How to Track AI Visibility: A Plain-English Guide

More buyers now ask ChatGPT, Perplexity, Gemini or Claude what to buy. So a fair question is: when they ask, does your product come up?

That's what "AI visibility" means. This guide shows you how to measure it in a way that holds up, how to do it for free, and how the paid tools compare.

Want a starting point? The free MadeTheAnswer check asks ChatGPT, Perplexity, Gemini and Claude 5 questions like the ones your buyers ask, one run each, and shows whether your product gets named. We work out your category and suggest likely competitors. Your score shows on screen right away, and the full report comes by email. No card. It's a one-run snapshot, not a verdict, since AI answers change from run to run.

Why AI visibility is hard to measure

With Google, you can look up where your page ranks. AI answers don't work like that.

In January 2026, SparkToro and Gumshoe.ai had 600 volunteers run 12 prompts 2,961 times across ChatGPT, Claude and Google's AI tools. There was less than a 1 in 100 chance of getting the same list of brands twice, and about a 1 in 1,000 chance of getting the same list in the same order. (SparkToro) (Gumshoe sells an AI tracking tool, which the post discloses.)

The study also found that real people's prompts for the same need were very different from each other. They scored just 0.081 on a 0-to-1 similarity scale.

Two big lessons come from this:

1. Don't track "rank" in AI answers. The order changes almost every time. SparkToro's Rand Fishkin called ranking-position claims "full of baloney."
2. Do track how often you show up. SparkToro found that the percentage of runs where a brand appears is a reasonable thing to measure.

The metrics that matter

Here are the numbers worth tracking, in order of importance.

1. Visibility rate (the main one)

This is how often you're named, out of all your tries. For example: "We were named in 9 of 30 answers this month." That's 30%.

It's simple, and it holds up even though answers change from run to run.

2. Who gets named instead

List the competitors that show up when you don't. This is often called "share of voice." It tells you who you're really up against in AI answers. That may not be who you expect.

3. Cited sources

Many AI answers link to the pages they used. Write these down. They tell you why a brand got picked.

This matters a lot. Onely, an SEO agency, studied 31 "best agency" questions. In ChatGPT's answers, 57% of cited pages were the brands' own sites and 22% were "best of" lists. In Perplexity, Gemini and Google AI Mode, lists made up 59% to 77% of citations. (Onely) That was one agency's study on one type of question, so check your own category.

4. Accuracy and tone

Is the AI describing you correctly? Is it positive, neutral or negative? AgencyAnalytics lists the same four core metrics: visibility, position, sentiment and citations. (AgencyAnalytics) We'd treat "position" with care, for the reason above.

Step 1: Build your prompt set

Your prompt set is the list of questions you'll ask every time. It's the most important part of tracking.

• Use buyer questions, not your brand name. "Best CRM for a 5-person sales team" tells you far more than "What is [brand]?" Both Onely and AgencyAnalytics recommend unbranded questions.
• Cover the buying journey. Mix "what is the best…", "[competitor] alternatives", and "[tool A] vs [tool B]" questions.
• Group them by topic. This makes patterns easier to see.
• Pick a size you can keep up with. AgencyAnalytics says agencies often use 25 to 50 prompts per client. If you're doing it by hand, 10 to 20 is a fine start.
• Freeze the list. Once you start, don't keep changing the questions. AgencyAnalytics recommends locking the prompt set, the AI tools and the location so your results stay comparable. If you must add questions, add them as a new group.

Step 2: Run each prompt more than once

Because answers change so much, one run per question isn't enough.

• Ask each question at least three times.
• Spread runs over different days if you can. Onely suggests testing several times over days.
• Use a fresh, logged-out session or temporary chat. Memory and past chats can change what you see.
• Track each AI tool on its own. ChatGPT, Perplexity, Gemini and Claude often disagree. AgencyAnalytics also recommends reporting by engine.

Step 3: Log the results the same way every time

A simple spreadsheet works. Use one row per answer, with these columns:

Then add a summary tab that shows, for each AI tool:

• Visibility rate this month vs last month
• Top competitors named
• Pages cited most often

That's it. It's slow, but it costs nothing, and you'll learn a lot by reading the answers yourself.

Step 4: Turn results into action

Tracking is only useful if it leads to changes.

• If the same list articles keep getting cited, check whether you're on them and whether your facts are right.
• If the AI cites your own site but says something wrong, fix that page.
• If you never show up in ChatGPT search answers, check that your robots.txt allows OAI-SearchBot. OpenAI says sites that opt out of it won't be shown in ChatGPT search answers. Changes can take about 24 hours. (OpenAI)
• If a competitor wins on "vs" and "alternatives" questions, write your own honest comparison pages.

Then keep running the same prompts and watch the trend.

Tools that can do this for you

Doing it by hand gets slow once you have many prompts, several AI tools and more than one brand. Here's a fair look at some options.

Please read this first: prices change often. Every price below was read on that vendor's own pricing page on Oct 31, 2026. If a vendor's own site doesn't show a price, we say so rather than use someone else's number. Always check the vendor's site before you buy.

Prices as of Oct 31, 2026, from each vendor's own pricing page. Entry plan shown first.

How to choose

• Just want to know where you stand? Start with a free check or the DIY spreadsheet. Remember that one run is a snapshot, not a trend.
• Need daily data or lots of prompts? Look at tools with higher prompt limits and daily refreshes.
• Need Claude or Gemini? Check which engines are included. Some tools charge extra for them.
• Running an agency? Look for multi-client setups and reports you can share with clients. (MadeTheAnswer's Agency plan is $199/month for 120 questions across up to 10 client sites.)
• Want to check the numbers yourself? Ask whether the tool shows the real AI answers behind each number, and how many runs it used.
• Be careful with any tool that sells "AI rank." Given how much answers change, visibility rate over many runs is more honest.

How agencies can report AI visibility to clients

Clients are asking about this. AgencyAnalytics surveyed 494 agency respondents. 66% said clients ask for AEO (answer engine optimization) work. 48% said AI discovery is their biggest challenge in showing results. (AgencyAnalytics)

A simple monthly report could include:

1. Visibility rate by AI tool, this month vs last month
2. Top competitors named, and any changes
3. Most-cited sources, plus any new ones the client now appears on
4. Accuracy issues found and fixed
5. Work done this month and what's planned next

Keep the same prompt set from month to month. Explain that AI answers vary, so you're showing trends, not exact ranks. Don't promise placements. Nobody controls these answers.

Show clients the proof. Clients may ask, "Is this real?" Link each number to the actual AI answers, and say how many times you asked. That builds more trust than a score alone.

MadeTheAnswer's Agency plan is built for this. It covers 120 questions across up to 10 client sites, checked weekly, with client-ready reports you can share by link or PDF. You can also start with the free check on a client's site. We work out your category and suggest likely competitors. The report shows whether AI names you, which competitors get mentioned, and which sites it cites, and each assistant's score links to the actual AI answers. See a sample report → (demo data)

Frequently asked questions

What is AI visibility?

It's how often AI tools like ChatGPT, Perplexity, Gemini and Claude mention or recommend your brand when people ask questions related to what you sell.

Can I track my "ranking" in ChatGPT?

Not in a reliable way. SparkToro's research found the list of brands, and their order, changes almost every time. Track how often you're named instead.

How many prompts should I track?

Start with 10 to 20 if you're doing it by hand. AgencyAnalytics says agencies often track 25 to 50 per client. What matters most is that you use the same set every time.

How often should I check?

Run each prompt several times, ideally across different days. Then compare results weekly or monthly. AgencyAnalytics suggests reporting to clients monthly.

Is checking by hand good enough?

Yes, to start. It's free and you learn a lot by reading the answers. It gets slow as your prompt list grows, which is when a tool starts to save time.

Why do different AI tools give different answers?

They work differently and lean on different sources. Onely found ChatGPT cited brand websites the most, while Perplexity, Gemini and Google AI Mode cited list articles much more. Track each one on its own.

What does MadeTheAnswer check?

The free check asks ChatGPT, Perplexity, Gemini and Claude 5 buyer-style questions in your category, one run each, and shows whether your product gets named. We work out your category and suggest likely competitors. Your score shows on screen right away, and the full report comes by email. No card. It's a one-run snapshot, so use it as a starting point. You get one free check per website and one per email each week. There are also 5 free checks a day for everyone. If today's are full, come back after the daily reset at midnight UTC (5 PM Arizona time).

The report shows whether AI names you, which competitors get mentioned, and which sites it cites. Each assistant's score links to the actual AI answers. Paid weekly plans: Starter ($29/month, 20 questions), Growth ($79/month, 50 questions) and Agency ($199/month, 120 questions across up to 10 client sites). Every paid plan covers ChatGPT, Perplexity, Gemini and Claude, and asks each question 3 times, so you see a trend over time. A new report comes every Monday about 8 AM in the time zone set in the dashboard. Your first report starts right after checkout once your site is added. See plans and prices → · View the sample report → (demo data)

Related guides

• How to Report AI Visibility to Clients — How to turn those checks into a report a client can follow.
• Otterly AI Alternatives: What to Pick, and When — A side-by-side look at tracking tools, using each vendor's prices.
• Is My Brand in ChatGPT? How to Check, and What to Do if You're Missing — A free, by-hand way to check one assistant first.
• Why Did My ChatGPT Traffic Drop? — What to check when the numbers move from one week to the next.

Get your baseline

Every good tracking plan starts with knowing where you are today.

Check your AI visibility for free. The free MadeTheAnswer check asks ChatGPT, Perplexity, Gemini and Claude 5 questions your buyers might ask, one run each, and shows whether your product comes up. We work out your category and suggest likely competitors. Your score shows on screen right away, and the full report comes by email. No card. No ranking promises, just a one-run starting point you can measure from.

---

Sources

• SparkToro (Rand Fishkin), "New Research: AIs are highly inconsistent when recommending brands or products," Jan 28, 2026. Co-researched with Gumshoe.ai. https://sparktoro.com/blog/new-research-ais-are-highly-inconsistent-when-recommending-brands-or-products-marketers-should-take-care-when-tracking-ai-visibility/
• Onely (Bartosz Góralewicz), "How ChatGPT decides which brands to recommend," Oct 16, 2026. Agency research. https://www.onely.com/blog/how-chatgpt-decides-which-brands-to-recommend/
• AgencyAnalytics (Francois Marchand), "AI search visibility," Sep 14, 2026. https://agencyanalytics.com/blog/ai-search-visibility
• OpenAI, "Overview of OpenAI Crawlers." https://platform.openai.com/docs/bots
• Vendor pricing pages, each read again Nov 9, 2026 (Otterly.AI Nov 10, 2026): Otterly.AI https://otterly.ai/pricing · LLMrefs https://llmrefs.com/pricing · Omnia https://www.useomnia.com/pricing · Peec AI https://peec.ai/pricing · Profound https://www.tryprofound.com/pricing · Rankscale https://rankscale.ai/pricing · SE Visible https://visible.seranking.com/pricing/ · SISTRIX https://www.sistrix.com/prices/ · Semrush https://www.semrush.com/pricing/seo-ai-search/ · Scrunch https://scrunch.com/pricing · BrightEdge https://www.brightedge.com/ · AgencyAnalytics https://agencyanalytics.com/pricing