Category: AI Visibility

You can track hundreds of prompts and still learn very little about how buyers see your company. If the questions have little connection to the concerns people work through before choosing a provider, more tracking just gives you more data around the wrong thing.

A stronger approach starts with the buying decision itself. Look at what prospects ask, what makes them hesitate, what they compare, and what they need to feel confident moving forward, then turn those recurring concerns into prompts you can track consistently over time.

---

What Prompt Tracking Can Tell You

Prompt tracking is the everyday form of AI visibility tracking, and it shows you how often your company gets named when the same group of buyer questions is asked again and again, next to the other companies that come up.

One answer on its own doesn’t tell you much, because AI can return different recommendations even when you ask the same question more than once. SparkToro’s research found that repeated prompts returned different combinations and orders of brands, although certain companies still showed up much more often across the full set of results. 

That’s why prompt tracking is only useful when you look at repeated results over time. You are watching how consistently your company becomes part of the answer when buyers ask the questions that matter to their decision.

The Problem With Starting From Suggested Prompts

Suggested prompts can be a useful starting point, since they often reflect your services, your category, or the topics your website already covers.

The catch is that relevance to your business does not always mean relevance to the buyer, so the next step is to ask how closely a question reflects something a real prospect needs to work out before choosing a company.

That context is what makes a question worth tracking, since buyers may be comparing approaches, worrying about cost, wondering how well a provider can handle their situation, or trying to avoid making the wrong choice.

A prompt can therefore look perfectly relevant on paper and still miss the concern that made someone start looking for help in the first place, which is why suggested questions work best when you compare them with what prospects are already asking your business.

Your Best Questions Are Probably Already in the Business

Start with the conversations that happen around a purchase.

Useful places to look include

• Sales calls
• Questions from new prospects
• Objections before a deal closes
• Notes from won and lost opportunities
• Contact and enquiry forms
• Support or onboarding conversations that reveal earlier buying concerns

You are looking for uncertainty that keeps coming back. What does a prospect need to understand before moving forward? What makes them hesitate? What are they comparing? What are they afraid of getting wrong?

Those conversations give you the substance of the prompt, even if they do not tell you the exact words someone will type into AI. People with the same buying concern rarely phrase it the same way, so the decision behind the question matters more than any single wording.

The goal is to turn that recurring concern into a question you can keep testing consistently.

Not Every Buyer Question Is Worth Tracking

Once you start reviewing conversations, you may find far more questions than you need, so the next step is deciding which prompts to track. Give priority to the ones that come up repeatedly, reveal a real point of hesitation, or shape whether someone moves forward.

Then look for overlap, because different questions can point to the same concern. If several versions really focus on trust, price, or fit, you don’t need to track each one separately.

Build a smaller set around the questions that add something distinct, which gives you a clearer group to return to without filling your list with minor variations of the same issue.

Turn the Buying Decision Into the Prompt

Consider a broad category query such as

best IRS tax relief companies

It identifies the service, but it says very little about why the person is looking.

Now consider the same category through a buyer’s situation.

I owe the IRS $40,000 and got a letter about my bank account. Can a tax relief company stop that, or could I get scammed?

The person has a specific problem and a real financial consequence in mind, which makes the question much closer to a real buying decision.

You do not need to copy a sales call word for word. A more specific prompt gives you a cleaner test, since you are comparing answers to a realistic buying situation rather than a broad category label.

The same research shows why that situation matters. When people asked about the same headphone purchase in different words, AI tended to return brands from a fairly consistent group, but once the use case shifted to something like gaming, podcasting, or noise suppression, the recommendations changed a lot. The situation behind the question can shape which companies enter the answer.

Keep One Set Still While You Learn From Another

Once you have a useful group of questions, keep a core set the same so you have something reliable to compare over time. If you keep changing the questions, it becomes harder to tell whether your visibility changed or the test itself changed.

That does not mean the list can never evolve, as new objections, offers, or buyer concerns may give you good reasons to test new questions. Keep those separate at first, which lets you learn from them without losing the comparison you have already built.

Over time, some of those new questions may prove useful enough to join the core set, while others may turn out to add very little. 

See What AI Says About Your Real Buyer Questions

You do not need a long list to start your prompt tracking. A few questions tied to moments that matter commercially are enough, since those are the situations where being named keeps your company in the running and being absent hands the opening to someone else.

Begin with the questions you already hear from prospects, put a few of them into the Thallo AI Visibility Scan, and see whether your company shows up in the answers. Then keep the ones that reflect a real buying decision, return to them over time, and watch how the pattern shifts.

It’s a regular workday, and the head of operations at a growing logistics company has finally set aside an hour to find a new IT provider. Instead of opening ten browser tabs, she types her question into an AI assistant, adding a few details about her team size and budget. A few seconds later, the answer recommends one of your competitors, noting their experience with similar companies and pointing to their strong client reviews. By the time she closes her laptop, she already has a name to call, and your company never came up.

What makes this moment so easy to miss is that it leaves no trace. There’s no lost visit in your analytics and no dip in your sales reports because she never reached your site in the first place. That’s why more marketing teams are turning to AI competitor analysis, which shows them which companies AI recommends to their buyers, why it selects them, and what it would take to earn a place in those answers.

---

Start With the Questions Your Sales Team Hears Every Week

Choosing the questions is the step many teams rush through, even though it ends up shaping almost everything the analysis tells you. AI only responds to what it’s asked, and buyers tend to write to it the way they’d explain their situation to a consultant, with details about their industry, team size, or budget. Each of those details changes which companies come up, which is why broad questions usually return broad answers that say very little about your market.

Your sales team is the best source for better questions because they hear how buyers describe their problems long before a deal closes. Questions like “Do you work with companies our size?” or “How is this different from what we use now?” reveal the criteria buyers use to decide. 

Ten to fifteen questions are usually sufficient, provided they cover both the early stage (when buyers are still defining the problem) and the later stage (when they are comparing providers). Run each question a few times in more than one AI assistant, starting a new conversation each time. Answers shift from one session to the next, and the repeating patterns are the ones worth acting on.

Note Which Competitors AI Names and the Reasons It Gives

As you run each question, log the answers in a simple spreadsheet, noting the question, the companies AI recommends, and their order of appearance. After a dozen questions, the patterns become undeniable. Certain competitors appear in almost every answer, while some questions never include you at all.

The names are just the starting point; the reasons AI gives for each recommendation offer the most insight. It helps to copy those reasons word for word, as phrases like “specializes in healthcare clients,” “known for fast onboarding,” or “praised for transparent pricing” often echo language found in reviews, articles, and the competitor’s own website.

Once grouped, these reasons tend to fall into themes such as specialization, client results, price, or speed. Each theme reveals what buyers in your market value and which competitor has best associated their name with it. If AI consistently describes a competitor as the go-to option for mid-sized manufacturers, and you serve that same market, you know exactly which story you need to tell more clearly and in more places.

Apply the same attention to your own mentions. Appearing fifth with a vague description carries much less weight with a buyer than appearing first with a clear reason to choose you. Therefore, how AI talks about your company matters as much as whether it mentions you at all.

Check the Sources Behind Each Answer

Many AI assistants display the sources behind their answers, usually as links next to the text or listed at the end. Opening these links one by one is one of the most revealing parts of the process, as they show you where AI gathers its information about your market.

This is where competitor analysis for AI search differs most from traditional SEO. AI heavily relies on what others publish about a company, such as news articles, reviews, industry roundups, and comparison lists, giving much less weight to what a company says about itself. In practice, a mention in a trusted publication works for AI the same way a referral works for a buyer.

So, when a competitor repeatedly appears, examine where AI found them. They may have been featured in an industry publication, listed in a directory your buyers trust, or included in several “best providers” articles. Each of these sources is a place where your company could also appear, and together they form a clear map of where you need to be present.

Look for Questions Where No Brand Stands Out

Not every question has a clear winner, and these are some of the most valuable to find. As you review your answers, look for moments when AI seems unsure about who to recommend. This usually manifests as:

• Listing several companies without favoring any of them.
• Giving general advice without naming anyone.
• Recommending a different company each time you ask.

These open spots tend to appear in the most specific questions—those that mention a particular industry, location, or problem. Consider the difference between these two:

• “What’s the best accounting firm?”
• “Which accounting firm works with construction companies in Texas and handles payroll across several states?”

The first question likely has well-known names already, while the second may still be up for grabs. That’s why these gaps are some of the most valuable findings in any AI competitor analysis. Claiming a question that nobody owns yet is usually much faster than dislodging a competitor AI already trusts. There’s also a bonus: buyers who ask very specific questions are often closer to making a decision, meaning the leads they bring tend to be better.

Decide What to Fix First

By now, you have a much clearer picture than most companies ever achieve. You know which questions you’re losing, who’s winning them, why AI prefers them, and where it finds its information. The good news is that you don’t need to fix everything at once. In most cases, these three steps will move you forward the fastest:

Show up where AI is already looking. Revisit the publications, directories, and review sites that appeared most often in your analysis, and start there. Each time you get featured, listed, or reviewed in one of those places, AI gains another trusted reason to include you in its answers.

Give AI a clear answer to point to. For every question you’re losing, check whether there’s a page on your site that answers it directly. If buyers want to know whether you work with companies their size and nothing on your website says so, AI has very little to go on. Creating that page is often a quick and meaningful win.

Set the record straight. From time to time, AI will describe your services inaccurately, mention old pricing, or connect you with a market you’ve moved away from. Updating your website, directory profiles, and other public information gives AI the correct details, and those errors tend to fade over time.

After that, it’s worth making a habit of tracking competitors in AI search every few months. Answers change as other companies publish new content and earn new coverage, so checking in regularly helps you see what’s working, who’s gaining ground, and where to focus next.

See Who AI Recommends in Your Place

Doing this analysis by hand takes time, though the first step only takes a couple of minutes. Thallo’s free scan shows you how AI responds when buyers ask about your category and which competitors appear in your place. You can try the free scan here.

If you want the full picture, Thallo’s audit takes it further. It runs a complete AI competitor analysis across the questions your buyers ask, compares you with your main competitors, and provides a clear plan for what to fix first.