How to choose which prompts to track

There's no limit to how many ways people can phrase a question to AI, but your prompt tracking list does have one. The maximum number of prompts you can track depends on your plan. This article helps you narrow down an unlimited number of possible prompts to a focused set that fits your limit and still covers what matters to your brand.

Prompt types to cover

Most accounts default to "best [category] tool" style prompts and stop there. That covers only one slice of how people actually talk to AI. Aim to include all five types below:

Match prompts to the buyer journey

Group prompts by where the buyer is in their decision, so you know what each result is actually telling you:

• Awareness: exploring a problem, no shortlist yet
• Consideration: comparing options, persona-specific (for example, "CRM for freelancers with international clients")
• Purchase: ready to act, mostly ecommerce or local

For a list of around 100 prompts, a workable split looks like this:

• 30 to 40 awareness prompts
• 40 to 50 consideration prompts
• 10 to 20 brand-evaluation prompts, kept separate from category prompts
• A purchase-stage slice only if you're an e-commerce or local

Scale the proportions to your own limit — the ratios matter more than the absolute numbers.

Where to find prompt ideas

Use a few of these sourcing methods together rather than relying on just one:

• Convert existing keywords: export your top non-branded keywords and rewrite them as natural questions. For example, "CRM small business" becomes "What's the best CRM for a small business?"
• Mine People Also Ask and AI Overviews: these are already phrased as questions and mirror how people prompt AI.
• Ask an AI model directly: "What questions do people ask when researching [topic]?" gives you a fast conversational list.
• Check Reddit, Quora, and forums: real complaints and requests in people's own words closely match how AI expects to be prompted.
• Look at competitor paid keywords: if a competitor is bidding on a 3+ word term, that topic likely converts.
• Reverse-engineer a website's structure: look at how a company segments its customers and which features solve which problems. Each segment or feature can become a prompt.
• Check Insights & recommendations: this section of AI Results Tracker surfaces prompts where your brand already shows up in AI answers but isn't tracked yet, so you're not guessing. See AI Results Tracker: Insights & recommendations for details.

How suggested prompts are generated

When you set up a project, you can either add prompts manually or pick them from a list of suggested prompts. Knowing where those suggestions come from tells you how far to trust them.

The suggestions are generated by an AI model at setup time, from three inputs and nothing else:

• your project's domain;
• the brand name you entered;
• your selected language.

The model first works out what business categories your brand operates in, using only the brand name and the website, and then generates a set of prompts for each of those categories. There is no separate industry or category field for you to fill in — the categories are inferred.

When to trust the suggestions: they are a reasonable starting point when your website makes your category obvious and your brand name is distinctive. Treat them as a first draft — run them through the four checks in Filter your list down to your limit below rather than tracking them as they come.

When to write prompts yourself: if your site does not explain your niche clearly, if you serve a segment narrower than your category, or if your brand name is ambiguous, the inferred categories will be off and the suggestions will be generic. In those cases the sourcing methods above will beat the suggested list. The same applies to anything the model cannot know from your website — new products, a shift in positioning, or the specific personas you are chasing this quarter.

How your brand configuration affects suggestions and mention detection

The brand name you configure does two jobs, and both are worth getting right at setup:

• It shapes your prompt suggestions, together with your domain, as described above.
• It is what mention detection looks for. Whether an AI answer counts as mentioning you is decided by matching your configured brand name against the mentions extracted from that answer.

Matching ignores case and tolerates a straight or curly apostrophe, but it is otherwise a literal match on the name you entered. There is no alias or synonym list to configure, so the name has to be the form that actually appears in AI answers. "ION" and "ION Group" are two different strings as far as detection is concerned.

Practical guidance:

• Use the form AI actually writes. Run two or three of your own prompts, read the answers, and enter the brand exactly as it appears there — not necessarily your legal entity name.
• If your brand name is a common word or a personal name, expect the number to be noisy in both directions: unrelated uses of the word can register as mentions, while genuine references phrased differently will not. In that situation lean on link presence and the Sources report rather than reading mention presence on its own, and consider prompts that pair the brand with your category.
• If several brands are stored on the project, the reported mention position is calculated against the first one, so the brand you report on must be first in the list.

If mentions look wrong across every prompt, a brand-name mismatch is the first thing to check — see AI Results Tracker FAQ.

Group prompts by topic

Once you've gathered candidate prompts, organize them into groups. You can group by:

• topic or content pillar
• goal, for example, prompts where you want a brand mention versus prompts where you want to be cited as a source

Groups make it easier to:

• filter your data by topic in the "Rankings" tab
• spot which topics are covered and which have gaps
• manage a large prompt set without losing track of what each prompt is for

You can create and organize groups from prompt management or while setting up your project.

Filter your list down to your limit

Once you've got more ideas than you need, run each candidate through these four checks. If it fails one, cut it or narrow it.

1. Competitive relevance: would this prompt realistically surface your brand or a direct competitor? If it's too broad or off-topic, drop it.
2. Influenceability: look at what sources AI currently cites for this prompt. If it's dominated by sites you can't realistically outrank (government, Wikipedia, major publishers), it's not a good use of a tracking slot.
3. Business alignment: every prompt should map to a real product, persona, use case, or funnel stage. If it doesn't tie back to something you'd act on, cut it.
4. Scope: narrow anything too broad by adding specifics: industry, team size, budget, or location.
   • "Best CRM software" → "Best CRM for UK-based startups under 20 people"
   • "Best email marketing tool" → "Best email marketing tool for Shopify stores under 10k subscribers"