Is your brand visible in AI search results? The AI visibility check

The short answer

An AI visibility check takes about fifteen minutes and gives you one of four answers. That is the whole exercise.

Ask five questions your buyers ask, in three engines. Record two things each time: whether the answer named you, and whether it cited a page you own.

Those two columns produce four outcomes, and each one is a different problem. Most people skip the second column, which is why most brands cannot tell which of the four they have.

Key takeaways

• Named and cited are separate. Across 3,981 measured appearances, only 13.2 percent were both.
• Being used without credit is the normal case. 61.7 percent were citations with no mention of the brand.
• One blended score hides all of this. It cannot tell an awareness problem from a content one.
• Engines disagree sharply. ChatGPT cites far more than it names. Gemini does the reverse.
• Fifteen minutes is enough to triage. Measuring properly comes later, and only if the check says you need it.

Where this comes from. We run organic for more than 250 enterprises at Pepper and track over 10 million prompts across every major engine. This is the first thing we run on a new account, before anything is quoted. That work shapes the view here, along with the client reviews we sit in and the talks we have at the events we run.

What is an AI visibility check?

An AI visibility check is a short manual test that tells you whether AI engines name your brand and cite your pages when buyers ask questions in your category.

It is deliberately smaller than a measurement programme.

What it is not

• It is not tracking. You run it once to find out whether you have a problem. Tracking comes after, and only if you do.
• It is not a tool. Free checkers run prompts you did not choose, which measures their sample rather than your buyers.
• It is not a score. The output is a diagnosis, which is more useful than a number out of 100.

The full measurement method, for when you are ready to track properly, sits in how to measure AI search visibility without expensive tools.

See where you show up. Pepper’s GEO platform tracks Brand Visibility, Domain Prompt Presence and Share of Voice across ChatGPT, Perplexity, Gemini, Claude and Google AI Overviews. Your team can log in, connect Search Console and GA4, manage the prompt set and run your own agents in the Agent Atlas. A growth team works the same account alongside you. Book a growth audit or see where you show up.

The check, in fifteen minutes

Four steps. You need a browser and a notepad.

1. Write five real questions

Not keywords. Questions, phrased the way a buyer would type them.

Good ones sound like “best data observability tool for a small team”, “how much does sales enablement software cost”, “is there a free alternative to the market leader” and “which vendor suits a regulated industry”.

Avoid your brand name. You are testing whether you come up unprompted.

2. Run them in three engines

Google, ChatGPT and Perplexity covers most buyer behaviour. Add Gemini if your audience leans Google.

Use a fresh browser session so personalisation does not flatter the result.

3. Record two columns, not one

For each question, write down two things.

• Named. Did the answer mention your brand by name?
• Cited. Did it link a page on your domain as a source?

Keeping these apart is the entire point of the exercise. A single “did I appear” column throws away the diagnosis.

4. Note who appeared instead

In the same pass, write down the sources the answer did use. That list becomes your target list later, and it costs nothing to collect now.

Reading the result

Your two columns give four outcomes. Each is a different problem with a different fix.

Named and cited

The engine says your name and links your page. This is the target state and it is rarer than people assume.

What to do: keep going, and start tracking properly so you can see it move.

Cited but not named

The engine uses your page and never says who you are. This is the most common result by a wide margin.

What to do: this is a brand problem, not a content one. More publishing will not fix it. Work on third-party coverage, covered in earned authority.

Named but not cited

The engine mentions you and links somebody else. Your reputation is working and your pages are not.

What to do: a content problem, and the cheaper of the two to fix. Answer in the first two sentences under each heading, and make sections liftable. The evidence is in our analysis of 1.4 million prompts.

Neither

You do not appear at all. Before concluding anything, check that engines can reach you.

What to do: run the technical check first, because a blocked crawler produces this exact result and is free to fix. The method sits in our crawler and rendering audit, and the crawler reference lists the bots by name.

Why two columns rather than one

Because the gap between them is large, and it is where the diagnosis lives.

Semrush and Kevin Indig measured 3,981 domain appearances across 115 prompts, 14 countries and four engines, published June 2026.

• 61.7 percent were citations with no brand mention.
• 25.1 percent were mentions with no citation.
• 13.2 percent were both.

So the common case is being used without credit. A single column asking “did we appear” records that as a win, when it is actually a brand problem with a slow and expensive fix.

The engines pull in opposite directions

ChatGPT cited at 87 percent and named at 20.7 percent. Gemini named at 83.7 percent and cited at 21.4 percent.

Run your check in one engine only and you will reach a confident conclusion about the wrong half.

Read the limits. The study does not state a data collection window, and 115 prompts is a small set. We use the proportions as directional, and the engine comparison as the reliable part, since it holds within one consistent sample.

Our methodology: how we weight the check

Ways to check at a glance

How to choose what to do next

The result of the check decides this, which is the point of running it.

The scorecard

Score your check out of 100 before trusting its conclusion.

The live test, over 90 days

If the check says you have a problem, upgrade it to a proper reading.

Fix 30 prompts in your buyers’ words and hold them steady. Real ones, such as “best data observability tool for a small team”, “how much does sales enablement software cost”, “is my brand visible in AI search” and “which vendor works for a regulated industry”.

Run them across Google, ChatGPT and Perplexity monthly for 90 days. Three readings is the shortest honest window. One is a snapshot, two could be noise, three shows direction. Keep the set fixed, because changing it resets the comparison.

Keep the same two columns. The trend that matters is whether the gap between named and cited is closing.

One thing to watch while you do it. Read the gap per engine rather than pooled, because the two largest pull in opposite directions and an average across them describes neither. A brand that looks balanced overall is often badly short on one surface and perfectly fine on another, which sends the budget to the wrong half entirely. Brand visibility covers what each number means once you are tracking it.

Weak approach versus strong approach

The weaker approach: paste your domain into a free checker, read a score out of 100, screenshot it for the deck. Those prompts were chosen by the tool, so the number describes its sample rather than your category, and it cannot tell you which problem you have.

The stronger approach: five questions you wrote, three engines, two columns, fifteen minutes.

One produces a number. The other produces a diagnosis, and only the second tells you what to fund.

Red flags

• A visibility score with no prompt list. If you cannot see the questions, the number describes somebody else’s sample.
• One column for “appeared”. It records a ghost citation as a success.
• A single engine checked. ChatGPT and Gemini disagree by roughly sixty points on naming.
• Your brand name inside the questions. That tests recall, not discovery.
• An industry benchmark quoted as a target. Nobody publishes a comparable one.
• Guaranteed visibility. Nobody controls generated output, and Google warns against guaranteed rankings.

Five questions worth asking

1. “Who wrote these prompts?” If a tool did, it is measuring itself.
2. “Are named and cited recorded separately?” One column loses the diagnosis.
3. “How many engines did we check?” One is not enough here.
4. “Which sources appeared instead of us?” Free to collect, and it becomes the plan.
5. “What does this result tell us to do?” If the answer is unclear, the check was the wrong shape.

Reduced to one principle: record named and cited apart, or the check tells you nothing.

One closing note that costs us something. This check is free, takes fifteen minutes, and most teams can run it without us, ours included in the general point that a diagnosis is the easiest thing in this category to sell and the cheapest to do yourself. Pay for the fixing.

What does this cost?

The first row is free and it decides whether any of the rows below apply to you.

What nobody should promise you

A meaningful industry benchmark for mention or citation rate. Nobody publishes a comparable one.

That one score captures AI visibility. The two halves move apart, often in opposite directions on different engines.

Guaranteed visibility from any programme. Nobody controls generated output, and Google warns against providers guaranteeing rankings.

Historical data from before you started checking. There is no backfill in this category, which is the honest argument for running the check today.

When the result is good enough

The answer that costs us the sale. If you ran the check and you were named and cited on most of your five questions, you do not need an AI visibility programme and we would tell you to spend the budget elsewhere.

That outcome is uncommon, since the measured rate for both together is about one in eight. It does happen though, usually to brands with strong earned coverage already. Recheck in six months rather than buying something now, because this moves fast enough that a clean result today is not a permanent one.

Frequently asked questions

How do I check if my brand is visible in AI search?

Write five questions your buyers would ask, without your brand name in them. Run each in Google, ChatGPT and Perplexity, then record two things separately: whether the answer named you, and whether it cited a page you own.

Why record named and cited separately?

Because they are different problems. Across 3,981 measured appearances, 61.7 percent were citations with no brand mention and 25.1 percent were mentions with no citation. Only 13.2 percent were both.

What does it mean if I am cited but not named?

The engine trusts your page and not your name. That is a brand problem rather than a content one, and it is fixed by third parties describing you accurately rather than by publishing more yourself.

What does it mean if I am named but not cited?

Your reputation is working and your pages are not being quoted. That is a content problem, and it is the cheaper of the two to fix by moving answers to the top of each section.

What if I do not appear at all?

Check the technical floor before concluding anything. A blocked crawler or a page that only renders after JavaScript produces exactly this result, and both are free to find and fix.

Should I use a free AI visibility checker instead?

For a first glance, fine. As a diagnosis, weak. Those tools run prompts you did not choose, so the result describes their sample rather than the questions your buyers actually ask.

How many engines do I need to check?

At least two, and ideally three. ChatGPT cited at 87 percent while naming at 20.7 percent, and Gemini did almost the reverse, so a single engine will give you a confident answer about the wrong half.

How often should I run this check?

Once to triage. If it finds a problem, upgrade to thirty fixed prompts run monthly for ninety days, because three readings is the shortest window that shows direction rather than noise.

Where to go next

Write five questions your buyers ask, run them in three engines, and keep two columns.

Fifteen minutes from now you will know which of four problems you have, and that is a better starting point than any score out of 100.

Measure it properly · Run the technical audit · Book a growth audit

Sources and further reading

Primary studies

• Semrush and Kevin Indig. “The Ghost Citations Study”, published 9 June 2026, run with Growth Memo. 3,981 domain appearances across 115 prompts, 14 countries and four engines. 61.7 percent citations without a mention, 25.1 percent mentions without a citation, 13.2 percent both. ChatGPT cited at 87 percent and mentioned at 20.7 percent, while Gemini mentioned at 83.7 percent and cited at 21.4 percent. The authors do not state a data collection window, and 115 prompts is a small set, so we treat the proportions as directional. Link
• Ahrefs. Study of 1.4 million ChatGPT prompts. 88.46 percent of citations came from pages retrieved through the search channel, which is why a technical check comes first when you appear nowhere. Link

Further reading

• Google Search Central. AI search optimisation guidance, 15 May 2026. Warns against providers guaranteeing rankings.
• Pepper. How to measure AI search visibility without expensive tools for the ongoing method, the free AI visibility audit for the technical floor, and AI trust signals for why the two columns diverge.
• Deliberately excluded: any industry benchmark for a good mention or citation rate. Nobody publishes a comparable one, the measured rates vary enormously by engine, and a number without a matching denominator would mislead more than it helped.