LLM Visibility / AI Visibility

TL;DR — LLMA large language model (LLM) is a deep-learning model trained on massive text corpora to predict the next token and generate human-like text. LLMs use the transformer architecture and power AI search features like Google's AI Overviews (Gemini) and Bing Copilot (GPT-4). visibility (also called AI visibility) is how often your brand shows up inside AI answers — the ones from Google AI OverviewsAI Overviews are the AI-generated summary box Google shows above or within its regular search results, written by Gemini models from pages retrieved out of Google's normal Search index. It's a Search feature, not a separate platform or index., ChatGPT, Perplexity, Copilot, and Gemini. It’s like organic visibility, but for written AI answers instead of a list of blue links. And here’s the catch: ranking #1 in Google doesn’t guarantee you show up in AI answers. They run on different signals.

What LLM visibility is

LLM visibilityLLM visibility (or AI visibility) is the aggregate measure of how often and how prominently a brand or page shows up in AI-generated answers — across AI Overviews, ChatGPT, Perplexity, Copilot, and Gemini. It's the AI-search analog of organic visibility, but it's driven by different signals. is an umbrella measurement for observed presence in model-assisted answers, not a standardized provider metric. Evidence for this claim ChatGPT search can answer with information from the web and provide linked sources. Scope: ChatGPT search; appearance in a sampled answer is product-, query-, locale-, and time-dependent. Confidence: high · Verified: 2026-07-14 OpenAI: ChatGPT search Crawl permission, retrieval, mention, and citation are distinct states. Evidence for this claim OpenAI distinguishes crawler controls for search inclusion from controls for model training. Scope: OpenAI's documented crawlers and controls; allowing a crawler does not guarantee retrieval, citation, or answer inclusion. Confidence: high · Verified: 2026-07-14 OpenAI: Crawlers

When you search now, you often get an AI-written answer at the top instead of (or above) the usual list of links. LLM visibility is the measure of how much your brand or your pages show up inside those answers.

It’s the AI-search version of a metric SEOs already know — organic visibility, the rolled-up sense of how present you are across search results. Same idea, new surface. Instead of “where do I rank for my keywords,” the question becomes “how often does AI mention or cite me when people ask about my topic.”

The part that surprises people

You’d think the pages that rank #1 in Google would be the ones AI quotes. Mostly they aren’t. Only about 38% of pages cited in Google’s AI OverviewsAI Overviews are the AI-generated summary box Google shows above or within its regular search results, written by Gemini models from pages retrieved out of Google's normal Search index. It's a Search feature, not a separate platform or index. also rank in the traditional top 10 — and that number used to be 76%. So your normal ranking reports don’t tell you how visible you are in AI. It’s a separate thing you have to measure separately.

The other surprise: most of your AI visibility comes from other people’s websites, not yours. When a site like Forbes or TechCrunch or a Reddit thread talks about you, that’s what AI tends to pull from. Roughly 82–89% of AI citations come from this kind of earned media, not from a brand’s own pages. As I put it in my Ahrefs Evolve 2025 talk on AI SEO, the shift is from “optimize your site” to “optimize how the internet talks about you.”

Three things, not one

“Showing up in AI” actually means three different things, and they’re easy to mix up:

1. Retrieved — the AI fetched your page to use as raw material. (You can see this in your server logsLog file analysis is reading a web server's raw access logs to see exactly which URLs search engine crawlers actually requested, when, how often, and what status code they got. Unlike crawl tools or Search Console, logs are the unsampled, ground-truth record of what really happened..)
2. Mentioned — your brand name shows up in the answer text.
3. Cited — your link shows up as a listed source.

Most tools only catch the third one. So if you only count citations, you’re seeing a small slice of your real influence.

And sometimes there’s nothing to catch at all — the AI answers straight from what it already learned during training, without fetching anything new. Those answers aren’t moved by content changes, no matter how good the content is. More on that distinction in the Advanced tab.

What to actually do

• Pick a list of questions your customers ask (say 250–500), and check how often AI answers mention or cite you. That percentage, tracked over time, is your score.
• Don’t check it once. AI answers change a lot — run the same question on different days and you’ll get different brands. One run isn’t a measurement.
• Build your reputation off-site too: get mentioned in articles, on Reddit, on YouTube. That’s what feeds the AI answers.

Want the full version — the correlation numbers, the platform-by-platform breakdown, and the exact measurement stack? Switch to the Advanced tab.

TL;DR — LLMA large language model (LLM) is a deep-learning model trained on massive text corpora to predict the next token and generate human-like text. LLMs use the transformer architecture and power AI search features like Google's AI Overviews (Gemini) and Bing Copilot (GPT-4). visibility is the aggregate of how often and how prominently a brand appears in AI-generated answers (AI OverviewsAI Overviews are the AI-generated summary box Google shows above or within its regular search results, written by Gemini models from pages retrieved out of Google's normal Search index. It's a Search feature, not a separate platform or index., AI Mode, ChatGPT, Perplexity, Copilot, Gemini, Grok). It is not organic visibility — only ~38% of AI Overview citations come from the traditional top 10 (was 76%), and the strongest correlate is branded web mentions (0.664), not Domain Rating (0.326). It splits into three states — retrieved → mentioned → cited — that need different tools; citation-only measurement understates the picture badly. It behaves radically differently per platform, it’s volatile enough that single prompt runs aren’t data points, and the right unit of measurement is a defined prompt pool tracked as Share of Model Voice over time. Tools like Brand Radar give directional indicators, not exact traffic counts.

What LLM visibility is (and isn’t)

Measurements are conditional on prompts, model versions, time, locale, and sampling. Evidence for this claim ChatGPT search can answer with information from the web and provide linked sources. Scope: ChatGPT search; appearance in a sampled answer is product-, query-, locale-, and time-dependent. Confidence: high · Verified: 2026-07-14 OpenAI: ChatGPT search Provider crawlerA crawler — also called a spider or bot — is an automated program that fetches web pages, extracts their links, and queues new URLs to visit. Search engines use crawlers to discover and download content for their index. controls describe access policy, not guaranteed inclusion in answers. Evidence for this claim OpenAI distinguishes crawler controls for search inclusion from controls for model training. Scope: OpenAI's documented crawlers and controls; allowing a crawler does not guarantee retrieval, citation, or answer inclusion. Confidence: high · Verified: 2026-07-14 OpenAI: Crawlers

LLM visibilityLLM visibility (or AI visibility) is the aggregate measure of how often and how prominently a brand or page shows up in AI-generated answers — across AI Overviews, ChatGPT, Perplexity, Copilot, and Gemini. It's the AI-search analog of organic visibility, but it's driven by different signals. — AI visibility, same thing — is the rolled-up measure of how often and how prominently a brand, domain, or page shows up in the answers AI searchAI search uses large language models and retrieval-augmented generation (RAG) to synthesize an answer from multiple sources rather than returning a ranked list of links. Examples include Google AI Overviews, ChatGPT Search, and Perplexity. systems generate. It’s a category, not a single number, the same way organic visibility is. The difference is the surface it measures: a synthesized, written answer instead of a ranked list of links.

The thing I’d burn into your brain first: LLM visibility is not the same as organic visibility. In mid-2025, ~76% of AI Overview citations came from pages ranking in the organic top 10. By early 2026 that’s down to roughly 38% in our data (and other studies put it even lower). Organic rank used to be a decent proxy for AI presence. It isn’t anymore. These are two different signals, and you need two different measurement programs.

Why the divergence? Because the signals that drive AI visibility aren’t the ones that drive rankings. In Ahrefs’ study of 75,000 brands by Louise Linehan and Xibeijia Guan — findings I later discussed at Ahrefs Evolve 2025 — the order came out like this for Google AI Overview visibility:

The #1 signal is brand presence off your own site. That’s the structural inversion: unlinked mentions — which pass no PageRankPageRank is Google's original recursive link-graph algorithm: a page's score depends on the scores of the pages linking to it, and in the published model each page's score is split across its outbound links (the simplified version: links are weighted votes). Google says it's evolved since launch but still part of its core ranking systems. and barely register in traditional SEO — are the strongest correlate of AI visibility. Which is exactly why I keep saying the job shifted from “optimize your site” to “optimize how the internet talks about you.” It lines up with the earned-media data too: 82–89% of AI citationsAn AI citation is the visible source link an AI answer engine shows next to its generated text — the clickable reference that credits the web page it used. A citation's presence is a separate thing from whether the cited page actually supports the statement, and from being retrieved (read behind the scenes) or merely mentioned (named without a link); citation is driven more by brand mentions and being retrievable than by traditional ranking. come from third-party sources (Forbes/TechCrunch/WSJ-type outlets), not brand-owned pages. One concrete example from the talk — Zapier had 16 pages mentioning Ahrefs, and those 16 pages were cited across 1,431 Ahrefs AI responses. Other people’s content about you does most of the work.

Three states: retrieved → mentioned → cited → clicked

“AI visibility” hides three genuinely different states. Conflating them is where most measurement goes wrong:

And then a fourth, downstream of all three: clicked — the rare case where someone actually visits your site from the answer (GA4 / web analytics).

Most tools only see the cited state. The retrieved and mentioned states are largely invisible to citation-only tools — OtterlyAI’s read of Bing’s data put it bluntly: “99.6% of your AI influence is invisible.” So if your dashboard is counting citations and calling that your AI visibility, you’re understating it by a lot. And the reverse failure exists too — Superlines found ~73% of AI presence was citations without a brand mention (“ghost citations”). Run mention tracking and citation tracking together or you get a false picture either way.

A question underneath all three: is this even a retrieval-driven answer?

Retrieved, mentioned, and cited all assume the AI went and fetched something for this specific query. A meaningful share of AI answers don’t — the model answers from what it already learned during training (its parametric memory), with no live retrieval step at all. See RAGRAG is the retrieve-then-generate pattern behind AI search: the system retrieves relevant passages from an external index at query time, injects them into the model's context, and generates an answer grounded in those sources — without changing the model's weights. for the mechanics: production systems run a query classifier that decides, per query, whether to search — it isn’t a step that happens on every request.

This matters for measurement because it changes what’s actionable. A retrieval-driven answer can, in principle, be moved by on-page and off-page work — better content gets fetched, better mentions get pulled in. A memory-driven answer can’t be moved that way: the model already “learned” what it knows about your brand at training time, and no page edit changes that until the model is retrained — which happens on the provider’s schedule, not yours. If you’re seeing a stable brand description across many prompt runs with no citations and no sign of retrieval, you may be looking at a memory-driven answer. Track it, but don’t spend editing effort expecting it to move.

There’s no fully reliable way to tell which is which purely by reading the output — it takes deliberately testing for it: comparing a platform’s search-on vs. search-off behavior where that’s exposed, watching whether citations appear at all, or checking whether the answer changes when the underlying source page changes. Treat “no citations at all” and “citations present” as different measurement regimes with different remedies, not two scores on the same scale.

Platform by platform — why visibility varies so much

There is no single “AI visibility” number, because the platforms behave completely differently. They draw from different sources and cite at wildly different rates:

The headline gap: Perplexity cites ~21.87 sources per response while ChatGPT cites in 0.59% of responses. The same content can be highly visible on one platform and effectively invisible on another. And cross-platform overlap is poor — only 7 of the top 50 most-cited domains appear across all three major platforms (AI Overviews, ChatGPT, Perplexity). LLM visibility is really three (or seven) separate visibility profiles. Measure each platform; don’t average them into one number and pretend it means something.

It’s volatile — single runs aren’t data

AI answers churn. Roughly 40–60% of cited sources change month to month, and for the same query a large share of the AI Overview content changes between runs. SparkToro found less than a 1-in-100 chance of getting identical brand lists across 100 runs of the same ChatGPT prompt. The methodological consequence is simple and non-negotiable: running a prompt once is not a data point. You need a defined prompt pool, many runs, and trend windows — not point-in-time snapshots.

How to measure it — the stack mapped to the states

No single tool covers all four states. Build a stack where each layer maps to one:

• Layer 1 — server logs → Retrieved. The -User bots (ChatGPT-User, Perplexity-User, Googlebot-Extended) are live inference fetches — someone asked a question and the AI went to get your page right then. That’s different from training/indexingStoring a crawled page in the search index so it can appear in results. Crawled is not the same as indexed — Google selects what to keep, and indexing isn't guaranteed. bots (GPTBot, ClaudeBot). Every error you serve a -User bot is a missed citation.
• Layer 2 — brand monitoring → Mentioned. Ahrefs Brand Radar does string-match mentions across the major platforms and has a “found but not cited” filter for isolating influence-without-attribution. Add manual prompt testing for QA.
• Layer 3 — citation tracking → Cited. Brand Radar’s Cited Pages / Cited Domains reports; Bing Webmaster ToolsMicrosoft's free portal for monitoring and improving how a site appears in Bing search — the peer to Google Search Console, plus IndexNow instant indexing, richer backlink data, and keyword volumes. Because Bing's index also feeds Microsoft Copilot, it doubles as a window into AI-search visibility.’ AI Performance reportThe Google Search Console report that shows how your site actually performed in Google Search, built from real impressions and clicks. It reports four metrics — clicks, impressions, average CTR, and average position — and keeps the most recent 16 months of data. (the first official platform source of citation data — citations, groundingGrounding is anchoring an AI model's answer to source documents it retrieves at the moment you ask — not to the patterns frozen into its weights during training. Retrieval-Augmented Generation (RAG) is the most common way to do it. queries, average cited pages); and Google Search ConsoleA free Google service that reports how a site performs in Google Search and surfaces problems with how Google crawls, indexes, and serves it. It's first-party data straight from Google — but you don't need it to appear in results.’s AI features filter / Gen AI Performance reports (impressions only — no clicks).
• Layer 4 — web analytics → Clicked. GA4’s AI Assistant channel, plus a custom channel group. Expect a dark-traffic problem: a large share of AI-sourced visits arrive with no referrer and land in Direct, and AI Overview/AI Mode clicks merge into google/organic — so the clicked layer is the least clean.

One caveat I want to be honest about: Brand Radar (and tools like it) report directional indicators, not exact traffic counts. Treat the trend and the relative share seriously; don’t treat any single figure as a precise tally.

TIP Classify a sampled answer without inventing missing-provider data

The pictured run is deliberately narrow: one simulated provider response, retrieval off, and five providers marked not evaluated. That makes it useful for classification practice—not for claiming market-wide visibility.

Check whether a sampled answer mentions the brand and contains a supplied known fact with my free AI Brand Visibility Checker Free

1. Supply the answer text, target brand, and one fact whose presence can be checked directly.
2. Record mention and known-fact coverage independently, along with provider and retrieval state.
3. Repeat the controlled capture across the prompt pool and providers before trending visibility.

Tap to enlarge

A simulated Llama answer mentions the fictional Acme Analytics brand by exact match, while the known-fact check marks Founded in 2018 as missing. Five other provider profiles are explicitly not evaluated and warn against inferring zero mention or citation rates.

The prompt-pool method and Share of Model Voice

The unit of measurement is the prompt pool: define ~250–500 high-intent queries, run them across the LLM endpoints on a weekly or monthly cadence, record where your brand appears, and track Share of Model Voice (SOMV) — brand appearances ÷ total tracked prompts × 100. Pull prompts from real demand: sales transcripts, support tickets, Reddit, G2 reviews, PAA, autocomplete. Benchmarks to calibrate against: average mention rate is around 17.2% across relevant prompts (AthenaHQ), and 40–70% is considered strong.

For platform mechanics, Brand Radar runs on a 400M+ search-backed prompt corpus (up from 350M+ earlier this year — it’s a growing, live-updating index, so treat the exact figure as directional), refreshes monthly for ChatGPT/Perplexity/Gemini/Copilot and continuously for AI Overviews/AI Mode, and distinguishes “cited” (linked URLs) from “found but not cited” (string matches without links).

A measurement edge case worth knowing

AI doesn’t just under-report you — sometimes it invents you. We had thousands of visits going to pages on Ahrefs that didn’t actually exist because an AI synthesized plausible-looking URLs. Hallucinated citations are a real dimension of AI-visibility measurement, and most tools won’t filter them out for you — so sanity- check the URLs you’re “cited” on.

What this is and isn’t worth

Be clear-eyed about value. Even being cited rarely produces a click — users clicked a cited source in only 1% of Google visits with an AI summary in Pew’s March 2025 U.S. browsing study, and the vast majority of Google AI Mode sessions end with no site visit. LLM visibility is primarily a brand awareness and authority metric, not a direct-traffic metric. It still matters: a large share of AI Mode users accept the curated shortlist without doing further research, so being in the answer carries weight even without the click.

Where this fits

This is the definition-and-measurement hub for the cluster. For the deep dive on the three states, see Retrieved, Mentioned, CitedThree distinct states of AI visibility: retrieved (an AI fetched your page as source material), mentioned (your brand appears in the answer text), and cited (your URL is linked as a source). They don't always happen together, and each is measured with a different tool.. For the GA4/analytics layer specifically — the dark-traffic and attribution mess — see AI traffic attributionAI traffic attribution is the practice of correctly identifying and measuring website visits that come from AI tools — ChatGPT, Perplexity, Gemini, Claude, AI Overviews, and AI browsers. It's hard because many of those tools strip the referrer header, so the visits land in your analytics as Direct traffic with no source.. For the off-site signal that actually drives this (entities and mentions), see Entity SEOEntity SEO is the practice of helping search engines and AI systems clearly identify, classify, and trust the entities you represent — your brand, your people, your products — rather than just matching keyword strings. The goal is to be an unambiguous, well-corroborated entity in machine knowledge systems so AI can cite you with confidence. and the broader AI search hubAI search uses large language models and retrieval-augmented generation (RAG) to synthesize an answer from multiple sources rather than returning a ranked list of links. Examples include Google AI Overviews, ChatGPT Search, and Perplexity..

AI summary

A condensed take on the Advanced version:

• LLMA large language model (LLM) is a deep-learning model trained on massive text corpora to predict the next token and generate human-like text. LLMs use the transformer architecture and power AI search features like Google's AI Overviews (Gemini) and Bing Copilot (GPT-4). visibility = how often/prominently you appear in AI answers (AI OverviewsAI Overviews are the AI-generated summary box Google shows above or within its regular search results, written by Gemini models from pages retrieved out of Google's normal Search index. It's a Search feature, not a separate platform or index., AI Mode, ChatGPT, Perplexity, Copilot, Gemini, Grok). It’s a composite category, the AI-search analog of organic visibility.
• It’s not organic visibility. Only ~38% of AI Overview citations come from the traditional top 10 (was 76%). Different signals, separate measurement needed.
• The #1 correlate is branded web mentions (0.664) — well above Domain Rating (0.326). 82–89% of AI citationsAn AI citation is the visible source link an AI answer engine shows next to its generated text — the clickable reference that credits the web page it used. A citation's presence is a separate thing from whether the cited page actually supports the statement, and from being retrieved (read behind the scenes) or merely mentioned (named without a link); citation is driven more by brand mentions and being retrievable than by traditional ranking. are earned media, not brand-owned pages. “Optimize how the internet talks about you.”
• Three states, three tools: retrieved (server logsLog file analysis is reading a web server's raw access logs to see exactly which URLs search engine crawlers actually requested, when, how often, and what status code they got. Unlike crawl tools or Search Console, logs are the unsampled, ground-truth record of what really happened.) → mentioned (Brand Radar string match) → cited (Brand Radar cited / Bing WMT / GSCA free Google service that reports how a site performs in Google Search and surfaces problems with how Google crawls, indexes, and serves it. It's first-party data straight from Google — but you don't need it to appear in results.) → clicked (GA4). Citation-only tools miss most of it (“99.6% invisible”); watch for “ghost citations” too.
• Underneath all three: some answers skip retrieval entirely. A memory-driven answer comes from the model’s training, not a live fetch — no citations, no retrieval signal, and no way to move it with content changes. Retrieval-driven answers are the ones on-page/off-page work can actually shift.
• Platforms differ radically: Perplexity ~21.87 citations/response vs. ChatGPT ~0.59% citation rate; only 7 of the top 50 cited domains span all three majors.
• Volatile: 40–60% of citations change month to month; a single prompt run is not a data point.
• Method: a 250–500 prompt pool, run on a cadence, tracked as Share of Model Voice (avg ~17.2%; 40–70% strong). Brand Radar data is directional, not exact traffic counts.
• Value: mostly brand/authority — clicks are rare (Pew observed cited-source clicks in 1% of Google visits with an AI summary).

Official documentation

What the engines themselves say about AI visibility and how to measure it.

Google

• AI features and your website (AI optimization guide) — Google’s core position: AI OverviewsAI Overviews are the AI-generated summary box Google shows above or within its regular search results, written by Gemini models from pages retrieved out of Google's normal Search index. It's a Search feature, not a separate platform or index. and AI Mode are rooted in core Search ranking, so standard SEO best practices remain the foundation. No special markup (llms.txt), chunkingChunking is splitting a document into smaller passages so AI systems can embed, index, and retrieve the single most relevant piece — not the whole page — in response to a query. It's a foundational step in RAG pipelines and the conceptual cousin of Google's passage ranking., or “AI-specific writing” is required or beneficial.
• AI features in Google Search results — how AI OverviewsAI Overviews are the AI-generated summary box Google shows above or within its regular search results, written by Gemini models from pages retrieved out of Google's normal Search index. It's a Search feature, not a separate platform or index./AI Mode surface content and what’s eligible.
• Gen AI Performance reports in Search Console (June 2026) — the AI features data in GSCA free Google service that reports how a site performs in Google Search and surfaces problems with how Google crawls, indexes, and serves it. It's first-party data straight from Google — but you don't need it to appear in results.: impressions for AI Overviews, AI Mode, and Discover generative features. Impressions only — no click data.

Bing / Microsoft

• Introducing AI Performance in Bing Webmaster Tools (public preview, Feb 2026) — the first official platform tool exposing AI citationAn AI citation is the visible source link an AI answer engine shows next to its generated text — the clickable reference that credits the web page it used. A citation's presence is a separate thing from whether the cited page actually supports the statement, and from being retrieved (read behind the scenes) or merely mentioned (named without a link); citation is driven more by brand mentions and being retrievable than by traditional ranking. data: Citations, GroundingGrounding is anchoring an AI model's answer to source documents it retrieves at the moment you ask — not to the patterns frozen into its weights during training. Retrieval-Augmented Generation (RAG) is the most common way to do it. Queries, and Average Cited Pages.
• Microsoft Copilot guidance for public websites — how Copilot Studio’s generative-answers node indexes and ranks web content used as a knowledge source (via sitemapsA sitemap is a file that lists the pages, images, videos, and other files on your site so search engines can discover them. It helps discovery, but submitting a sitemap doesn't guarantee crawling or indexing./IndexNowIndexNow is an open push protocol that lets you instantly tell participating search engines (Bing, Yandex, Naver, Seznam, and Yep) which URLs you've added, changed, or removed via a simple HTTP request — and one submission is shared across all of them. Google does not use it., then ranked by relevance, freshness, and engagement).

Quotes from the source

On-the-record statements from the engines. Each deep link jumps toward the relevant passage on the source page.

Google — standard SEO still applies

• “The best practices for SEO continue to be relevant because our generative AI features on Google Search are rooted in our core Search ranking and quality systems.” — Google Search Central. Jump to quote

Bing — why AI Performance exists

• “As AI becomes a more common way people discover information, visibility is not only about blue links. It is also about whether your content is cited and referenced when AI systems generate answers.” — Bing WebmasterMicrosoft's free portal for monitoring and improving how a site appears in Bing search — the peer to Google Search Console, plus IndexNow instant indexing, richer backlink data, and keyword volumes. Because Bing's index also feeds Microsoft Copilot, it doubles as a window into AI-search visibility. Blog, “Introducing AI Performance in Bing Webmaster ToolsThe free, first-party consoles search engines give site owners — Google Search Console and Bing Webmaster Tools — to see how their pages are crawled, indexed, and ranked, and to fix problems. The legacy name is 'webmaster tools.'” (Feb. 10, 2026). Jump to source

LLM visibility audit checklist

A pass to set up — and sanity-check — an AI visibility measurement program. It maps to the three states, so you cover more than just citations.

Scope

• Defined which platforms you’ll track (at minimum: AI OverviewsAI Overviews are the AI-generated summary box Google shows above or within its regular search results, written by Gemini models from pages retrieved out of Google's normal Search index. It's a Search feature, not a separate platform or index., ChatGPT, Perplexity; ideally + Copilot, Gemini, Grok).
• Listed the brand entities, regions, and languages in scope.

Retrieved (server logsLog file analysis is reading a web server's raw access logs to see exactly which URLs search engine crawlers actually requested, when, how often, and what status code they got. Unlike crawl tools or Search Console, logs are the unsampled, ground-truth record of what really happened.)

• Checked server logs for AI inference botsA crawler — also called a spider or bot — is an automated program that fetches web pages, extracts their links, and queues new URLs to visit. Search engines use crawlers to discover and download content for their index. (ChatGPT-User, Perplexity-User, Googlebot-Extended) — separate from training bots (GPTBot, ClaudeBot).
• Confirmed you’re not serving errors/blocks to -User fetch bots (every error is a missed citation).

Mentioned + Cited (Brand Radar / Bing WMT / GSCA free Google service that reports how a site performs in Google Search and surfaces problems with how Google crawls, indexes, and serves it. It's first-party data straight from Google — but you don't need it to appear in results.)

• Built a baseline in Ahrefs Brand Radar (mentions + cited pages/domains).
• Ran the “foundA 302 (\"Found\") is a temporary redirect: it forwards users to a new URL while telling search engines the original URL should stay in the index. It's a weak canonicalization signal, not the zero-equity dead end of SEO folklore. but not cited” filter to find influence-without-attribution.
• Pulled Bing Webmaster ToolsMicrosoft's free portal for monitoring and improving how a site appears in Bing search — the peer to Google Search Console, plus IndexNow instant indexing, richer backlink data, and keyword volumes. Because Bing's index also feeds Microsoft Copilot, it doubles as a window into AI-search visibility. AI Performance (citations, groundingGrounding is anchoring an AI model's answer to source documents it retrieves at the moment you ask — not to the patterns frozen into its weights during training. Retrieval-Augmented Generation (RAG) is the most common way to do it. queries, avg cited pages).
• Checked GSC AI features / Gen AI Performance (impressions — remember, no clicks).

Prompt testing

• Tested your key prompts manually across platforms — multiple runs, not one (results are volatile).
• Checked accuracy and sentiment, not just presence (fix misinformation).
• Spot-checked cited URLs for hallucinated/non-existent pages.

Clicked + tracking

• Reviewed GA4 AI Assistant channel / custom channel group (expect dark trafficAI traffic attribution is the practice of correctly identifying and measuring website visits that come from AI tools — ChatGPT, Perplexity, Gemini, Claude, AI Overviews, and AI browsers. It's hard because many of those tools strip the referrer header, so the visits land in your analytics as Direct traffic with no source. in Direct).
• Set a recurring cadence — monthly minimum, weekly for competitive categories — with trend windows, not single snapshots.
• Benchmarked Share of Model Voice vs. competitors.

LLM visibility vs. organic visibility — cheat sheet

The fast comparison. The point isn’t that one replaces the other — it’s that they’re different metrics on different signals and need separate measurement.

Three states to track (don’t collapse them): Retrieved (server logsLog file analysis is reading a web server's raw access logs to see exactly which URLs search engine crawlers actually requested, when, how often, and what status code they got. Unlike crawl tools or Search Console, logs are the unsampled, ground-truth record of what really happened.) → Mentioned (Brand Radar string match) → Cited (Brand Radar cited / Bing WMT / GSC) → Clicked (GA4).

Don’t: treat citation counts as total AI visibility (you miss retrieved + mentioned), assume one platform represents the others, or treat a single prompt run as a measurement. And llms.txt does not affect Google ranking — Google says it ignores it.

Tools for measuring LLM visibility

No single tool covers all four states — most lean toward the cited layer. Pair a brand-monitoring tool with the official platform reports and your own logs.

• Ahrefs Brand Radar — 400M+ search-backed prompts (up from 350M+ earlier this year; it’s a growing, live-updating indexStoring a crawled page in the search index so it can appear in results. Crawled is not the same as indexed — Google selects what to keep, and indexing isn't guaranteed. — treat the exact figure as directional); tracks mentions, citations, impressions, and Share of Model Voice across ChatGPT, Perplexity, Gemini, Copilot, AI OverviewsAI Overviews are the AI-generated summary box Google shows above or within its regular search results, written by Gemini models from pages retrieved out of Google's normal Search index. It's a Search feature, not a separate platform or index., AI Mode, and Grok. Monthly refresh for the conversational platforms, continuous for AI OverviewsAI Overviews are the AI-generated summary box Google shows above or within its regular search results, written by Gemini models from pages retrieved out of Google's normal Search index. It's a Search feature, not a separate platform or index.. Has a “foundA 302 (\"Found\") is a temporary redirect: it forwards users to a new URL while telling search engines the original URL should stay in the index. It's a weak canonicalization signal, not the zero-equity dead end of SEO folklore. but not cited” filter. Directional indicators, not exact traffic counts.
• Semrush AI Visibility Toolkit — citation frequency/position, share of voice, and competitive AI-market-share tracking over a large LLMA large language model (LLM) is a deep-learning model trained on massive text corpora to predict the next token and generate human-like text. LLMs use the transformer architecture and power AI search features like Google's AI Overviews (Gemini) and Bing Copilot (GPT-4).-prompt corpus.
• Profound — enterprise-leaning, programmatic API querying of LLM endpoints for Share of Model Voice tracking, plus AI-crawlerA crawler — also called a spider or bot — is an automated program that fetches web pages, extracts their links, and queues new URLs to visit. Search engines use crawlers to discover and download content for their index. analytics.
• Scrunch — AI-search visibility platform with citation tracking, bot-crawl observability, and content optimization for AI answers; reducing brand hallucinationsAn AI hallucination is when a large language model generates output that is confidently stated but factually wrong, made up, or unsupported by its source. It's a side effect of next-token prediction — not a bug that can be fully eliminated. is one of its stated outcomes.
• Peec AI — daily updates and source-gap analysis on a smaller prompt set; lightweight entry point.
• Waikay.io — multi-platform (ChatGPT, Gemini, Claude, Perplexity) rolled up into an AI Brand Score.
• Bing Webmaster ToolsMicrosoft's free portal for monitoring and improving how a site appears in Bing search — the peer to Google Search Console, plus IndexNow instant indexing, richer backlink data, and keyword volumes. Because Bing's index also feeds Microsoft Copilot, it doubles as a window into AI-search visibility. — AI Performance (official, free) — citations, groundingGrounding is anchoring an AI model's answer to source documents it retrieves at the moment you ask — not to the patterns frozen into its weights during training. Retrieval-Augmented Generation (RAG) is the most common way to do it. queries, average cited pages. Best window into Copilot’s cited state.
• Google Search ConsoleA free Google service that reports how a site performs in Google Search and surfaces problems with how Google crawls, indexes, and serves it. It's first-party data straight from Google — but you don't need it to appear in results. — AI features / Gen AI Performance (official, free) — impressions for AI Overviews/AI Mode/Discover. No clicks.

AI-visibility measurement mistakes

Reporting one prompt run as market share

Outputs vary by platform, model, time, location, and phrasing. Use a defined prompt pool and repeated observations before describing a trend.

Combining retrieval, mentions, and citations

A system can fetch a page without naming the brand, name a brand without linking, or cite a URL. Track each state separately because the measurement source and business meaning differ.

Comparing tools as if their panels were identical

Visibility platforms use different prompts, models, schedules, and scoring formulas. Keep one methodology stable for trend reporting and disclose what it can and cannot observe.

Build a controlled visibility prompt pool

Paste verified customer jobs, category terms, comparison needs, and funnel stages:

Create a balanced AI-visibility prompt pool using only the supplied research. Group
prompts by informational, comparison, recommendation, and problem-solving intent.
Keep wording neutral, avoid naming our brand unless the task is branded measurement,
and flag prompts that depend on location, freshness, or user context. Return a stable
prompt ID, wording, intent, audience, and reason for inclusion.

Classify captured AI answers

Paste answer text, citations, platform, model/version, query time, and prompt ID:

Classify our brand as absent, mentioned, or cited. Record cited URLs exactly, describe
position and sentiment without assigning a numerical score, and separate observable
facts from interpretation. Flag hallucinated claims and ambiguous brand references.
Do not infer retrieval unless server-log evidence is supplied.

The Pool, Capture, Classify, Trend framework

1. Pool: Freeze a representative set of prompts with stable IDs, intents, and audience context.
2. Capture: Record platform, model/version, date, answer, and citations for every run under consistent conditions.
3. Classify: Separate absence, mention, citation, sentiment, factual accuracy, and competitor presence without inferring unseen retrieval.
4. Trend: Compare like-for-like observations over time and annotate methodology or product changes that break continuity.

The framework produces directional visibility evidence. It does not convert prompt observations into exact traffic or market-share estimates.

Test yourself: LLM visibility

Resources worth your time

My speaking

• GEO, AEO, LLMO — What’s With All This AI Stuff? (Ahrefs Evolve 2025) — the slides where I discuss Ahrefs’ correlation study (branded web mentions 0.664 vs. DR 0.326) and present the “optimize how the internet talks about you” framing. Video · Webinar version.

My related writing

• AI Visibility Audit: A Step-by-Step Guide — the 8-step audit methodology (scope → baseline → accuracy/sentiment → unbranded gaps → top-cited pages → mentions → competitors → strategy).
• How Brand Radar works (methodology) — prompt selection, platforms tracked, refresh cadence, and the “cited” vs. “foundA 302 (\"Found\") is a temporary redirect: it forwards users to a new URL while telling search engines the original URL should stay in the index. It's a weak canonicalization signal, not the zero-equity dead end of SEO folklore. but not cited” distinction.
• Insights from 56 Million AI Overviews — the overlap-with-organic data.
• AI search traffic converts 23x better — why AI visibility is worth measuring even with few clicks.
• Generative engines are breaking web analytics — the attribution/dark-traffic problem.
• GEO: brand mentions vs. DR correlation study.

From others

• Danny Sullivan / Google: “Good SEO is good GEO” — the official line that AI visibility rides on SEO fundamentals.
• Duane Forrester — 12 new KPIs for generative AI search — a broader metric framework including chunk retrieval frequency, attribution rate, and AI citationAn AI citation is the visible source link an AI answer engine shows next to its generated text — the clickable reference that credits the web page it used. A citation's presence is a separate thing from whether the cited page actually supports the statement, and from being retrieved (read behind the scenes) or merely mentioned (named without a link); citation is driven more by brand mentions and being retrievable than by traditional ranking. count.
• OtterlyAI: The AI Citations Report 2026 — “99.6% of your AI influence is invisible” (Bing data); found 73% of sites have crawlability issuesCrawlability is how well search engine crawlers can discover, access, and fetch a site's pages. A crawlability issue is any technical condition — blocked access, broken links, server failures, or bloated URL inventory — that stops pages from reaching the index. affecting citations.
• SparkToro: AIs are highly inconsistent when recommending brands — less than 1-in-100 chance of identical brand lists across 100 runs of the same ChatGPT prompt; key caution for measurement methodology.
• Superlines: AI Search Statistics 2026 — platform-by-platform citation rates (Grok 27%, ChatGPT 0.59%); documents the “ghost citation” phenomenon (73% of AI presence lacks a brand mention); 70% AI OverviewAI Overviews are the AI-generated summary box Google shows above or within its regular search results, written by Gemini models from pages retrieved out of Google's normal Search index. It's a Search feature, not a separate platform or index. content churn per query.
• Discovered Labs: How each platform cites sources differently — ChatGPT cites 15% of retrieved pages; 82.9% of B2B citations from third-parties.
• Princeton GEO paper (Aggarwal et al., KDD 2024) — rank-5 pages showed +115.1% citation frequency vs. rank-1 pages at -30.3%; original data on what content attributes actually drive AI citationsAn AI citation is the visible source link an AI answer engine shows next to its generated text — the clickable reference that credits the web page it used. A citation's presence is a separate thing from whether the cited page actually supports the statement, and from being retrieved (read behind the scenes) or merely mentioned (named without a link); citation is driven more by brand mentions and being retrievable than by traditional ranking..

Stats worth citing

• Branded web mentions correlate 0.664 with AI OverviewAI Overviews are the AI-generated summary box Google shows above or within its regular search results, written by Gemini models from pages retrieved out of Google's normal Search index. It's a Search feature, not a separate platform or index. visibility — vs. anchor text 0.527, branded search 0.392, and Domain Rating 0.326. The off-site signal dominates. (Ahrefs’ 75,000-brand Google AI Overview study by Louise Linehan and Xibeijia Guan.) Source
• Only ~38% of pages cited in AIThree distinct states of AI visibility: retrieved (an AI fetched your page as source material), mentioned (your brand appears in the answer text), and cited (your URL is linked as a source). They don't always happen together, and each is measured with a different tool. Overviews rank in the traditional top 10 — down from ~76% in mid-2025. Organic rank no longer proxies AI visibility. Source
• 82–89% of AI citationsAn AI citation is the visible source link an AI answer engine shows next to its generated text — the clickable reference that credits the web page it used. A citation's presence is a separate thing from whether the cited page actually supports the statement, and from being retrieved (read behind the scenes) or merely mentioned (named without a link); citation is driven more by brand mentions and being retrievable than by traditional ranking. come from earned media (Forbes/TechCrunch/WSJ-type), not brand-owned pages. Zapier’s 16 pages mentioning Ahrefs → 1,431 cited Ahrefs AI responses. Source
• Average brand mention rate ≈ 17.2% across relevant prompts; 40–70% is strong (AthenaHQ, State of AI SearchAI search uses large language models and retrieval-augmented generation (RAG) to synthesize an answer from multiple sources rather than returning a ranked list of links. Examples include Google AI Overviews, ChatGPT Search, and Perplexity. 2026).
• 40–60% of cited sources change month to month — AI visibility is volatile, so measure with prompt pools and trend windows, not single runs (EMARKETER).
• Platforms differ ~45x: Perplexity ~21.87 citations/response (~13% citation rate) vs. ChatGPT ~0.59%; only 7 of the top 50 most-cited domains appear across all three major platforms.
• Brand Radar corpus: 400M+ search-backed prompts (up from 350M+ earlier this year — a growing, live-updating indexStoring a crawled page in the search index so it can appear in results. Crawled is not the same as indexed — Google selects what to keep, and indexing isn't guaranteed.) — monthly refresh for ChatGPT/Perplexity/Gemini/Copilot, continuous for AI OverviewsAI Overviews are the AI-generated summary box Google shows above or within its regular search results, written by Gemini models from pages retrieved out of Google's normal Search index. It's a Search feature, not a separate platform or index.. Directional, not exact traffic counts. Source
• Users clicked a cited source in 1% of Google visits with an AI summary in Pew’s March 2025 U.S. browsing study. In Semrush’s separate clickstream sample, 92–94% of Google AI Mode sessions did not lead to an external-domain visit. LLMA large language model (LLM) is a deep-learning model trained on massive text corpora to predict the next token and generate human-like text. LLMs use the transformer architecture and power AI search features like Google's AI Overviews (Gemini) and Bing Copilot (GPT-4). visibility is primarily a brand awareness metric, not a traffic metric.
• Less than 1-in-100 chance of identical brand lists across 100 runs of the same ChatGPT prompt — AI recommendations are highly inconsistent, making single-run measurements statistically unreliable. (SparkToro) Source
• 99.6% of AI influence is invisible to citation-only measurement tools, per OtterlyAI’s analysis of Bing data — retrieved and mentioned states dwarf the cited state in volume. Source

LLM Visibility

LLM visibility (or AI visibility) is the aggregate measure of how often and how prominently a brand or page shows up in AI-generated answers — across AI Overviews, ChatGPT, Perplexity, Copilot, and Gemini. It's the AI-search analog of organic visibility, but it's driven by different signals.

Related: AI Share of Voice, Retrieved vs. Mentioned vs. Cited, Measurement and Reporting

LLM Visibility

LLMA large language model (LLM) is a deep-learning model trained on massive text corpora to predict the next token and generate human-like text. LLMs use the transformer architecture and power AI search features like Google's AI Overviews (Gemini) and Bing Copilot (GPT-4). visibility is a composite, not a single number. It captures how present your brand or content is inside the written, synthesized answers that AI searchAI search uses large language models and retrieval-augmented generation (RAG) to synthesize an answer from multiple sources rather than returning a ranked list of links. Examples include Google AI Overviews, ChatGPT Search, and Perplexity. systems produce — Google AI OverviewsAI Overviews are the AI-generated summary box Google shows above or within its regular search results, written by Gemini models from pages retrieved out of Google's normal Search index. It's a Search feature, not a separate platform or index. and AI Mode, ChatGPT, Perplexity, Bing Copilot, Gemini, and Grok. It sits next to “organic visibility” as a category of measurement, but it measures presence in a fundamentally different space: an answer someone reads, not a list of links they choose from.

It breaks into three distinct states that get conflated all the time. Retrieved means an AI’s retrieval system fetched your page as source material (you see this in server logsLog file analysis is reading a web server's raw access logs to see exactly which URLs search engine crawlers actually requested, when, how often, and what status code they got. Unlike crawl tools or Search Console, logs are the unsampled, ground-truth record of what really happened.). Mentioned means your brand name appears in the generated text. Cited means your URL is explicitly linked as a source. Most tools only see the cited state, which dramatically understates real influence — a lot of retrieval and mention never produces a visible citation.

The big surprise for SEOs is that LLM visibility and organic visibility are not the same thing. Only about 38% of pages cited in AIThree distinct states of AI visibility: retrieved (an AI fetched your page as source material), mentioned (your brand appears in the answer text), and cited (your URL is linked as a source). They don't always happen together, and each is measured with a different tool. Overviews also rank in the traditional top 10 (down from ~76% in mid-2025), and the strongest correlate of AI Overview visibility is branded web mentions (~0.664), not Domain Rating (~0.326). It’s measured platform by platform with a prompt pool, tracked over time as Share of Model Voice — and because AI answers are volatile, a single prompt run is never a data point.

Related: AI Share of Voice, Retrieved vs. Mentioned vs. Cited, Measurement and Reporting

Changelog