AI visibility study: 1,256 AI answers about one SaaS brand (2026)

AI visibility is how often AI search engines name your brand when buyers ask for advice.

We measured it for one SaaS brand, across 1,338 answers from six AI search engines.

Our verdict: getting cited isn’t enough. AI often reads your website, but that doesn’t put you on the shortlist.

The brand is Kaizo, a customer service quality platform that scores every support conversation, from human agents and AI agents alike (kaizo.com). Kaizo gave permission to publish its own numbers.

The five key findings:

1. On 218 buying questions without a brand name, AI named Kaizo in 17.4% of answers.
2. Zendesk QA showed up in 67.8% of the same answers, MaestroQA in 62.7%.
3. Perplexity cited kaizo.com in 36.8% of its answers, but named Kaizo in only 21.5%.
4. Vendor websites supply 46.8% of cited links. Review sites 6.3%, Reddit and YouTube together 1.2%.
5. Claude named Kaizo most often (22.0%), Google AI Mode least often (10.1%).

How we measured: 223 English-language questions, asked from the United States on October 20 and 21, 2026, to ChatGPT, Claude, Gemini, Perplexity, Google AI Overviews and Google AI Mode. The full method is at the bottom.

How often does AI name a brand when buyers ask for advice?

AI named Kaizo in 219 of the 1,256 answers to buying questions without a brand name: 17.4%.

That’s roughly one in six answers. It puts Kaizo in ninth place among the tools we tracked.

What counts as a buying question? Questions like “What are the best alternatives to Klaus (Zendesk QA) for support quality assurance?” or “Which tools can automatically score 100% of customer support conversations with AI?”.

Here’s what the ranking looks like:

Customer Impact measurement platform, 218 buying questions without a brand name, US, October 21, 2026. Kaizo highlighted.

The leaders get named in two out of three answers. Kaizo in one out of six.

What it means: in AI search, the category’s already been carved up. If you’re in the top three, you’re included by default in every comparison answer.

So doesn’t a smaller brand get found at all?

It does, but unevenly across the questions.

Kaizo was named at least once for 124 of the 218 questions, by at least one AI search engine. For the other 94 questions (43.1%), not a single engine named the brand.

The difference lies in the topic. On questions about QA software in general, AI named Kaizo in 22.8% of answers. On questions about scoring AI agents and chatbots, in 8.9%.

That’s striking, because that’s exactly where Kaizo positions itself: “QA for the agentic era”. Your positioning won’t show up in AI answers until the web repeats it.

Is being cited the same as being recommended?

No. In our study, AI search engines cited kaizo.com 82 times without naming Kaizo in the answer.

The reverse happened only 41 times: named, but not cited.

Perplexity shows the gap most clearly. It cited kaizo.com in 36.8% of its answers. But in the text itself, it named Kaizo in only 21.5%.

All 223 questions, including branded questions. AI Overviews: only the 170 questions where Google showed one.

Why does that happen? Perplexity searches the web live and attaches numbered sources to every answer, as Perplexity itself explains.

So your page can supply a fact, like a definition or a price, without your brand making it into the answer.

Rule of thumb: always track two numbers separately. How often you’re cited, and how often you’re named. Our guide explains how to measure AI citations.

Which sources do AI search engines cite for buying questions?

Mostly the vendors’ own websites: 46.8% of the 8,610 cited links.

That cuts against popular advice. Reddit and YouTube together supplied just 1.2% of citations.

Customer Impact measurement platform, 8,610 cited links in 1,285 answers, US, October 20 and 21, 2026.

The most cited domains were intryc.com (599 times), kaizo.com (448), zendesk.com (405) and lorikeetcx.ai (394).

G2 was the only review site in the top 10, with 223 citations on g2.com and another 148 on learn.g2.com.

What it means: in a software category, AI search engines read the players’ own comparison pages, pricing pages and blogs first. If you don’t have those pages, you’re leaving the story to your competitors.

Review sites still play a part. Our guide on G2 and Capterra reviews for AI visibility shows how to put that channel to work.

How many sources does each engine use?

It varies by up to a factor of three:

• ChatGPT and Perplexity: 10 sources per answer on average.
• Gemini: 6.2 sources.
• AI Overviews: 5.8 sources.
• Claude: 5.0 sources.
• AI Mode: 2.9 sources.

The fewer the sources, the smaller the chance your page is among them. That fits what we saw: AI Mode named Kaizo least often.

Does AI visibility differ by AI search engine?

Yes, a lot. Claude named Kaizo in 22.0% of answers to buying questions without a brand name, Google AI Mode in 10.1%.

Customer Impact measurement platform, 218 buying questions without a brand name, US, October 21, 2026.

Position in the answer differs too. When ChatGPT named Kaizo, it sat at position 2.5 on average. With Claude, at position 4.7.

Claude and Perplexity simply name more tools per answer: 6.0 and 5.7 on average. ChatGPT and AI Mode name 4.0 and 3.7.

For AI Overviews and AI Mode, Google uses a technique called “query fan-out”: several searches across subtopics at once, according to Google Search Central. Read more about query fan-out and intent.

What it means: measuring one engine isn’t enough. A cross-model analysis shows where you win and where you disappear.

Does Google always show an AI Overview for buying questions?

No. Google showed an AI Overview for 170 of the 223 questions: 76.2%.

For the other questions, the buyer saw regular search results. There, your classic SEO ranking still counts in full. Here’s how to get a grip on that: getting found in Google AI Overviews.

The other five engines answered every question.

Why do almost all AI answers name a shortlist?

Because the question asks for one. Between 86.5% and 92.8% of answers named at least one of the 23 tracked vendors.

A buyer who asks “which tool is best for X?” doesn’t get an explanation. They get a list.

All answered questions. “Tracked tools”: Kaizo plus 22 competitors in our measurement platform, other names not counted.

What it means: for B2B software, the AI answer is the new longlist. If you’re not on it, the buyer starts their comparison without you.

That’s why it pays to know which questions your customers ask ChatGPT and whether your competitor gets named and you don’t.

What can you do with these numbers for your own brand?

Measure first, then pick your questions, and only then build the pages AI search engines read.

Four steps that follow from the data:

1. Track named and cited separately. A high citation score without mentions means AI uses your facts but doesn’t recommend your brand.
2. Find your gap questions. For Kaizo, that was 43.1% of buying questions. That’s where your first new pages belong.
3. Write the comparison yourself. Vendors supply almost half of all sources. A fair comparison or alternatives page is a source AI reads.
4. Repeat your positioning on other sites. Your own claim alone won’t cut it. Reviews, partners and media have to carry it too.

How to build a page like that is covered in our guide on citable content for AI.

Compare your own numbers with our AI visibility benchmark by industry and work out your share with the method for AI share of voice.

How did we run this study?

We used our own measurement platform, which we use to track brands’ AI visibility every week.

Questions: 223 English-language questions in the customer service quality software category. 218 contained no brand name. 5 did name Kaizo, such as “Kaizo vs MaestroQA”. We don’t count those last ones in the numbers on unbranded buying questions.

Engines and models:

• ChatGPT: gpt-6 via the API.
• Claude: claude-haiku-5-5 via the API.
• Gemini: 3.5 Flash-Lite via the API.
• Perplexity: sonar via the API.
• Google AI Overviews and AI Mode: the US search results.

Period and market: 8 questions on October 20 and 215 on October 21, 2026, asked from the United States.

What we counted: whether an answer names Kaizo, in which position, whether kaizo.com appears among the sources, and which of 22 preselected competitors get named.

Limitations: this is one software category, one market and one point in time. API models can answer differently from the chat apps people use. And AI answers vary from run to run, so read the numbers as a snapshot, not a fixed share.

Frequently asked questions

What is AI visibility?

AI visibility is how often AI search engines like ChatGPT, Perplexity and Google AI Overviews name or cite your brand when someone asks a question. You measure it with a fixed set of buyer questions, repeated over time.

How do you measure AI visibility?

Draw up a fixed list of buyer questions, put them to several AI search engines and count how often your brand is named and cited. Our guide on measuring AI visibility without tools shows how to do it by hand.

What’s a good AI visibility score?

It depends on your category. In this study, the leaders reached 63% to 68% of answers, the ninth brand 17.4%. So always compare yourself with your direct competitors, not with a generic number.

Does Reddit help you get named by AI?

Barely, in this study: Reddit and YouTube together supplied 1.2% of cited links. For B2B software, vendor websites and G2 carried far more weight.

Can you run this study for my own brand?

Yes. Our GEO checker gives you a free first measurement. If you’d like the full study on your own questions, take a look at our approach as a GEO agency.

Want to know how visible your brand is in AI?

We measure your brand on the questions your buyers actually ask, across six AI search engines, and show you which pages to build first.

Book an intro call and we’ll walk through your first measurement together.