Blog / Guides

Branded vs non-branded prompts: how to track both

Lars Koole · · 7 min read

A single AI visibility score that blends prompts naming your brand with prompts that do not will flatter you. Here is how to tell them apart, split them and read each on its own.

A laptop screen showing analytics charts on a desk in an office
Summarize with

Most AI visibility reports start with one number: the share of tracked prompts where your brand appears. If the prompt set includes questions that already contain your brand name, the number says more about how the set was written than about how visible you are.

This guide explains the two kinds of prompt, why they behave so differently, and how to choose and read them separately.

The two definitions

A branded prompt contains the name of a brand. Usually it is your own, but it can be a competitor's. Examples:

  • "Is Acme good for small sales teams?"
  • "Acme vs Beta: which one should I pick?"
  • "Acme alternatives"
  • "Does Acme integrate with Slack?"

A non-branded prompt describes a need, a problem or a category without naming any brand. Examples:

  • "Best CRM for a five-person sales team"
  • "How do I stop losing track of leads in spreadsheets?"
  • "What should I look for in a CRM for a small business?"

Remove the brand names and see whether the question still makes sense. A branded prompt falls apart. A non-branded prompt does not.

Why branded prompts nearly always mention you

When a prompt asks about your brand, the engine is being told to talk about you. The answer will almost always name you, because the question requires it. The only open questions are what the engine says and whether it is accurate.

That makes branded prompts useful for one job: checking reputation. You can see how the engine describes your product, whether it gets the facts right, and which sources it uses when it talks about you. Those are worth knowing. But a mention in answer to "Is Acme good for X?" is not evidence that anyone will discover Acme.

Counting those mentions as visibility is circular: the prompt supplied the name.

Why non-branded prompts are where discovery happens

A buyer who does not know your brand cannot type its name. They ask for a category, a use case or a comparison of approaches, and the engine decides which brands to put on the shortlist. If you are not named in those answers, you are absent from the part of the buying process where options are formed.

This is also where the visibility gap shows up. Non-branded prompts are the only honest measure of it, which is why they are the core of any tracking plan.

A real example of the gap

AI Tools Bakery ran Pineprompt on their own site while writing their guide to AI search monitoring tools. Across 12 tracked prompts, overall visibility came out at 33%. Their four branded prompts scored between 91% and 100%. Of their eight non-branded prompts, seven scored 0% and one scored 9%, an average of around 1%. In their words, branded prompts were carrying the headline number.

The blended score averaged a segment where mentions are close to guaranteed with a segment where they are earned, and hid the one that mattered.

How to build a prompt set

Treat the following as a starting point, not a rule. The right mix depends on how well known your brand is and how much of your market is still choosing a category.

Mix. Make non-branded prompts the majority of the set, since they carry the discovery signal. A smaller group of branded prompts is enough to monitor reputation and comparisons. A set that is mostly branded will look healthy and tell you little. If your brand is new and unknown, lean further toward non-branded prompts; if you are an established name with many comparison queries, you may want more branded ones.

Source the prompts from real language. Invented prompts tend to sound like marketing copy. Real questions are better:

  • Sales calls: the questions prospects ask before they have heard your pitch, and the alternatives they say they are comparing.
  • Support questions: how customers describe the problem in their own words, including the words you would not use.
  • Search Console queries: the phrases people already use to reach your site. Question-shaped queries convert easily into prompts, and branded queries show which branded prompts are worth tracking.

Our post on what prompt fan-out is explains why engines expand one question into several searches, which is a good reason to cover a topic with a few differently worded prompts rather than one.

One country and language per set. Answers differ by market, and brands named in one country may not appear in another. Keep each set to a single country and language so the results are comparable. If you sell in several markets, build a set for each rather than mixing them.

How to read the results separately

Report the two segments side by side and never average them.

  • Visibility. The share of runs in which your brand is named. For branded prompts, expect a high figure; a low one is a red flag about accuracy or recognition. For non-branded prompts, this is your real discovery rate.
  • Position. Where you appear in the answer. Read it from non-branded answers only. In a branded answer you are usually first because the question is about you. Our glossary entry on position in answer covers the measure.
  • Share of voice. Your mentions divided by all brand mentions in the segment. In non-branded prompts it shows how you compare with competitors. See how to measure share of voice in AI search for the method.

Read each prompt over several days rather than from a single answer, because answers vary from run to run, and read results per engine as well as per segment. The difference between brand mentions and citations applies in both segments: check whether you are named, and separately whether your pages are cited.

Common mistakes

Tracking only branded prompts. The dashboard looks good because the prompts guarantee it. You learn nothing about discovery, and you miss the moment a competitor starts to take your category.

Judging position from branded answers. Being listed first in an answer to "Acme vs Beta" is not a ranking win. Position matters only where the engine chose the order itself.

Changing prompts too often. A trend line needs a stable set. If you swap prompts every few weeks, a rise or fall may only reflect the new wording. Make a change deliberately, note the date, and treat the data before and after as two series. Add new prompts rather than rewriting old ones.

Branded and non-branded prompts in Pineprompt

Every new Pineprompt project starts with a set of tags that includes branded and non-branded, and you can attach them to prompts. The tag filter on the overview narrows the numbers to the prompts carrying that tag, for example non-branded prompts only. On the prompt list, the same filter narrows which prompts are shown. The analytics available through the API accept the same tag filter.

Pineprompt tracks ChatGPT, Gemini, Claude, Google AI Overviews and Google AI Mode. Tag your prompts when you add them and read the two segments separately from the start. The glossary has short definitions of branded vs non-branded prompts and AI visibility if you want to share them with your team.

Summary

Branded prompts tell you how engines describe you. Non-branded prompts tell you whether buyers will find you. Keep them in separate segments, build the non-branded set from real customer language, hold the set steady, and report each segment on its own.

If you want to see where you stand on the prompts that matter, get started with Pineprompt.

Written by

Lars Koole

Co-founder & Head of Engineering

Lars co-founded Pineprompt and leads its engineering. He writes about the mechanics of AI search: how platforms retrieve, rank, and attribute the sources behind their answers.

Keep reading

Track your brand across every AI platform.

Pineprompt monitors five live AI platforms daily, in the country and language your buyers use. See our methodology for what we capture and how.