aeo
Branded vs Unbranded AI Prompts: Which to Track First.
Branded vs unbranded AI prompts are the first split any team building an AI visibility programme has to get right, and most get it backwards. A branded prompt asks an engine about you by name; an unbranded one describes a buyer's problem and lets the engine choose who to recommend. Tracking only the first tells you that ChatGPT knows your company exists. Tracking the second tells you whether it would ever put you in front of a buyer who has not heard of you yet, which is the number that actually matters for pipeline.
What is the difference between branded vs unbranded AI prompts?
A branded prompt names your company directly, something like "Is [Your Company] good for supply chain visibility software" or "How does [Your Company] compare to [Competitor]." An unbranded prompt never names anyone: "What's the best supply chain visibility software for a mid-market manufacturer" or "How do I reduce late shipment penalties with better tracking." The engine has to pick a source on its own, and that pick is built entirely from what it can find and trust off-site.
Branded prompts confirm your reputation with people who already know you. Unbranded prompts test whether you exist at all to people who do not, which is the harder and more valuable test.
Why do unbranded prompts matter more for growth?
Unbranded prompts matter more because they measure new-buyer discovery, the exact moment a prospect who has never heard of your company forms their first impression of who belongs on a shortlist. A branded prompt where the engine praises you accurately is nice to have; an unbranded prompt where the engine never mentions you at all is a lost deal you will never see in your CRM, because the buyer never reached your site to convert.
This is also where most B2B teams have the least visibility today. SourceRank AI baseline audits show the average B2B company is cited in fewer than 5% of relevant AI prompts, and the gap is almost always concentrated in the unbranded set. Companies rarely fail to show up when a buyer asks about them by name. They fail to show up when a buyer has not decided who to ask about yet.
What are competitive prompts and when should you add them?
Competitive prompts name a rival directly and ask the engine to compare, choose, or list alternatives: "[Competitor] vs [Competitor], which is better for enterprise teams" or "top alternatives to [Competitor]." Add this category once your branded and unbranded numbers are established, because a competitive prompt only becomes actionable after you know whether you are missing from the category conversation entirely or losing a direct, named comparison.
Running competitive prompts too early wastes effort. If an engine never mentions your company on unbranded prompts, it has no reason to raise you in a head-to-head comparison either, and fixing that comparison result requires the same off-site authority work as fixing the unbranded gap.
Branded, unbranded, and competitive prompts compared
| Prompt type | What it measures | Example prompt shape | Buyer moment | Priority in a first tracked set |
|---|---|---|---|---|
| Branded | Reputation and accuracy when named directly | "Is [Company] a good fit for X" | Late evaluation, post-shortlist | Low, add after baseline |
| Unbranded | Whether you get discovered at all | "Best [category] for [use case]" | Early research, pre-shortlist | Highest, build the set here first |
| Competitive | Whether you win a named comparison | "[Competitor] vs [Competitor]" | Mid evaluation, active comparison | Medium, add once branded and unbranded are measured |
| Problem-aware | Whether you're found before category is known | "How do I fix [symptom of the problem]" | Earliest research, no category yet | Medium, strongest long-term compounding |
What is the right ratio of branded to unbranded prompts to track?
The right starting ratio for most B2B teams is roughly a quarter branded and three-quarters unbranded and competitive combined, because that mirrors how buyers actually split their research: most questions describe a need, only a minority name a vendor before a shortlist exists. A team that tracks a 50/50 or branded-heavy mix ends up with a report that looks reassuring and misses the discovery gap doing the real damage to pipeline.
This is also where a lot of free or DIY AI visibility checks quietly mislead: a tool that only lets you type in your own brand name is only ever going to run branded prompts, no matter how good its dashboard looks.
How do you build your first tracked prompt mix?
- 1. Write 10 to 15 unbranded prompts that describe your buyer's problem in their own words, at different stages from vague symptom to specific category search
- 2. Write 5 to 10 problem-aware prompts that never mention a category, only a pain point, since these compound fastest with off-site authority work
- 3. Add 5 to 10 competitive prompts naming your two or three real alternatives by name
- 4. Add 5 branded prompts to confirm baseline reputation, and no more, since this category needs the least ongoing attention
- 5. Run the full set across ChatGPT, Perplexity, Gemini, and Copilot on a fixed cadence, because a single engine's result is a sample, not a measurement
That sequencing, roughly 30 to 40 prompts weighted toward unbranded and problem-aware, is close to what a SourceRank AI audit runs as a starting baseline before scaling the set with a client's own sales and search data.
Does the right prompt mix change by engine?
Yes, because ChatGPT, Perplexity, Gemini, and Copilot pull from different sources and behave differently on the same prompt category. Perplexity, which favours fresh, well-structured web content and always shows its sources, tends to reward unbranded and problem-aware prompts fastest once a page is well structured. ChatGPT leans more on training data and established authority, so unbranded gains often take longer to show but hold once they land. Gemini and Copilot sit closer to their respective search indexes, so unbranded results there track more closely with existing SEO authority than the other two engines do.
Running the same prompt mix across all four is still the right approach. The differences show up in how fast each category moves, not in which categories are worth tracking.
Frequently asked questions
What is a branded AI prompt?
A branded AI prompt is a test question that names your company directly, such as "What does [Company] do" or "Is [Company] good for [use case]." It measures reputation and accuracy, not discovery.
What is an unbranded AI prompt?
An unbranded AI prompt describes a buyer's problem or need without naming any company, such as "best software for reducing late shipments." It measures whether an engine discovers and recommends you on its own, which is the harder test and the one tied most closely to new pipeline.
Why do unbranded prompts matter more than branded ones?
Unbranded prompts matter more because they represent the moment a prospect who has never heard of your company forms a first impression. A missed branded prompt loses reputation points with someone who already knows you. A missed unbranded prompt loses a buyer entirely, before they ever reach your website.
What ratio of branded to unbranded prompts should I track?
Most B2B teams should start with roughly 25% branded and competitive prompts and 75% unbranded and problem-aware prompts, since that mix mirrors how buyers actually research before a shortlist exists.
When should I add competitive prompts to my tracked set?
Add competitive prompts once you already have a branded and unbranded baseline. A competitive prompt only tells you something actionable once you know whether the underlying visibility gap is category-wide or specific to a named comparison.
How many prompts do I need before the data means anything?
A starting set of 30 to 40 prompts, weighted toward unbranded and problem-aware categories and run across all four major engines, is enough to establish a reliable baseline. Fewer than that and single-response variance can make a real gap look like noise, or noise look like a trend.
Can a free AI visibility checker test unbranded prompts?
Most free checkers are built to test your own brand name only, which limits them to the branded category by design. That makes them useful for a quick reputation check but blind to the discovery gap, which is usually where the real opportunity sits. A full AEO audit tests the unbranded and competitive categories a free tool cannot.
Does the branded and unbranded split matter differently across ChatGPT, Perplexity, Gemini, and Copilot?
The categories matter the same way across all four engines, but the speed of movement differs. Perplexity tends to reflect unbranded gains fastest because it favours fresh, well-structured content and always cites its sources; ChatGPT, which leans more on training data and established authority, tends to take longer to shift on the same work.
Getting the branded vs unbranded AI prompts split right before you buy or build a tracking programme is the difference between a report that flatters you and one that tells you where the pipeline is actually leaking. If you want a baseline built the right way from the first prompt, see how a SourceRank AI audit works or get in touch to scope one for your category.