What is AI visibility, and why it decides deals you never see

Picture the buying process you never get to observe. A founder needs software in your category. She opens ChatGPT, describes her situation in two sentences, and asks what she should use. The engine names four products, summarizes their trade-offs, and suggests one to start with. She visits that vendor's site, takes the trial, and buys. Your analytics recorded nothing. Your sales team never got a call to lose. If your brand was absent from that answer, the deal ended before your funnel began, and no dashboard you own will ever show it.

This is not a hypothetical minority. Forrester's 2026 Buyers' Journey Survey of 18,000 buyers found 94 percent used AI during their most recent purchase, 55 percent compared vendors inside AI tools, and 47 percent built internal business cases before ever contacting a vendor. G2's research adds the sharpest numbers: 69 percent of B2B software buyers chose a different vendor than they originally expected because of AI chatbot output, and 33 percent bought from a brand they had never heard of before the AI named it. One in three purchases going to a previously unknown brand is shortlist formation by algorithm.

Now the uncomfortable pairing: G2 also found that 51 percent of B2B tech brands have zero citations across ChatGPT, Perplexity and Gemini. Half of every category is structurally absent from the conversations where these decisions form. AI visibility is the discipline of knowing whether you are in that half, and doing something about it.

AI visibility is how often, and how favorably, AI engines mention, recommend or cite your brand when users ask questions in your category. Three parts of that definition carry weight. It is measured on questions, because engines answer questions rather than rank keywords. It is measured per engine, because each engine has its own sources and habits. And it is measured over repeated runs, because a single answer is one draw from a probability distribution rather than a fact about the world.

The definition also splits into two currencies. A mention is your brand named in the answer text: "for small agencies, consider Acme." A citation is your page linked as a source under the answer. They have different mechanics and different payoffs: mentions shape shortlists even when nobody clicks, while citations carry the traffic and compound authority. A complete visibility program tracks both, and we separate them properly in mentions vs citations: the two currencies of AI visibility.

What AI visibility is analogous to, and what it replaces, is the rank report. For two decades, position in Google was the proxy for whether you existed at the moment of consideration. In categories where buyers now ask assistants first, presence in the answer is that proxy, and the rank report measures a stage of the journey that a growing share of buyers skip.

The surfaces where AI visibility plays out are no longer niche. ChatGPT reached roughly 900 million weekly active users in early 2026, about double a year earlier and on pace to cross a billion. Google's Gemini app surpassed 750 million monthly users. ChatGPT Search alone processes an estimated 250 to 500 million queries weekly, with Perplexity around 50 million. One 2026 estimate puts AI assistant sessions at about 45 billion worldwide, equal to roughly 56 percent of global search engine volume, though that figure is best treated as an estimate rather than a census.

Behavior tracks the surfaces. McKinsey reported in October 2025 that 50 percent of consumers already use AI-powered search intentionally as a primary way to find information and make buying decisions, and Aeolyft's 2026 U.S. Search Trends Report found 58 percent of Americans use AI weekly. Inside classic search, Google's AI Overviews trigger on roughly 48 percent of tracked queries by early 2026, with the honest caveat that panels disagree and some measure 13 to 25 percent depending on the query mix.

In B2B software specifically, the migration is measured twice over: G2 found 72 percent of buyers use ChatGPT during vendor evaluation, 44 percent also use Perplexity during shortlisting, and comparing vendor strengths and weaknesses is the top AI use case at 41 percent. The share starting research with a chatbot more often than Google jumped from 36 to 51 percent in seven months. Whatever the exact panel numbers, the direction and speed are beyond argument.

To manage AI visibility you need its mechanics, and they reduce to two channels. Channel one is training data: what the model learned about your brand before its knowledge cutoff. These mentions are durable and slow, changing only when the model is retrained, on schedules the AI labs do not publish. Channel two is live retrieval: at answer time the engine fetches pages from a search index and composes from them, usually citing sources. Retrieval-grounded answers can change within one to two weeks of the right page being published and indexed, which makes them the actionable half of the discipline. The strategic split is unpacked in two ways AI knows your brand.

The retrieval infrastructure is racing. ChatGPT's search launched on Bing's index, and OpenAI now runs its own crawler, OAI-SearchBot, which per Botify's analysis has roughly tripled its crawl since August 2025. Perplexity maintains an index reported above 50 billion pages. Google's AI features cite from Google's organic index. The consequence for brands is blunt: pages missing from Google's and Bing's indexes are invisible to most retrieval, so indexing remains the entry ticket it always was.

Scale figures and citation shares as published in the studies listed under sources; the ChatGPT citation numbers come from 5W Research's 2026 analysis and the Perplexity figures from cross-platform citation studies. Two patterns deserve emphasis. First, community content is disproportionately powerful: Reddit's citation share grew about 73 percent in commercial categories through 2025 and 2026, and it is Perplexity's single largest source. Second, prestige media underperforms its reputation, with WSJ, NYT and Bloomberg missing ChatGPT's top 20 cited domains entirely.

Choosing which engines to invest in first is its own decision, covered in which AI engines actually matter for your brand. The short answer for most brands: start with ChatGPT, which holds about 92 percent of trackable LLM referral traffic, then add engines as your category's buyers demonstrate they use them.

AI visibility measurement fails in predictable ways, and the failures all flatter the brand. The first trap is the single run: SparkToro measured under a 1 percent chance that two identical ChatGPT runs return the same brand list, so one favorable answer proves nothing and one unfavorable answer proves equally little. Honest measurement runs a fixed question set repeatedly and reports the trend.

The second trap is the branded question. Ask an engine "is Acme good for agencies" and the answer will discuss Acme by construction. Scores that mix branded and unbranded questions inflate quietly, which is why Reachroller excludes branded questions from its headline score entirely. The third trap is unverifiable scoring: a number with no raw answers behind it cannot be audited, and in a probabilistic surface an unauditable number is indistinguishable from a guess. Reachroller's answer is to count a mention only when the brand name literally appears in stored answer text, and to link every score to the answers it came from, a design documented on the methodology page.

Whether you use a tool or a spreadsheet, hold the measurement to those three standards: repeated runs, branded exclusion, auditable answers. Anything less produces a score that moves with the weather and gets quoted in board decks anyway.

The improvement playbook rests on the one controlled study the field has: the Princeton and Georgia Tech GEO research published at KDD 2024, which tested nine methods across 10,000 queries. Adding statistics, quotations and cited sources lifted a source's visibility in generative answers by up to roughly 40 percent, while keyword stuffing landed below baseline. Translated into practice: publish a page per losing question that answers it in the first paragraph and argues with numbers and named sources. The discipline built around this evidence is generative engine optimization, and it is the production half of any visibility program.

The second lever runs through pages you do not own. Given that Wikipedia and Reddit alone exceed a quarter of ChatGPT citations per 5W Research, earning honest presence on the third-party sources an engine already trusts for your questions often moves answers faster than publishing. The third lever is plumbing: submit new pages to Google Search Console and Bing Webmaster Tools, then allow one to two weeks before judging results, because retrieval reads indexes and indexes take time. Nothing in the evidence supports overnight flips, and vendors promising them are promising around the plumbing.

The visitors who arrive from AI answers convert unusually well, with the studies disagreeing on how well. WebFX, analyzing 2.3 billion sessions across 2024 and 2025, measured AI traffic growing 796 percent and AI-referred visitors converting about 1.2 times organic. Adobe found AI traffic converting 42 percent better in March 2026. Semrush reported around 4.4 times standard organic, and Ahrefs measured up to 23 times, a figure widely considered an outlier and cited here with that caveat. SE Ranking's study of 101,574 sites found ChatGPT referrals jumping 36.7 percent in May 2026 to an all-time high.

Two qualifiers keep the picture honest. AI referrals remain small in volume, around 0.18 percent of all sessions in some panels, and much AI-sourced traffic hides in direct where attribution cannot see it. And a large share of AI visibility's value never becomes a referral at all: it operates as shortlist presence in zero-click conversations. The click-based numbers are the measurable shadow of a larger effect, which is exactly why measuring the answers themselves, rather than only the traffic, is the point of the discipline.

Start manual and small. Write the 20 questions a buyer would ask an assistant before choosing in your category, unbranded and phrased like a human. Ask them in ChatGPT and Perplexity, save every answer, and record three things per question: does it name you, who does it name, and which sources does it cite. That single afternoon produces your baseline, your competitor map and your pitch list, and the full protocol is written up in the AI visibility audit.

When the manual loop proves the point, automate it. Reachroller runs your question set on schedule through official engine APIs, scores mentions auditably with branded questions excluded, generates a publish-ready fix page for every question you lose, complete with slug, title tag, meta description, schema markup and indexing steps, and rechecks whether the answer flipped. Starter is $29 per month for 400 credits and 25 tracked questions. The fair caveat: it is a young product, with ChatGPT live today and Claude, Gemini, Perplexity and Grok built and rolling out. The trial is three days, 50 credits, every feature, no card, which covers the full first report that tells you which half of your category you are in.