How to create tracked prompts that actually measure your AI visibility

Good tracking starts with good prompts: the questions your buyers ask, not the ones your site answers.

Measuring AI visibilityBy Tilio10 min readLast updated 19 October 2026

In this post

This guide walks through what tracked prompts are, the questions worth asking before you write any, and a step-by-step way to build a set you can trust. There’s a free prompt set generator near the end if you’d like a head start.

What is a tracked prompt?

A tracked prompt is a question, phrased the way a real person would ask it, that you run on a recurring schedule across AI answer engines such as ChatGPT, Claude, Perplexity, Gemini, Copilot and Google AI Overviews. Each time it runs, you record whether your brand appears, how it’s described, which sources are cited, and how you compare with competitors.

It’s the closest thing AI search has to a keyword rank tracker, with one big difference: there’s no fixed results page. The answer is generated fresh each time, varies between engines, and can change from one day to the next. That’s exactly why tracking on a schedule matters.

Why tracked prompts matter

Your prompt set is the denominator for every AI visibility metric you’ll ever report. Appearance rate, share of voice, citation counts, competitor benchmarks: all of them are calculated against the prompts you chose. If those prompts don’t reflect what buyers ask, every number built on them will point you in the wrong direction.

Three more reasons to give this proper thought:

AI answers vary. Ask ChatGPT the same question twice and you can get two different answers, with different brands mentioned and different sources cited. One spot check tells you very little. What matters is how often you appear across many runs over time, and only recurring tracking shows you that.

Engines behave differently. A brand that appears in 70% of relevant Perplexity answers can be missing from Claude or Gemini on the same prompts, because each engine draws on different sources and weighs them differently. Running the same prompts everywhere shows you the gaps a single-engine view would hide.

Buyers ask AI at decision moments. Research from Ahrefs shows AI answers substantially reduce clicks to traditional results, so the answer itself is increasingly where the shortlist gets made. If you’re not in the answer when someone asks “best [your category] for [their situation]”, you were never in the running, and your analytics won’t show it either.

Want a quick read on how easily AI can understand your website? Our free AI visibility checker takes about a minute and doesn’t ask for an email.

Prompts are not keywords

If you come from SEO, the easiest mistake to make is treating prompts as long keywords. They behave differently in four ways.

People write them like speech. Nobody asks ChatGPT “accountancy software UK small business”. They ask “what accounting software should a small UK business use if we mostly invoice in euros?” Prompts carry context, constraints and intent. Yours should too.

Engines expand your question behind the scenes. Modern answer engines often rewrite a prompt into several background searches (sometimes called query fan-out) and combine the results. One well-chosen prompt effectively tests your visibility across a cluster of related searches, not a single string.

There’s no search volume to sort by. You can’t look up how many people asked a specific prompt last month. Some AI visibility platforms estimate how often themes come up, which helps with prioritisation, but prompt selection is mostly a judgement about your buyers, not a column sorted high to low.

Journey stage matters more than wording. Two prompts with similar wording can sit at very different points in the buying journey. “What is answer engine optimisation?” and “answer engine optimisation agency for a UK law firm” deserve different priority, different competitor sets and different expectations. If you’d like the academic grounding on how generative engines select and cite sources, the original GEO paper from Princeton researchers is a good read.

The questions to ask before you write a single prompt

This is where prompt sets are won or lost. Work through these before writing anything. We use a version of this list with every client, and the answers do most of the work.

Who is actually asking? Pick two or three real buyer personas and write in their voice. A finance director, a marketing manager and a founder ask about the same product in very different ways. If every prompt sounds like your own team wrote it, that’s the thing to fix.

Where are they in the journey? Awareness prompts (“why isn’t my brand showing up in ChatGPT?”), consideration prompts (“how do I improve AI search visibility?”), comparison prompts (“how do X and Y differ for mid-size firms?”) and purchase prompts (“best AEO agency UK”) each tell you something different. Cover the whole journey, weighted towards the stages where deals are decided.

Would a real person say this? Read each prompt aloud. If it sounds like a keyword with filler words attached, drop it. “AEO agency UK best 2026” isn’t a prompt. “Who are the best AEO agencies in the UK right now?” is.

What’s at stake if you’re missing from this answer? Some prompts are interesting, some are revenue. “What does AEO stand for?” builds category awareness. “Which agency should we hire to fix our AI visibility?” is a deal happening in real time. Track both, but know which is which and weight accordingly.

Who are the competitors for this prompt? Competitors vary by prompt. On a broad category prompt you might be up against global players; on a local or niche prompt, two specialists. Assign competitor sets per topic cluster rather than one blanket list.

Which engines does your audience use? A developer-heavy audience leans towards Claude and Perplexity. A consumer audience leans on ChatGPT, Gemini and Google AI Overviews. Track everywhere you can, but read the results through the lens of where your buyers actually are.

What’s your branded and unbranded balance? Branded prompts (“is [your brand] any good?”) measure reputation and accuracy. Unbranded prompts measure whether you get discovered at all. Most brands over-track branded prompts because the results feel good. Try to resist.

Does geography or language change the question? “Best payroll software” and “best payroll software for UK businesses” produce different answers with different brands in them. If you sell regionally, say so in your prompts.

How many prompts, and how often? Cover your important buyer needs and repeat the same questions. Choose a schedule you can sustain, record the engine and market, and compare patterns across multiple runs.

A tracked prompt taxonomy: the six categories every set needs

Here’s the taxonomy we build every client prompt set around. The examples use a fictional UK project management tool called “Planbird” so you can see the pattern.

Category	Example prompt	What it tells you	Journey stage
Unbranded category	What's the best project management tool for a UK agency of about 20 people?	Whether you get discovered at all, and who wins instead	Consideration to purchase
Comparison	How do Planbird and Monday.com differ for a small creative team?	How engines position you head to head, and what they say tips the balance	Purchase
Problem and solution	My team keeps missing deadlines across client projects, what should we change?	Whether you appear before buyers know the category exists	Awareness
Branded	Is Planbird any good? What are its weaknesses?	Reputation, accuracy of claims, whether engines describe you the way you describe yourself	Evaluation
Long-tail use case	Project management tool that handles retainer work and ad hoc requests in one view?	Visibility on high-intent niches where smaller brands can win	Purchase
Local and geo	Best project management software companies based in the UK?	Whether regional intent works for or against you	Consideration

As a starting point, aim for roughly 60% unbranded and problem-led prompts, 25% comparison and 15% branded. Treat that as a guide rather than a rule. A challenger brand nobody knows yet should lean even harder towards unbranded discovery. An established brand dealing with misinformation might justify more branded tracking. If you’d like a ready-made starter set across all six categories, the free prompt set generator further down builds one from your brand name and what you sell.

How to build your tracked prompt set, step by step

1. Start with real language. Before writing anything, collect how buyers actually talk: sales call notes, support tickets, the questions prospects ask in first meetings, Reddit and industry forums, People Also Ask boxes, your site search logs. Google Search Console queries make useful seeds too, once you translate them back into natural questions.

2. Draft broadly. Build a candidate set across the categories relevant to your business. Involve someone who talks to customers every day, then check the questions against actual sales and support conversations. Tilio’s Prompt universe research maps modelled buyer questions across an industry and examines sampled answers. It can inform the shortlist; it does not measure how often people ask each question.

3. Test them live. Run each candidate in two or three engines. You’re checking two things: does the engine interpret the prompt the way you intended, and does the answer mention brands like yours at all? A prompt that returns generic how-to advice with no brands is educational rather than competitive, and belongs in a different bucket.

4. Score and prioritise. Rate each surviving prompt on relevance to what you actually sell, commercial intent, and winnability (is there a realistic path to appearing in this answer with your available resources?). Prompt volume estimates help where you have them, but don’t let them override your judgement about intent.

5. Balance the set. Apply the category split above, check you’ve covered each journey stage, and make sure no single topic dominates. Fifty prompts about one product line and five about everything else is a blind spot, not a strategy.

6. Assign competitors per cluster. For each topic group, list the two to four competitors you genuinely lose deals to. This is what turns appearance data into share of voice you can act on.

7. Set volume and cadence. Choose enough prompts to cover your priority audiences, products and buying stages. A larger set is useful only when it adds relevant coverage; no prompt count guarantees a representative result. Track daily if your platform allows it, because daily runs are what separate a real shift from ordinary noise. If you’d rather have specialists build and run this for you, explore Tilio’s ongoing AI search programme.

Common mistakes to avoid

Keyword-stuffed prompts nobody would ask. If a prompt reads like it came out of a keyword tool, the data it produces describes a conversation that never happens.

Tracking only branded prompts. Appearing in 95% of answers about your own brand feels great and tells you very little. Discovery happens on unbranded prompts, and that’s where most brands are losing without knowing it.

Too few prompts. With 15 prompts, one unusual answer can swing your visibility score by several points and send everyone chasing a trend that isn’t there. Volume is what makes the data trustworthy.

Set and forget. The questions people ask AI change quickly, new competitors appear in answers, and engines change how they source. A prompt set nobody has reviewed in six months is measuring last year’s market.

Ignoring engine differences. An average across all engines can hide the fact that you’re strong in one and absent in another. Read the results per engine, then prioritise the engines your buyers use.

Chasing prompts you can win but don’t need. A high appearance rate on commercially irrelevant prompts is a vanity metric. Keep the set tied to what you actually sell.

Measuring and iterating

Once your set is live, these are the numbers worth watching:

Appearance rate: the percentage of runs where your brand appears in the answer at all
Share of voice: your appearances relative to your assigned competitors, per topic cluster
Citations: which of your pages (and which third-party pages) engines cite when you do appear
Sentiment and accuracy: how you’re described, and whether it matches reality
Position: whether you’re the lead recommendation or a footnote

A sensible cadence is daily automated runs, a monthly review of movement and citations, and a quarterly refresh where you retire dead prompts and add new ones based on what sales and support are hearing.

Start with a clear baseline: where you appear, where you don’t, and which sources shape the answers. Agree priorities around the gaps that matter to your business, then review share of voice and citations as changes go live. Use that evidence to adjust the work and its pace.

Build your starter set with the free prompt set generator

Enter your brand name and what you sell, add a competitor and location if you have them, and the generator builds a starter set across all six categories. Download it as a CSV or copy the prompts straight into your tracking platform. Treat it as a first draft to edit and add to, not a finished set: swap in your buyers’ real language wherever you can.

Prompt set generator

Enter your details to build a starter set of tracked prompts across all six categories, then download it as a CSV or copy the prompts.

Your brand name

What you sell, as a buyer would say it

A natural noun phrase works best, the way a buyer would describe it, not a keyword.

Main competitor (optional)

Location (optional)

Enter your brand name and what you sell to preview your prompt set. Adding a competitor and location unlocks comparison and local prompts.

The CSV uses standard columns (prompt, category, journey stage) that open cleanly in Excel or Google Sheets and bulk upload into most tracking platforms. Everything runs in your browser, and nothing you type is stored or sent anywhere.

You don’t have to build every prompt by hand, either. Some GEO/AEO platforms, such as Answyn, use AI to generate prompts informed by the context of your brand, then let you refine them within the platform. Answyn also lets you add prompts manually or bulk upload them mapped to columns, with a handy template provided, so a set like the one above is easy to bring across.

Frequently asked questions

How many prompts should I track?

Fifty is a working minimum for a single-product brand, 100 gives most businesses reliable signal, and larger or multi-market brands often need 200 or more. Below about 30, normal answer volatility drowns out real trends.

How often do AI answers change?

Constantly. The same prompt can produce different answers on consecutive days as engines re-source, update and vary their generation. That is why daily tracking across many prompts beats occasional manual checks.

Should I track branded prompts?

Yes, but keep them a minority of your set. Branded prompts tell you whether engines describe you accurately. Unbranded prompts tell you whether you get discovered at all, and discovery is where the growth is.

Do I need different prompts for each engine?

No. Track the same set across every engine, because the comparison is the insight. Add a handful of engine-specific prompts only if one platform matters disproportionately to your audience.

Can I check my AI visibility for free?

Partly. Our free AI visibility checker reviews how easily AI can read your website, with no sign-up required. It does not test live AI answers, so use tracked prompts to see where your brand actually appears.

Start with prompts that tell you the truth

Tracked prompts aren’t glamorous, but everything else in AEO stands on them. Choose them like a buyer, test them like a sceptic, and refresh them as the market moves.

If you’d rather not build this alone, Tilio’s managed programme includes AI search analytics, with an agreed set of buyer questions, answer engines and markets, competitor comparisons, dashboard access and findings to guide your priorities. We establish your baseline during onboarding and use the findings to prioritise content, technical improvements and off-site work.