building the measurement while the channel is small means you understand it before it gets large.
Keep going
More on how AI answers pick their sources, and what it means for what you measure.
Sources
-
Semrush, 2026 AI Visibility Index, analyzing 126 million AI search prompts. Source for 45% of marketing leaders being unable to accurately measure their brand in AI answers, 9% tracking all relevant metrics, and ChatGPT citing an average of 15 sources per response versus Gemini’s 3. Prompts analyzed January through April 2026. ↩ ↩2
-
Ahrefs, AI Search Citations vs. Google Rankings, 15,000 long-tail queries, July 2025. Across ChatGPT, Gemini, and Copilot, about 8% of cited pages ranked in Google’s top 10 and roughly 80% did not rank in the top 100; Perplexity cited a top-10 page 28.6% of the time. ↩
-
Ahrefs, How Google Search Rankings Correlate With AI Overview Citations, 1.9M citations from 1M AI Overviews, July 2025. 76.10% of AI Overview-cited pages ranked in the top 10, with a median cited position of 3. ↩
-
Similarweb, AI Referral Traffic Winners By Industry. AI platforms generated over 1.13 billion referral visits in June 2025, up 357% from June 2024, against 191 billion referrals from Google search in the same month. ↩ ↩2
-
Similarweb, Generative AI Stats 2026. Total AI referral visits across the web grew more than 3x between September 2024 and September 2025. ↩
-
Previsible, AI Traffic Report, 6.77 million standalone-LLM sessions across 166 GA4 properties, November 2024 to May 2026. ChatGPT averaged 92.4% of standalone AI-assistant referral traffic across the window. The December 2025 snapshot: ChatGPT 84%, Perplexity 8.9%, Gemini 4.5%, Copilot 2.1%, Claude 0.6%. ↩
-
Visibility Labs, reported by Search Engine Land, ChatGPT vs. non-branded organic search conversions. Across 94 ecommerce brands in 2025, ChatGPT traffic converted at 1.81% versus 1.39% for non-branded organic, and drove $474,000 in revenue against $32.1 million from organic. ↩
-
Ahrefs, AI Search Traffic Converts 23x Better. On Ahrefs.com, AI search accounted for 0.5% of traffic but 12.1% of signups over 30 days, a 23x higher conversion rate than organic. Single site, signups as the conversion event. ↩
-
Amsive, Does LLM Traffic Convert Better Than Organic?. Across 54 sites, LLM traffic converted at 4.87% versus 4.60% for organic, a difference that was not statistically significant (p = 0.794). ↩
-
Ahrefs, What Correlates With AI Overview Brand Mentions, 75,000 brands. Branded web mentions correlated with AI Overview visibility at 0.664, against 0.326 for Domain Rating, 0.295 for referring domains, and 0.218 for backlinks. ↩
-
Ahrefs, AI Overviews Are Reducing Clicks, 300,000 keywords. The presence of an AI Overview was associated with a 34.5% lower clickthrough rate for the top-ranking page. ↩
Working on this same shift?
I write about SEO, GEO, and getting found by AI search.
If this resonated, I'd love to compare notes.