AI visibility is how often AI assistants such as ChatGPT, Gemini and Perplexity can find your business, describe it accurately and name it when buyers ask questions in your category. Being found in search means a page of yours can appear in a list. Being visible to AI means the answer your buyer actually reads includes you.
Imagine asking an assistant who does the work you do, and watching a familiar competitor appear. Again. Your work didn't get worse this month. You weren't rejected. You were omitted.
Omission has a mechanism, and the mechanism is fixable. Between a question and a named answer, the assistant finds you, reads you and names you. Below: how each step works, why the same competitor keeps showing up, and how to measure where you stand. The basics and a ten-minute self-check are in our guide to what AI visibility is.
Found, Read, Named: What Happens Between a Search and an Answer
Found is the technical floor. Google's documentation for its AI features says that to be shown as a supporting link in AI Overviews or AI Mode, "a page must be indexed and eligible to be shown in Google Search with a snippet." OpenAI's crawler documentation sets a similar floor for ChatGPT: sites that opt out of its search crawler, OAI-SearchBot, "will not be shown in ChatGPT search answers." If a security setting at your host blocks that crawler, you can be absent without anyone deciding anything.
You can check this layer yourself in an afternoon, or ask whoever runs your website to. Confirm your important pages are indexed in Google Search Console. Read your robots.txt file and your hosting or CDN bot settings for OAI-SearchBot by name. Make sure the content that explains what you do is on the page as text, and not locked inside an image or a PDF. None of it is glamorous, and all of it sits underneath everything else.
Read is where most of the action is. Google says AI Overviews and AI Mode "may use a 'query fan-out' technique," issuing multiple related searches across subtopics and data sources to build one response. Your buyer types one question. The assistant runs several. Each of those searches looks for something specific: who you are, what you do, where, for whom, and what others say about you.
That changes what "being found" has to mean. A single strong page that ranks for a single phrase covers one of those searches. The trail that wins covers many of them, consistently.
Named is the last step, and the one buyers see. The assistant writes an answer with a few names in it, or none. A business can clear the first two steps and still not be named, because being described is a lower bar than being recommended. Our piece on why AI can cite your content and still recommend someone else takes that gap apart in detail.
Why Being Found Is No Longer the Same as Being Chosen
For twenty years, being found was most of the job. A buyer saw a list, clicked a few links, and decided. The answer layer changes the shape of that moment.
Pew Research Center tracked the browsing of 900 US adults in March 2025. When a Google search showed an AI summary, people clicked a traditional search result in 8% of visits. Without a summary, they clicked in 15%. Links inside the summary itself were clicked in just 1% of visits.
The same study found something every specialist should notice. Longer searches were far more likely to produce an AI summary. Only 8% of one- or two-word searches did, against 53% of searches of 10 words or more. And 18% of all searches in the study produced one.
Your buyers rarely ask short questions about hard problems. They ask long, situational ones: the circumstances, the constraint, the fear. Those are exactly the searches where an answer now sits above the list, and where fewer people scroll past it.
Try this with your own work. Write down the long version of the question your best client asked before they first called you, in their words, with the circumstance included. Then ask it of two or three assistants. What comes back is a fair picture of the read and named layers for the buyer you most want.
So a ranking still matters. It is how you get found, and it feeds the read layer, because the assistant's searches still draw on the same index. It just no longer guarantees you get read by the human, because the human is increasingly reading the machine's summary of everyone.
Why the Familiar Competitor Keeps Appearing
When a competitor's name comes back and yours does not, it feels like a ranking of merit. It is a ranking of what the machine can verify.
AI can't feel your reputation in a room. It can only work with what it can verify. Twenty years of referrals, client loyalty and hard-won judgment do not reach an assistant unless they are written down somewhere it can read, in words that match everywhere it looks.
The competitor who keeps appearing usually has a clearer trail: specific pages, repeated facts, outside sources saying the same thing about them.
Start your own trail with the facts a machine checks first, and make them identical everywhere they appear: your business name, the services you actually offer, the people who deliver them, the places you serve, and the credentials you hold. Your website, your Business Profile, your directory listings and your social profiles should all say the same thing in the same words. Small mismatches cost more than they look, because the assistant cannot tell which version is true.
There's a clue in how citations behave over time. SISTRIX tracked 82,619 prompts for 17 weeks to April 2026, in six countries including the United States. Sources rotated fast: Google's AI Mode replaced 56% of cited sources every week, and ChatGPT Search 74%. Yet for 43% of brand queries, the brand's own domain stayed cited in all 17 weeks, while the sites cited alongside it rotated at about 70% a week.
That finding is about brand queries, not "who should I hire," so read it as direction. The direction is still useful. A clear, consistent record of your own tends to stay put while the noise around it churns. A scattered record has nothing to anchor it.
What Makes a Business Visible, and What Makes It Recommendable
Visible is the floor. Recommendable is the record standing on it.
AI recommends businesses it can verify (foundation), that others vouch for (validation), and that know something it doesn't (information gain). We build all three. Each one answers a different search inside the fan-out:
- Foundation answers "who is this, exactly?" Consistent names, services, people and locations, with structured data that matches what your pages say.
- Validation answers "does anyone else agree?" Independent sources that confirm your claims, so the machine isn't relying on your homepage alone.
- Information gain answers "what do they know that others don't?" Your real judgment, written down, so the assistant has something specific to say about you.
Plenty of established firms are strong on one and thin on the others, for a historical reason. Each of the three was historically a different industry's job: web teams owned the foundation, PR owned validation, and the expertise lived in the founder's head. Nobody was asked to connect them, so for most firms nobody did. Our piece on what AI needs before it recommends a business works through each layer.
Measurement shows the gap. Operations close it. Knowing which of the three is thin is the first useful thing a reading gives you.
How AI Visibility Is Measured Without a Fake Rank
Because answers vary so much, AI visibility only makes sense as a rate, read over time. SparkToro's January 2026 study ran 12 prompts through ChatGPT, Claude and Google's AI 2,961 times and found under a 1 in 100 chance of getting the same list of brands twice for ChatGPT and Google's AI, with Claude only slightly more likely to repeat. The study was run with an AI tracking vendor and isn't peer reviewed. Its conclusion still holds up: how often a brand appears across many runs is a reasonable measure, and a single "rank in ChatGPT" is not.
So a useful measurement looks like this. A set of real buyer questions, asked in your buyers' own words. Several engines, because buyers use several. Many runs, with the answers saved so you can compare them later. Then three numbers per question: how often you were named, how accurately you were described, and who was named instead. Repeat the same set on a schedule, monthly is a sensible rhythm, so that a change in the numbers means something changed in the world rather than in the questions.
At Probably Genius, measuring this rate is where every engagement starts. We test the five engines your buyers actually ask: ChatGPT, Gemini, Perplexity, Claude and Grok, with the questions they actually ask. The free Recommendation Check adds 109 specific checkpoints across eight zones, hand-scored, which score both your own site and what the engines answer. Our piece on measuring whether AI visibility work is working lists the questions to ask any provider who reports a score.
If you decide to close the gaps rather than just measure them, The Answer, our plan for competitive, winnable markets, is $2,500 setup + $1,500/month, twelve-month program, and the full ladder is on the pricing page. Start with the reading. The method behind it is public in our methodologies.
You're probably a genius at what you do. We make sure AI gets the memo.