What is AI visibility?

AI visibility, defined

When someone asks an answer engine "what is the best accounting software for a small agency", it does not return ten links. It writes an answer that names a few brands and links a few sources. AI visibility is how often you are in those answers, across the questions your market asks and the engines your buyers use.

It has four parts, and they move independently:

• Named: the answer mentions your brand.
• Cited: the answer links one of your pages as a source.
• Top pick: you are the first or only recommendation.
• Portrayal: what the answer says about you is accurate and fair.

The IAB's August 2026 framework, Measuring Visibility in the AI Era, groups the same ideas as presence, prominence, portrayal and persuasion. Whatever the labels, keep the parts apart: a brand that is named often but never cited is known to the model and absent from the pages it reads, and the fix for that is different from the fix for a brand that is cited but rarely recommended.

The work of growing it goes by several names, GEO, AEO and LLM SEO among them; GEO vs AEO vs SEO explains how they relate.

Not the security meaning. In IT security, "AI visibility" also means seeing which AI tools employees use. This guide is about marketing: how AI answers treat your brand.

Why it matters now

B2B buyers increasingly start with an answer engine: they ask for options, compare two vendors, or check pricing before they ever visit a website. The shortlist forms inside the answer, and a brand that is not named is not considered.

The traffic data shows the same shift. Seer Interactive measured organic click-through on queries with an AI Overview falling from 1.76% to 0.61% between mid-2024 and September 2025 (Seer Interactive). Ahrefs found an AI Overview went with a 58% lower click-through rate for the top-ranking page (Ahrefs, February 2026). Pew found people clicked a regular result in 8% of visits with an AI summary, against 15% without one (Pew Research). Google says AI Mode passed one billion monthly users within about a year of launch (Google, May 2026).

Our view: the click is no longer the unit to manage. Being in the answer is. Traffic still matters, but it is now a late and partial signal of whether you were considered at all.

AI visibility and SEO

SEO is still the foundation: engines that search the web read pages they can find, and Google's AI features use pages that are indexed and eligible for a snippet (Google Search Central). But ranking and being named are not the same result.

The overlap is real but partial. In our experiments of 1 October 2026, 66.3% of the brands an answer recommended ranked in the top 10 for one of the searches the engine ran, against 30.7% of the brands it left out (method). Ahrefs found 38% of the pages AI Overviews cite also rank in the top 10 (Ahrefs, March 2026).

How answer engines decide whom to name

An engine assembles an answer from three places. Each one can favour a different brand, which is why a competitor can lead in ChatGPT and trail in Perplexity.

What it remembers

The model's training gives it a view of every category before it searches anything. That view changes only when the model is retrained. It is also not the same as what the engine says once it searches: ChatGPT's recommendations with search off and on overlapped by only 32% in our tests.

The pages it reads

With web search on, the engine reads a handful of pages and leans on what they say. A brand named on six or more of the pages other answers cited was named in 34.9% of answers, against 11.9% for a brand on none of them. Most of those pages are not yours: Muck Rack found 84% of the links ChatGPT, Claude and Gemini cite go to earned media rather than brand or paid content (Muck Rack, May 2026).

The searches it runs

Before answering, many engines search, sometimes several times. The words they add tell you which pages to have: in our run, 45% of their searches contained "best", 30% the current year, 11% "vs" and 10% "alternative". For commercial questions, "best X" lists are the pages engines cite most (Ahrefs).

How to measure AI visibility

Measuring AI visibility means sampling answers, because there is no Search Console for ChatGPT and no public count of what people ask it. Four rules make the sample worth reading.

1. Group questions into topics. Measure "payroll software for small businesses" as a topic of many phrasings, not one prompt.
2. Ask many different questions. Two answers to the same question shared 68% of their brands in our tests, while answers to different questions shared 21%. Variety measures the topic; repeats mostly measure one answer.
3. Read share, not position. SparkToro found AI tools almost never return the same list twice and concluded that a visibility percentage across many prompts is a reasonable measure, while a "ranking position in AI" is not (SparkToro, January 2026).
4. Keep the range. A share from a few dozen answers is a direction, not a decision. The IAB framework draws the same line between directional and decision-grade measurement.

In Vercatus the result is Answer Share. Answer Share is the share of AI answers on your topics that name or cite you, shown with its range and next to your competitors on the same scale. How AI visibility tracking works covers the details, and a basic manual audit shows how to get a first reading by hand.

How to grow AI visibility

Ranked by what we have seen move answers, most reliable first. How to improve your brand's visibility in AI search takes each move step by step.

1. Let the answer engines' crawlers in

An engine cannot cite a page it cannot fetch. When we probed 64 well-known websites on 18 September 2026, 27 served their home page to none of the three AI assistant crawlers we sent and 4 to only some (the study). Often it is a firewall default, not a decision. The free check shows yours in a minute.

2. Get onto the pages engines read

Find the reviews, comparison lists, forums and publications the answers cite for your topics, and earn a place on them: a correction where you are described wrongly, an inclusion where you are missing, original data a journalist can quote. This is the slowest work and usually the largest effect.

3. Answer the shortlist questions on your own site

Publish the pages engines search for: honest comparisons with named competitors, alternatives pages, pricing with its conditions, integrations, and who you are best for. Answer in the first lines, with specific numbers and dates, so a passage can be quoted on its own.

4. Make your facts the same everywhere

Engines cross-check. If your pricing, positioning or headcount differs between your site, your profiles and directories, they hedge or pick the cleaner competitor. Give each important entity one defining page and keep the rest consistent with it (why it matters).

5. Put the answer in the HTML

Several AI fetchers read the HTML the server sends and run no JavaScript. Content that appears only after scripts run may be invisible to them (the evidence).

6. Change one thing at a time, and judge it

Answers drift on their own, so a before-and-after chart flatters every change. Compare the topics you changed with the ones you left alone over the same weeks. If you cannot tell, say so, and do not scale the change yet.

Five myths worth dropping

• "llms.txt gets you into AI answers." Google says Search does not use it, and we know of no answer engine that documents using it for ranking (the details).
• "You need special AI markup." Google states there are no extra requirements to appear in AI Overviews or AI Mode beyond being indexed and eligible for a snippet.
• "Track your position in ChatGPT." Answers rarely repeat their order. Share across many questions is the stable measure.
• "One engine stands in for the rest." Engines read different sources and favour different brands. Measure each.
• "AI visibility shows up as traffic." Most answers get no click. Measure the answers themselves; treat AI referral traffic as a floor, not the total.